Raw Data By P3 Adaptive
The Invention of Skippable YouTube Ads, w/ Shishir Mehrotra
Co-Founder and CEO, CodaListen Now:
Shishir is as ahead of the technology curve as it gets, some of his ideas have revolutionized the way that tech giants like Microsoft, Google, and YouTube operate. Now, he’s innovating again as the founder and CEO of Coda-an amazing integrated system that centers around creating Docs that are as powerful and actionable as Apps. He’s also one of the most down to Earth human beings we’ve ever had the pleasure of sitting down with!
References in this episode:
- 2:20 – Shishir’s data path intersects with Rob’s and the stories abound, Shishir passes on working for Google before it was Google
- 15:25 – Shishir has a random idea about advertising that eventually forms into some common advertising practices, Google woos Shishir back, and he ends up running YouTube!
- 27:25 – The value of a Computer Science degree is….debatable, an interesting definition and example of AI, and Nouns VS Verbs in naming products and features
- 41:00 – How Coda was formed and the amazing innovation that Coda is-it makes a doc as powerful as an app, and the importance of integration
Rob Collie (00:00:00): Hello, friends. Today's guest is Shishir Mehrotra, and let me tell you, Shishir is a ringer of a guest. We met back at Microsoft in the 2000s where he was already entrusted with some pretty amazing responsibilities and was doing very, very well in those roles. About the same time that I left Microsoft to start P3, Shishir left Microsoft to go ... Oh, that's right ... Run YouTube. And he was at the helm of YouTube during what he calls the hyper-growth years where YouTube really exploded and became the thing that we know it is today. During this conversation, I discovered that it certainly sounds like he invented something about YouTube that we absolutely take for granted today and has been seen by billions, used probably billions of times per day. That wasn't enough for him, so he left YouTube after a number of years and started a new company called Coda.
Rob Collie (00:00:55): And Coda is an incredibly ambitious product. You could say that in some sense, it's aimed at being a Microsoft Office replacement, but even that isn't quite right. It's in a little bit different niche than that. And, of course, we explored that in our conversation. We talk about his billion dollar mistake, quite possibly, literally, billion dollar mistake, not many people can make those. I was thrilled to discover that he and I have basically exactly the same philosophy about nouns and verbs in software. We talk about the antiquated notion that a computer science degree is somehow super important in product management roles, even at software companies. And just, in general, I couldn't get enough of it. He was super gracious to give us his time for this show, and I hope you enjoy it as much as we did. So, let's get into it.
Announcer (00:01:42): Ladies and gentlemen, may I have your attention please.
Announcer (00:01:48): This is the Raw Data by P3 Adaptive podcast with your host Rob Collie and your cohost Thomas Larock. Find out what the experts at P3 Adaptive can do for your business. Just go to p3adaptive.com. Raw Data by P3 Adaptive is data with the human element.
Rob Collie (00:02:11): Welcome to the show. Shishir Mehrotra, how are you today?
Shishir Mehrotra (00:02:15): Oh, I'm great.
Rob Collie (00:02:16): Are you coming to us from Silicon Valley?
Shishir Mehrotra (00:02:17): I am. Well, south of California. Been in my house and in this spot for about the last year.
Rob Collie (00:02:23): When did we meet?
Shishir Mehrotra (00:02:24): You were working on Excel and I think at the time I was working on WinFS, the early days of Microsoft.
Rob Collie (00:02:31): Oh, WinFS. Just completely unexpected sidelight. It was like 1998 or maybe 1999, we're in a review with Jim [Allchin 00:02:42] and all of his lieutenants. And the whole point of this meeting is to assassinate the technology I was working on. This was an arranged hit on MSI ...
Shishir Mehrotra (00:02:54): [crosstalk 00:02:54]. On MSI.
Rob Collie (00:02:55): ... On the Windows Installer, right?
Shishir Mehrotra (00:02:56): Okay. Yeah. Yeah.
Rob Collie (00:02:57): And there are factions in this room that have had their knives, they've been sharpening them and they've arranged this moment so they can kill us. And, at one point, one of the complaints about us was our heavy use of the registry. Just poisoning the registry. Do you remember a guy named Rob [Short 00:03:15]?
Shishir Mehrotra (00:03:15): Yeah, of course.
Rob Collie (00:03:16): I really liked Rob Short. I thought he was awesome. He was a tough guy, but also really fair and funny and friendly at the same time. And he's been sitting in this meeting for hours because he has to, and he's just totally tuned out. Of course he would be, right? It's not about him. And then, this mention of the registry as an attack on us comes up and Jim Allchin immediately whirls around to Rob and goes, "Now you see, this is what I'm talking about. Our storage system is such a piece of shit." And he starts ripping it to Rob and Rob's having to wake up from his trance. It's like suddenly the guns can swing so fast in those meetings.
Shishir Mehrotra (00:03:55): I mean, that was a use case that Bill and Jim and so on all tried to push on WinFS, but it was one we actively resisted. It's a hard one.
Rob Collie (00:04:02): It is. The worst thing in the world is to have state stored in multiple places that have to go together with each other. Right? That just turns out to be one of the hardest problems.
Shishir Mehrotra (00:04:11): It's such a critical element of the operating system. And you end up with all sorts of other issues of what can run on what and ...
Rob Collie (00:04:17): And it's funny. The registry was basically my introduction to the entire Win32 platform. When I was running the installer, that's all I knew about. I knew about the type library registrations and the registry. I knew it in class IDs. And I could follow those things. I could follow that rabbit's trail from one place to another without ever really understanding what a class ID was. Right? It was just the registration of an object, right?
Shishir Mehrotra (00:04:40): Right.
Rob Collie (00:04:40): I didn't learn that until years later. So funny. But then we crossed paths again. Right?
Shishir Mehrotra (00:04:45): SQL.
Rob Collie (00:04:46): I remember how it happened. Ariel [Nets 00:04:49] came into my office and said, "Hey, there's someone important who's going to need some information from you." And I go, "Okay." And he said something like, "He's a real rising star here, so make sure you give him everything he needs." And I'm like, "Okay."
Shishir Mehrotra (00:05:05): I don't think I know this half of the story. Okay.
Rob Collie (00:05:09): And I think you were somehow involved with the potential acquisition that was going on at the time. Is that true?
Shishir Mehrotra (00:05:14): You talking about in-memory BI?
Rob Collie (00:05:16): Yeah.
Shishir Mehrotra (00:05:16): Yeah. I was at the time ... Maybe for your listeners. So, my history, after WinFS folded and collapsed, and you can talk about that if you'd like, I ended up being unexpectedly merged into the SQL Server division. I ended up running what Microsoft called the program management team or SQL Server. And it was super interesting for me because I was never really a database guy. Everything I had worked on to that point was fairly end user-centered, infrastructure in the background. And I was surrounded by these people that really love databases. Actually, as a side note, I fell in love with databases because of Paul Flessner. Paul was on his way out. He was retiring that year and he had one last ... At the time we used to call them strategy days so that Bill and Steve and so on would post this annual review.
Shishir Mehrotra (00:06:01): And Paul Flessner, he decided this was going to be his last hurrah strategy, "I'm going to tell these people exactly what I think." He's in the middle of preparing for this and WinFS is folding up and he says, "While you're figuring out what you're going to do next, why don't you come help me write the strategy days presentation?" And he was really drawn to the idea of someone that actually wasn't in his organization doing it because I could speak my mind about whatever and I had no bias walking into it. And probably from his perspective, I would write whatever the hell he wanted and make it sound good. This guy, he's a database legend. He drove the Sybase acquisition that turned into SQL Server. And so, he had a list of ideas for how to think about the database market that many of which were pretty ascetical.
Shishir Mehrotra (00:06:44): And he spoke in very plain language when he's ... Actually, interestingly, he's retired. [inaudible 00:06:48] his woodworking. That's his thing. He builds chairs and tables are amazing. You can go buy them. As opposed to many techie database guys, he speaks in very plain language.
Rob Collie (00:06:55): I love that.
Shishir Mehrotra (00:06:56): And you just walk through like, "Here's how to think about the different workloads and here's what's happening in the industry and here's what's happening in data warehousing." Which wasn't really a term at the time and data warehousing was just emerging. And then, at the end of that process, we had a pretty successful strategy days and he said, "Why don't you run the PM team and help my new guy?" Ted Kummert came in to go and run SQL Server after Paul. And that's how I ended up in that spot. And as part of that, I ended up covering a lot of ... One of Paul's last statement was, "Data warehousing is not the same thing. Go do something different." And that's where people like Ariel and Amir and so on, that whole division, Tom ... And there was a bunch of people running that at that time ... Came into play.
Shishir Mehrotra (00:07:34): And then they had this idea that ... There's a lot of different things to know about SQL Server. SQL Server is not actually well-built for data warehouse and so most databases are not. And at the time, the raining wisdom was you needed a completely different architecture for business intelligence, which I guess we called OLAP back then. I don't know if that term is still used.
Rob Collie (00:07:54): Yeah. Oh, we still do. We just hide it. It's a dirty word.
Shishir Mehrotra (00:07:57): Yeah. For the geeky folks out there, and the key difference being that instead of storing things row by row, you store things column by column and you also precalculate aggregate. So, you have some sense of what, I guess, nowadays called the cube. These things are likely to be great for, "We're going to precalculate the sum of orders for customers by region or whatever it might be." And then, Ariel and his brother Amir had this idea and they said, "Hey, we've got this strategic advantage at Microsoft, which is we own the front end and the backend of this architecture. On the backend, we need to be able to scale better and we need to move to column storage and do all this fancy stuff with cubes. But if you ask anybody where all of their analysis actually gets done, what do they say?
Shishir Mehrotra (00:08:38): There's 1,000 reporting tools out there but everybody lives in Excel. And so, they said, "What if we were to find a creative way to pull these together? And I think at the time you were running this part of the Excel platform. And so, I was sent in to go figure out how to make this pitch. I mean, these guys really wanted to do an acquisition space and so on. And I was sent in to try to make the pitch. And, actually, the insight there was interesting. Amir came up with this chart, which I'm not really sure where it came from but he basically went and looked at the size of cubes of OLAP instances across a wide set of customers, including all of Microsoft. He pulled all of these different ones and he figured out that the biggest cube at Microsoft was this thing called MS Sales.
Shishir Mehrotra (00:09:20): It was all the customer data from Microsoft if you remember well. And he said, "If you compress this down with column storage, I'm going to get the numbers wrong." But it fit inside tens of megabytes of storage, which was previously much, much larger if you did as row storage. And he said, "This is so small that it can fit in memory on a client, which was unheard of. Usually, the whole idea behind these systems was you have to query a server. The server is really big. At that time, a lot of systems go up and scaled out. There's often very big hardware back there as well. And he said, "Hey, I bet we could move to a model where the primary way that people do this analysis actually happens in that place where they actually want to do their work in Excel. So, I think that's where the other half of that conversation from my side was coming from.
Rob Collie (00:10:06): Yeah. So, like you said, with Paul Flessner bringing you into right part of the strategy days stuff, Amir was, at that point in time, still using me in the same way. I had come over from the Excel world and so he was trotting me out every time he wanted someone to talk about Excel in a way that he couldn't be criticized. I was just almost the unfrozen caveman lawyer from Saturday Night Live, this Forrest Gump figure, "Listen, I don't know much, but I do know Excel and I know the people." You know?
Shishir Mehrotra (00:10:32): Yeah. Yeah.
Rob Collie (00:10:33): Usually, because on the SQL side of the house, you couldn't argue with me about Excel. If I go back to the Excel world, they'd all argue with me but on the sequel side, I was unquestioned. So, Ariel was right, he said, "This guy is a mover and shaker. He's going places." And then, an eye blink later, you're at YouTube. When did you end up at YouTube?
Shishir Mehrotra (00:10:53): So, there's a personal story arc that goes along with this. I started a company out of school called [Sintrata 00:10:58]. It was an early version of what became AWS, Azure, so on, to utility computing. There's a whole generation company that started back in that '99, 2000 period. All of us were seven to 10 years too early. There was no virtualization, no containers and none of the underlying technology that actually made the cloud take off existed yet. As that was wrapping up is how I got to Microsoft but in that period, Sintrata was funded by this famous venture firm called Kleiner Perkins.
Shishir Mehrotra (00:11:23): My primary investor was a [inaudible 00:11:24]. [inaudible 00:11:25] as Sintrata was wrapping up, he had suggested, "Why don't you go join another client or company?" And I said, "Which one?" And he said, "Well, you can look at all of them but the one that's really hot right now is these two Stanford guys are creating this new search engines called Google. Might want to check it out." And so this is back in 2002. And so, went over and spent some time with Larry and Sergey. And at the time, they hadn't hired a single outside product manager. And so, they wanted me to come in and start the product management team there. And, interestingly, I turned them down. My wife likes to call my billion dollar mistake. And instead I got drawn to Microsoft.
Shishir Mehrotra (00:12:01): As I got drawn to Microsoft, it's related to this story because I had an old boss of mine, I was an intern at Microsoft when I was in college, and he was starting this new thing called Gideon that was in the Office team actually. And the project would turn Office into a front end for business applications. So, it's had a lot of relevance to what ended up happening in that space.
Rob Collie (00:12:18): Who was running Gideon? Who was that?
Shishir Mehrotra (00:12:20): Satya was our skip-level boss and this was much, much earlier in his career. And the guy actually running the project was a guy named John [Lacada 00:12:27]. I think he's gone now. I don't know where he is. Yeah.
Rob Collie (00:12:29): I worked with John quite a bit over the years [crosstalk 00:12:32]. And this is how you know Danny Simmons. Right?
Shishir Mehrotra (00:12:34): That's right. Danny was part of that team.
Rob Collie (00:12:36): Oh my gosh! Yeah.
Shishir Mehrotra (00:12:36): Yeah. Danny was on that team. I ended up working with Danny multiple times. Mike Hewitt was the one who was my intern manager who pulled me over to the project. Actually, as a fun version of fate or whatever, Mike now works at Coda. [crosstalk 00:12:48]-
Rob Collie (00:12:48): Does he really?
Shishir Mehrotra (00:12:49): Yeah, he's an engineer here. He's great. He lives in Idaho. Once we really started hiring distributed, I finally managed to pull him into Coda. So, I turned on Google in that period and they didn't let up. Basically, every year they would call and say, "Hey, we got something down here for you." Gideon actually didn't have a very positive outcome. I showed up to work on this thing and nine months later, Sinofsky killed it. Given the priorities Office had at the time, it made reasonable sense, but it was my first education of big company politics and that's how I ended up working at WinFS.
Rob Collie (00:13:20): Sinofsky has delivered many such educations of big company politics.
Shishir Mehrotra (00:13:24): Yes. Yes. For sure. For sure.
Rob Collie (00:13:26): One of his primary contributions. Yes.
Shishir Mehrotra (00:13:27): So, are you reading his history of Microsoft [inaudible 00:13:30]?
Rob Collie (00:13:30): I haven't been but now I will be.
Shishir Mehrotra (00:13:32): Oh, you should. It's good. Steve and I didn't always see eye to eye on everything, but his sense of history is really good. I don't know how the hell he remembered so much stuff, but he's basically publishing a new thing every few days, I think, maybe every week, and it's really good.
Rob Collie (00:13:44): I both loved Stephen and was terrified of him at the same time.
Shishir Mehrotra (00:13:48): It's common.
Rob Collie (00:13:49): Yeah.
Shishir Mehrotra (00:13:50): So, I'm working on SQL Server, but the reason all that matters is I was committed to Seattle. I had convinced my, at the time, fiance now wife, to move up to Seattle. She's a physician. So, she was doing her residency at Children's Seattle. And I convinced her to stay and do her fellowship and that all ran out. So, my clock ran out on Seattle. Said, "All right, now we're ready to move." And we had presumed we were going to move to the Bay area. So, it was just implied at the time, if you're going to be a techie, you got to move down to the Bay Area at some point. And I thought I was going to start another company. I was ready to do it again but Jonathan Rosenberg, the guy at Google who ended up running product there, he called me, he said, "Oh, if you're thinking about coming back, why don't you just come meet a few people?"
Shishir Mehrotra (00:14:28): And I said, "No, I've been doing the big company thing for a while. I don't think I want to do that anymore." And he said, "No, no, no, no. Google is not that big a company." This is 2007, 2008. And he said, "Google is not that big of a company. Come just meet a few people and nothing else and have some good conversations." And so, I went down, met a bunch of people and this was Larry and Sergey but also Vic Gundotra was there then and Andy Rubin had just joined. And there was a bunch of ... That era of Google was being formed. And I end up, at the end of the day, in Jonathan's office and I tell him, "That was really entertaining, but it feels like a big company. I don't think this is for me." Jonathan's a pretty crass person. I won't use the same language he used but he said, "Oh, that's really effing stupid."
Shishir Mehrotra (00:15:06): And I said, "Why?" And he said, "Well, look, and I'll just give you a really simple reason. All those people, they probably talk to you about Android and Chrome and all this other stuff but what they forget is that, at the heart, Google sells advertising and all the money in advertising goes to television. And nobody even watches those stupid ads." This may sound dumb, but maybe not to this group. I didn't know that. For me, I'd never bought or sold an ad in my life. And the idea that all of the money and advertising goes to television was news to me. And I got on a plane after work back to Seattle. I do a lot of my thinking on planes for weird reasons. You may be the same. I don't know.
Shishir Mehrotra (00:15:40): But I get on the plane, I take out this little sheet of paper and this was a week after the Super Bowl, February of 2008, the Giants had just beaten the Pats in this epic Super Bowl. And I take out the sheet of paper, I write at the top, how come advertising doesn't feel like a Super Bowl every day? And the basic thing I was thinking about was we had our friends over for Super Bowl and while we're watching the game, the ad would come on, if somebody missed it, I would have to rewind for people to watch the ad again. It's like, "Oh, people actually like the ads in this one day of the year. What's different?" And so, I take out this sheet of paper, I end up writing this little position paper on what I think is wrong with advertising, without knowing really anything about advertising. Get home, it's pretty late. My wife's not up to tell me it was all stupid.
Shishir Mehrotra (00:16:19): And then I wake up the next morning and I write to Jonathan. I say, "Hey, look, I really enjoyed the time. I don't think Google's for me, but I had some thoughts on something you said that stuck with me about why advertising sucks. And I'm sure you guys are already thinking about it, but I'm happy to send it to you if you'd like." And he's pretty early morning guy and so he read it and said, "Actually, nobody's thinking about this. Maybe you should come and I'll give you a small team and you can start running this." There were three ideas in the paper but the most simple one was how come ads don't have a skip button on them? And then, if you skip the ad, why don't you make it so that if you skip the ad, the advertiser doesn't pay?
Shishir Mehrotra (00:16:50): You change all the incentives of advertising so that if the ads aren't good, then nobody gets paid if the ads are going to get better. And we're going to reset the balance and that's why it's going to feel like Super Bowl every day. He was like, "There's a lot of reward and be creative on the Super Bowl." So, J.R. convinced me. He's like, "Come down. Run this project." When I tell the story, it sounds eerily similar to how I ended up at Microsoft, like, "Oh, come run this small project." And it was this group of people, again, that misunderstood what ... This project was at the time called Mosaic.
Shishir Mehrotra (00:17:19): It became a product called Google TV. Chromecast, Google TV, Google Home, all comes out of that same group now. So, I showed up to work on that and very quickly in that process realized that this had actually very poor corporate sponsorship as well. In this case, Larry and Sergey thought this product was really, really dumb. I should have known as I was going through the interview process. And so, I told J.R. and I was excited about the project and I said, "Hey, maybe I should talk to Larry and Sergey about that, a bunch of ideas and other stuff if I met them." He's like, "Oh yeah, they're traveling this week." I was like, "Really? Okay." And every time I asked, he was avoiding me talking to them about the project. But, anyway, so I show up to work on that and it's very long story out, but this paper leads to me working on this project.
Shishir Mehrotra (00:17:57): And then, just, basically, we decided to merge the project into YouTube. And back in 2008, to a very side door, end up initially running the monetization team and eventually running the rest of the team for YouTube and then spending six years there and growing that business, which was ... At the time, when I joined YouTube, it was the weird stepchild of Google. It was generally thought of as the first bad acquisition that Google made. Until then we had this string of amazing acquisitions led to Maps and Android and all this stuff. YouTube was a weird one, right? It was the, we lost hundreds of millions of dollars a year. It was dogs on skateboards. We had a billion dollar lawsuit from Viacom.
Rob Collie (00:18:35): Mark Cuban famously said it's never going to go anywhere.
Shishir Mehrotra (00:18:38): I have very fun stories with Mark Cuban. It was two years after I left YouTube where he finally wrote me and said, "Actually, I think you might've been right." He was quite convinced we were wrong about it. But, anyway, so I ended up working on YouTube. I'd never bought or sold an ad in my life, knew nothing about video and an infrastructure guy in the previous career, and ended up working on YouTube for six years.
Rob Collie (00:19:02): It's a really interesting thing, right? Sometimes not knowing a lot about an industry or a topic is actually fantastic because you don't bring all the baggage and all the preconceptions. Of course, you can't just go all in on that. If you never know anything about anything, you're just someone wandering around the world with a loud voice. And so, getting the right balance between knowing what you should know and not knowing the things that will throw you off, if we could get that mix right at all times in our lives, we'd be in great shape, but it's tricky, isn't it?
Shishir Mehrotra (00:19:32): You've roughly described my career. Almost every job I took was in a space I knew nothing about. And it's a very positive interpretation of this person who has to learn every piece of this. But yeah, I think a beginner's eye allows you to look at a space a little bit differently and it certainly worked out at YouTube. And we were walking the trends of the video industry in every way, how we thought about content, how we thought about monetization, and what is good content? What is not good content? Our views on these things were diametrically opposite of every assumption that had been made by every experienced person in that industry. I think we turned out to be more right than wrong.
Rob Collie (00:20:07): Oh my gosh! Yeah. Now, a few things jumped out at me from that story. First of all, if we think about it with the perfection of hindsight, the clarity of hindsight, basically, Google ran this really sick reverse auction for your services where they like, "If you come here now we'll pay you a billion dollars." And you're like, "Hmm, no." Right? And then-
Shishir Mehrotra (00:20:30): It wasn't obvious that it was going to be a billion dollars.
Rob Collie (00:20:30): I know. Then they call you back a year later and they say, "Okay. Fine. How about 100 million?" And you're like, "Hmm, no." And they finally got it down low enough for you to take the job. I've never met anybody who has a story where you can even joke about a billion dollar mistake. So, I'll never have the opportunity to recruit you, but if I did, now I know how.
Shishir Mehrotra (00:20:56): [crosstalk 00:20:56] blowing your offer. That's right. That's right.
Rob Collie (00:20:59): And it's got to include the words, just come run this small, little team.
Shishir Mehrotra (00:21:03): Yeah. Yeah. I get drawn to projects. I don't get drawn to the rest of it. So far it's worked out okay. But yeah, I get drawn to ideas. I mean, this is really only the fourth company I've ever worked for yet every transition was drawn by some idea that I couldn't stop thinking about.
Rob Collie (00:21:17): That idea or position statement, is that in some way, at the beginning, the origin story of the skip button for ads?
Shishir Mehrotra (00:21:27): Oh yeah. I mean, the skip button for ads it's now called TrueView. Back to your point on beginner's mind. So, I show up, I've got this idea around the skip button and actually it makes more sense for YouTube than it does for this Google TV thing that we were working on. So, there's totally reasonable outcome. I show up and my first meeting with the sales team, I'm maybe six weeks in, the head of sales, Susie, she says, "Can you come give a talk to sales team and just tell a little bit about your vision for YouTube." And we had a nice ... And I said, "Look, I don't think this is a good idea. I don't know anything about this part of the industry. So, I'm going to make a fool of myself." And she's, "No, no, no. You have got all these great ideas and they're fresh and different and why don't you come talk to them?"
Shishir Mehrotra (00:22:04): And I go talk about a bunch of different ideas, and I talk about this one about skip buttons on ads. And one of the salespeople, who I've since become very good friends with, she raises her hand and she says, "Wait, I don't understand. Do you want none of us to make any money?" They thought this was the dumbest idea on the planet. You put a skip button on ads, people are going to hit the skip button. It's like that's what obviously is going to happen. And, basically, the entire sales force rejected this idea. And it took me three years to ship that feature because every person in the sales force thought it was such a dumb idea. I would get told, "You can come talk at the sales conference, but you're not allowed to talk about your stupid skip button idea. You have to talk about everything else."
Shishir Mehrotra (00:22:43): And what turned out was ... This is actually another fun story in great product managers. I don't know if you still think of yourself as a PM, but I consider you to be a really strong product manager as well. But this is a story about a guy, Lane Shackleton, who actually now runs product at Coda. So, Lane was a sales guy. He was actually our primary sales guy at YouTube. And he really wanted to be a PM. And at the time, we had this really stupid policy where you weren't allowed to be a product manager at Google unless you had a CS degree. It was just part of the early, early viewpoint the founders had.
Rob Collie (00:23:17): So relevant.
Shishir Mehrotra (00:23:21): Right. So, you commiserate with this a lot. So, Lane comes to me and says, "I want to be a PM. How do I do it?" And I said, "Hey, look, I mean, I love you and I think you could do a great job but I've got this policy. And I got to make a really strong case if I'm going to get over the policy." And he said, "How about I just do it on the side? Do it as a trial run." He gave me an idea. I said, "Okay, I'll make a deal with you. I'll let you try to be a PM, but you have to do it in your 20% time. And not in your 80-20% time, but you got to do a great job of your sales job and then you do this part. And the second criteria is you take whatever project I give you."
Shishir Mehrotra (00:23:52): And he said, "All right, deal. What's the project?" I said, "Okay, I want you to work on this thing called skippable ads." And I said, "Look, the sales team thinks it's really dumb because the way that the division work, the engineering leader was like, "I'm not allocating stuff that the sales team thinks is dumb. And so, I can give you one engineer who is a new grad and that's it." But I have a playbook for you. I think you need to go and you just go talk to the AdWords team and get this thing out of the buying experience and then work on this with the analytics and figure just these couple pieces out. And we'll be able to ship this thing and we'll slowly build up the business. It'll be fine."
Shishir Mehrotra (00:24:23): And so, he goes away and he comes back a couple of weeks later for his update. And I said, "Oh, how's it going? Did you talk to the AdWords team?" And he said, "No, actually, I decided that's not the problem here." And I said, "What do you mean? That was your job. Go talk to those different people." And he says, "Well, I've been thinking about it and I think the real problem here is the name is wrong." I was like, "The name? What are you talking about? We'll name this thing later. This is not that important." And he says, "No, no, I think the problem is that skippable ad is a value proposition to an end user but who buys advertising? The advertiser buys advertising. Skippable is actually a really poor value proposition to the advertiser. Why would I want my ad to be skipped? Right? And so, the reason you're hearing so much negative reaction if people don't understand why it's helpful to the advertiser."
Shishir Mehrotra (00:25:06): And so then he came up with this idea and said, "Why don't we name it TrueView?" And I'm skipping a whole bunch of parts in the story, but we call it TrueView. That's what the ad for one is actually called. You have no idea what ads are called, right? Oh, there's ads on Google. Nobody knows [crosstalk 00:25:18] from AdWord.
Rob Collie (00:25:18): Yeah. It's not a feature. Yeah.
Shishir Mehrotra (00:25:19): But what's a sponsored story? And you don't know any of that stuff. You just know it's an ad. And he said, "So, let's focus on the advertising." Came up with this name TrueView. And the idea is very simple is you only pay per true views. You don't pay for the junk, you only pay for the real ones. Right? And all of a sudden this thing went from being, I'm not allowed to talk about it at sales conferences to the number one thing on the entire sales force [inaudible 00:25:42] all of Google. Beyond anything the average team was working on.
Shishir Mehrotra (00:25:45): And it was such a simple idea. And, by the way, the way the math works is very simple, it's most people do skip the ad. It's about 80% skip rates on those ads. So, four out of five times you see an ad, you probably have a skip button, but it turns out that the 20% of the time you don't is such high signal and so effective an ad that you can often charge something like 20 times as much for that view. And so, what you end up with is you end up with you just take that math and say, [inaudible 00:26:09] four times better monetization with a skippable ad than without a skippable ad.
Shishir Mehrotra (00:26:13): It was not obvious that advertisers would be willing to pay that much more if they know you actually watched the ad but when you start ... But this is a good example, again, a beginner's mind and, Lane, I mean, this is one of his ... So, I've managed to convince the calibration committees and so on and turned to a product manager and turned into a great product manager. He joined me early on at Coda and now runs the product and design team here. Great example of coming fresh to a new problem.
Rob Collie (00:26:36): Yeah. Well, if only he'd had a computer science degree, that idea would have been so much smarter. You know?
Shishir Mehrotra (00:26:43): Yeah. The crazy part, this is one of the most technical guys I know and he's like, "I don't understand. I write this stuff on the side. Why do I need a stupid degree for that?" Right?
Rob Collie (00:26:53): I know. There was one time in my first three years at Microsoft where I used one piece of my computer science education, one time. I used O notation to prove that we shouldn't do it a certain way. And when I got my way after using O notation, it's like, "This is an O of N squared algorithm." I got to run around the hallways chanting, like, "Whoa, look, my education, it worked. It worked. It worked." And that was the only time I ever used any of that. So, no, that's a silly policy.
Shishir Mehrotra (00:27:25): Yeah. It was funny, when I was going to college, my parents were both computer scientists and I was one of those kids who grew up with a ... I never knew what I wanted to be. One week I was going to be a lawyer, then I was going to be a doctor, then I was going to be a scary period for my mom where I really wanted to be a taxi driver. I went through all the different periods. And then, I'm filling out my college applications and it says like, "What do you want to major in?" And I said, "Oh, I think I'll write down CS." I was into computers at the time and so I write down CS. And my dad says, "If you major in CS, I'm not paying for college." What are you talking about? I thought you'd be really excited.
Shishir Mehrotra (00:27:57): That's what you guys do. My dad now runs supercomputing for NASA. I thought this would be pretty exciting for you. And he says, "No, no, no. This is a practitioner's degree. I'm not paying for college unless you major in something where the books are at least 50 years old." And that was the policy. And so, I ended up majoring in math and computer science. And from his perspective, he paid for a math degree and I happened to get the CS degree for free. But his view was that ... Which is true ... Computer science changes so fundamentally every 10 years.
Shishir Mehrotra (00:28:22): And my classes the professors often taught out of the book that they're about to publish. The book wasn't even published yet and they're like, "Oh, here's the new way to think about operating systems." And it was totally different than what it was five years ago. I think there's a lot of knowledge in CS degrees but I actually think ... O notation is an example. I used to teach that class at school. That's math. That's not CS.
Rob Collie (00:28:42): I know. Yeah. Yeah.
Shishir Mehrotra (00:28:44): It's a very good way to think about isotonic functions but the actual CS knowledge is all but relevant by the time you graduate.
Rob Collie (00:28:51): One thing that you said to me about your time at YouTube that stuck with me years, years, years, years later is that here we are at the tip of the spear, the head of this giant organization and YouTube eventually became giant, and with all this amazing machine learning and just so much algorithmic, not even complexity, but also just we don't even know what it's doing anymore. It's so sophisticated that we can't even explain why it's making these decisions but they're doing well, and yet every day we get together, we're looking at simple pivot tables and there's these knobs on the sides of these giant algorithmic machines that some human being has to set to, like, "Should we set it to six or seven?" And it's just this judgment call. And I just love that. That was, in a weird way, so reassuring to me that even at the absolute top of the pyramid of the algorithmic world, there's still a need for this other stuff.
Shishir Mehrotra (00:29:43): The most fun example of this, backing for a moment, my dad, back to the story of me going into CS. At one point I had asked him, what is artificial intelligence? And he said, "Well, artificial intelligence is this really hard to describe field." I asked, "Why is that?" And he said, "Well, because it's got this characteristic that the moment something works, it's no longer AI." And so, AI is what's left is all the stuff that doesn't work. And so, you can use all these examples of when you have all regressions, it's like, "That's just math. That's not AI. We understand how it works." My favorite example with the kids is when you drive up to the traffic light, how does it know when to turn red and green and so on?
Shishir Mehrotra (00:30:19): Oh, there's a sensor there. It just senses the cars there and so then it decides to turn red or green. That's not AI. I know how that works. I can describe it. It's a sensor. And so, we went through, I think, decades of time where the moment something worked, it stopped getting called AI. And then, some point, 10, 15 years ago, I'd say 10, we flipped it. And now, all of a sudden, anything that does math is AI. And it's amazing to me that we would look at some of these systems and it was literally a simple regression and we say, "Oh, that's machine learning." And it became very invoked. I think about it that way.
Shishir Mehrotra (00:30:53): I mean, there are some really complicated machine learning techniques and the way our neural network works, which is the heart of how most of these machine learning techniques work is very complicated, but at the heart of what it's doing, it's approximation function for a multi-variable phenomenon. So, the most fun example I can tell you about your observation there is this project called DALS. DALS was an acronym for Dynamic Ad Load System where at the time, on YouTube, the rate at which we showed ads was contractually set. We would go negotiate with the creator and say, "Oh, ESPN, we want your content on YouTube." And we would say, "Look, our policy is we show ads every seven minutes." And they say, "No, our content is so good. We want it every two minutes."
Shishir Mehrotra (00:31:32): And then, the Disney folks would have their own number. And so, there is this long line of contractual stuff baked into our ad serving logic that's like, "Oh, it's been two minutes. You have to show an ad." Because they all just thought they knew better of how good their content was. And so, one of the engineers had this idea and said, "This is dumb." We know our intentions are well aligned. Almost all our deals were rev share deals. We made money when the creator made money. And we know whether or not this is a good time to show an ad or not, why don't we turn this into a machine learning system and guess whether or not we should show an ad? So, it's called Dynamic Ad Load System. DALS was its acronym. So, the team goes off and this engineer goes off and builds this thing.
Shishir Mehrotra (00:32:08): Lexi was his name. So, Lexi builds this thing and he brings it to one of our staff reviews. Every Friday, we had this meeting of IT staff. That's where we went through all the major stats for the business and including any major experiments that are running. If he brings something in and he says, "All right, before we launch this thing, I'd like to know what our trade-off function." The trade-off function in this case is, how much watch time are you willing to trade off for revenue? These are two primary metrics. At every moment we're going to decide, should we show an ad or not? And we have to make a guess at, "We think if we don't show an ad you'll watch for this much longer, if we do show an ad, there's a chance you'll leave but we'll make this much money. So, what's the number? How much should we trade off?"
Shishir Mehrotra (00:32:45): This is a very typical question I would get in this forum. It's impossible to answer, how much would you trade off? Watch time, revenue. And so, I came up with a number and I put a slope on this chart and we decided two for one. I can't remember whether it was two points of watch time for one point of revenue. But whichever way it was, I do a slope and we got a lot of reaction. They're like, "Okay. Great." And they ran away from the room. "Okay. We have a number. We can go do our thing." And so, they come back a few weeks later and say that we're ready to launch. And I said, "Okay, so did you hit the number?" And they said, "Well, actually, we have some interesting news for you. Turns out in our first tuning of the system, we actually have a tuning that is positive on both watch time and revenue. And somehow by redeploying the system, we make more money and people watch longer."
Shishir Mehrotra (00:33:25): And I said, "Really? How does that happen?" And they said, "Well, we don't really know yet, but can we ship because clearly better than your ratio?" And I said, "Well, okay, you can ship but next week I want you to come back and tell me why." And so, next week they come back and I said, "Do you know why?" And they said, "Well, we don't know why, but we have another tuning and it's even better on both watch time and revenue. I was thinking we ship this one."
Shishir Mehrotra (00:33:47): I was like, "Okay, but please come back next week." This went on for four weeks. Right? So every week they would come back and they'd say, "Okay, we got this thing. It's even better on both. And we still have no idea why." And, finally, they figured out why. And it turns out that basically what was happening was the system was learning to push ads later in people's sessions. If you watch YouTube for a while, early on, you'll see very little advertising. But if you sit there and watch for hours and hours and hours, the ad frequency will gradually increase with a viewpoint of, this person's not going anywhere. They're committed, which makes intuitive sense, but it wasn't an input that we handed the system.
Shishir Mehrotra (00:34:18): And how did we figure that out? The pivot table. I notice that the ... What did we do? We went and charted everything we could out of the experiment group and in our experiment group and we just guessed at what is the way to figure out why is this happening? Because it's not a signal that we were intentionally giving the system, it's just the system got every other signal it could. And we looked at everything. I mean, is it geography? Is it tied to content? Is it age? Is it ... How is it possible that we're showing more ads and people are watching for longer? That story is a lesson in a number of different things. I mean, I think it was a great lesson in how when people think about machine learning systems, they miss this element of ... Any machine learning system is just a function.
Shishir Mehrotra (00:34:54): All the ML system does is take a very large set of inputs, apply a function to it and generate an output. Generally, that output is a decision, show an ad, don't show an ad. Self-driving car turn right or turn left. It's some decisions of, is this image a person or an animal? And that system is trained and is trained on a bunch of data. And at some point, somebody, usually fairly low in an organization, makes the tuning decision and says, "I'm willing to accept this much being wrong for this much being right." Generally called precision recall. More layman's term for it is you figured out your false positive rate versus your false negative rate for whatever system you're trying to figure out. But somebody has to make a decision.
Shishir Mehrotra (00:35:30): It's usually three tunings, very deep in the system. And then, after that point, the system is unexplainable. You have no idea how this thing works. And so, what do you do then? You go look at a bunch of empirical data of what's happening and try to figure out, "What did I just do? I've got this thing and what's actually happening here?" And you try to figure out, is it doing what you actually want it to do? And all of that is done in fancy pivot tables.
Rob Collie (00:35:53): Yeah. It's so funny, the AI, and you've said before, your dad, as soon as it reaches a equilibrium, it's not AI anymore.
Shishir Mehrotra (00:36:00): Right. Not anymore.
Rob Collie (00:36:02): Now though, it seems like it's a funny thing that you built these systems that then figure things out and they seem to be working great but then they can't turn around and explain to you what they're doing. It's not built to explain. It's just built to do.
Shishir Mehrotra (00:36:15): It makes some sense how the human brain works. Why did you do that? I don't know. I just did it. And when you're running a business, that's not an acceptable answer. I need to know why did it go that way instead of ... Why did it turn right? I need to know why. So, you end up with this interesting tuning and then you're constantly looking at charts of output, what is going on here? To try to figure out whether it's working the way you want it.
Rob Collie (00:36:34): So, while we're on pivot tables for a moment, go back to your story about skippable ads. This is TrueView. Imagine how much better off we would be as a society if pivot tables had originally been named summary tables.
Shishir Mehrotra (00:36:50): Oh man.
Rob Collie (00:36:51): You know? That one was blown.
Shishir Mehrotra (00:36:53): Yeah.
Rob Collie (00:36:54): I actually tried to rename it stupidly. I mean, it was too late. It was way too late. And I fought that battle for way too long. It was a fool's errand to try to rename something that had been in the world for that long but what does it mean to pivot data? No one knows.
Shishir Mehrotra (00:37:08): It's now the insider's club handshake.
Rob Collie (00:37:12): I know. I know. I think we probably lost half of the people who would have used them just in the name.
Shishir Mehrotra (00:37:17): It's interesting you say that because the way we do the equivalence in Coda, we don't use the term pivot at all. We call it grouping. We don't even call it a thing. Right? We don't give it a noun name. We give it a verb name. And it just turns out that grouping a table is a very understandable phenomenon. In Coda, our model of grouping doesn't require aggregates also turns out ... And the reason I don't love the word summary is I actually think most commonly what you want to do is you take a set of records and you say, "I've got a bunch of tasks. Let me sort them by in progress and done." And I still want to be able to see the tasks. And one of the things pivot table, I think screwed up, is that you can't see the tasks anymore. The moment you're in that world ...
Rob Collie (00:37:55): Yeah. I agree. But given what was built, the pivot table implementation, right? Summary would have been the killer name, right?
Shishir Mehrotra (00:38:02): Would have been a much better name. What would you have named VLOOKUP to?
Rob Collie (00:38:05): Oh, I don't know. Pivots is still relevant to me, VLOOKUP not as much. No. But like Bill Gates always pressing for the unification of grouping in Excel with pivots. And we were always like, "Hmm, no." And it became a running joke after a while, he'd be like, "To the extent that you guys on Excel ever do anything that I ask you." That would be his preamble to some of the things he would say to us.
Shishir Mehrotra (00:38:33): I mean, I would say, nowadays, people use pivot for lots of things, but for our first year for the customer journey, our grouping feature was definitely the top of the list. And, honestly, there's a bunch of people who, like you said, never really understood pivot tables and could never compare the two, like, "Oh, that makes total sense to me. I drew up a table. That makes total sense." Then two, to show aggregates.
Rob Collie (00:38:50): The way you zeroed in on noun versus verb, that actually has come up multiple, multiple times on this show. It's one of my things. My new hires, when they'd come to work for me on the Excel team, I would sit them down and say, "Listen, you are not allowed to introduce nouns into this product. If you want a new noun, you've got to come to me. You got to fight me for it. You can verb all you want." That was hard one knowledge. I was a noun guy coming out of computer science school. Computer science people love them some nouns. Entities. Just say the word entity and you get all gooey inside, but no, it's a verb world.
Shishir Mehrotra (00:39:26): I make that specific statement, you can ask my team, all the time. You're going to add a new noun, you got to come through me. I mean, on YouTube, it was interesting because YouTube has three primary nouns, video, channel and playlist. And we spent forever ... For a long time video was the only noun that mattered. And it was a big debate over which one matters more, channel or playlist. And I made the team pick. You got to pick one. We picked channel, which is probably obvious. Playlists are these long forgotten feature of YouTube and channels are now a big deal. But that wasn't always true. Channels actually used to be a very small deal on YouTube. If you go back to what I do in 2008, yeah, you would publish a video, it's like, "A channel, whatever." Totally bit my fingers on this channel, but it has nothing else on it.
Shishir Mehrotra (00:40:05): And, nowadays, all people care about on YouTube is like, "This is my channel. How many subscribers I have." And the same way with Coda, we've put a lot of energy into as few nouns as possible. We'd use common language for nouns, only brand the ones that you really, really, really want to brand. Because there are very few branded nouns in Coda. There's lots of incentives in product development that lead to it. In a lot of companies, you get promoted on it. Like, "I invented this thing. It's now Power BI. And it's now this pivot thing." And you get a lot of feedback loop because nouns are distinguishable but it doesn't help your customers.
Rob Collie (00:40:37): Even the technology under the hood is screaming at you, "Noun me. Noun me." It's like, I've got this really cool data structure here. It's dying to be surfaced in the ... No, no, don't do that. That's not what we do. We do not surface the technology. That's not what we're here for, but it's a powerful instinct. Really powerful. Okay. So, Coda, that's the next chapter. And that's the next place where we crossed paths. So, I actually realized that it was six years ago. I visited you in the Valley six years ago. And the reason I know it was six years ago is because one of the people who was there in the early days with you, the very beginning.
Shishir Mehrotra (00:41:18): They're all still here, but yeah.
Rob Collie (00:41:19): Okay. Good. So, got the feel that they will be long-timers. Yeah. It was a tight bunch. It was a tight crew. The two of you were joking to them, "Maybe we should go to Burning Man this year." And I was sitting there thinking to myself, I had been invited that year to a friend's bachelor party. He was going to Burning Man. And I didn't even speak up because I was so terrified of going. I wasn't even sure if I was going to go.
Shishir Mehrotra (00:41:42): Did you go?
Rob Collie (00:41:42): I did. And that was 2015. So, that's how I know. It was also, I think, the first year that the Warriors had blown up down the NBA scene. So, we were sitting and watching the Warriors annihilate people after we talked. So, six years ago, you were pretty deep into this thing that's now called Coda. It was codename something else at the time that I kept getting wrong. Was it Krypton?
Shishir Mehrotra (00:42:03): Krypton. That's right.
Rob Collie (00:42:05): But I kept calling it Vulcan.
Shishir Mehrotra (00:42:08): The team had such a laugh out of that.
Rob Collie (00:42:12): I kept forgetting it was Krypton and calling it Vulcan. So, why don't you explain both to me and to our listeners what the original vision was and how and if that's evolved over time.
Shishir Mehrotra (00:42:25): By the way that meeting was, hey, super entertaining. Rob came in and described this as Vulcan as been repeated many times in the story. But it also was super informative because you came and gave a bunch of perspective. I think probably one of the most relevant to our last discussion, one of your most interesting observations that stuck with the team was you described this person and you said, "Hey, I can walk into a room and if I ask them just a couple of questions I can split the room into two groups of people very quickly." You used to call it the data gene. And your questions were, do you know what a VLOOKUP is or do you know what a VLOOKUP is? What a pivot table is? Bad for many of the reasons we just talked about, but for the perspective of understanding how humans are evolving and so on, it was actually quite insightful that these people you just can't keep them away. They will eventually figure these things out.
Shishir Mehrotra (00:43:11): And if you have that data gene, you will some point in your life intersect with these things and figure out what they are. The Coda founding story, so I was at YouTube and an old friend of mine, [inaudible 00:43:21] Alex DeNeui, now my co-founder at Coda, he and I have known each other for 20 plus years. We went to college together. And he's part of the founding team at Sintrata as well. Interestingly, we've worked every other job together, which is a fun pattern. So, he had started this company that got acquired by Google and he had just quit. And he was starting a new company and he'd come to me and he said, "Hey, my company's not doing that well. I'm thinking about pivoting to do something different. Can you help me brainstorm a new set of ideas?" So, we started brainstorming mostly about what he should do.
Shishir Mehrotra (00:43:49): I was still relatively happy at Google, but I had told him, "If you pick something interesting, I'd be happy to invest or advise or help out in some way." Said this long list of ideas and we started brainstorming and at one point, one of us writes this sentence on the whiteboard, what if anyone can build a doc as powerful as an app? And that sentence ended up becoming the rallying cry for what became Krypton and then Coda. It's a very simple statement but it comes out of two primary observations. One is, I think the world runs on docs not apps. That if you go ask any team how they operate, any business, company person, so on, if you ask them how they operate, they'll immediately rattle off all the different packet software they use. "Oh, we use this thing for CRM and this thing for inventory. And we use this thing for pass tracking and so on."
Shishir Mehrotra (00:44:32): And then, if you just sit behind them and watch them work for a day, what do they do all day? They're in documents, spreadsheets, presentations, and some communication tool. That's what they live in. And this first observation was one that was very deeply embedded in us because that's how we ran YouTube. I mean, YouTube, amongst other things was born right in the start of the Google Docs generation. I got the YouTube 2008, Google Docs is just coming out and, as I mentioned, we were the forgotten stepchild of Google, so we were allowed to do whatever we wanted but we could get no help in doing it. And so, we decided, for example, we would run our task management goal-setting process. We didn't like how OKRs worked.
Shishir Mehrotra (00:45:09): I actually just published a whole paper on this last week. You can take a look. But we didn't like how OKRs worked. We wanted to do a different way. And so, how do we do it? We do in a big spreadsheet? I ran compensation differently at YouTube. I had this philosophy I call level independent compensation and the Google HR team allowed us to do it, but said, "We're building zero software for it." So, we did it in a network of documents and spreadsheets. One of the most fun example is if you hit flag on a YouTube video, for years, a flag on a YouTube video would show up as a row in a spreadsheet [inaudible 00:45:37] the person's desk. That's how we ran all these systems. We used to get made fun of. People are like, "Oh, look at these people. They're duct taping together documents and spreadsheets to run what became a multi-billion dollar division." I used to say like, "I actually think this is our strategic strength."
Shishir Mehrotra (00:45:49): I mean, the reason we can plan so nimbly, the reason I can hire whoever I want, the reason we can adjust our flagging and approval system so quickly is because we didn't purchase some big bulky software to do it, we design it ourselves and turned it into something that then actually met our, at the time, current value system. So, this is observation number one, it's the world runs on docs not app, which is, by the way, not obvious to people but I feel fairly strongly about it. The second observation is that those documents surfaces haven't fundamentally changed in almost 50 years. The running joke at the company is that if Austin Powers popped out of his freezing chamber, he wouldn't know what clothes to wear or what music to listen to, but he could work a document, a spreadsheet, and a presentation just as well as anybody else could. Because everything we're looking at is metaphors that were created by the same people who created WordStar, Harvard Graphics and VisiCalc.
Shishir Mehrotra (00:46:39): And we still have almost the exact same metaphor, which just seems crazy to me. In that same period of time, every other piece of software stack is totally different. An operating system from the '70s versus Android and iOS is unrecognizable. Databases, which we thought were pretty fundamental are completely different than they used to be. Things like search engine, social networks, none of these things even existed and yet the way that slide decks are put together, the way you navigate the spreadsheet grid and the way you think about pages and document is exactly the same as it was in the 1970s.
Shishir Mehrotra (00:47:10): So, you take the two observations, you stick them together and you say, "Hey, we [inaudible 00:47:13] runs on these docs, not applications." And those surfaces haven't changed in almost 50 years. Something's broken. What if we started from scratch and built an entirely new type of doc based on this observation that what we are actually doing with our docs is a lot closer to what we're doing with applications than not? That was the thesis we started with. I got personally obsessed with it. I couldn't stop thinking about it. And this went from, hey, let me invest, let me help, to I quit Google and went and started but at the time with Krypton and then eventually became Coda.
Rob Collie (00:47:44): I'm sure he recruited you at some point by saying, "How about you just come run this small team over here?"
Shishir Mehrotra (00:47:48): Yeah. Exactly. Exactly. That's right.
Rob Collie (00:47:51): Those are the magic words.
Shishir Mehrotra (00:47:52): We won't pay you at all. That's the ...
Rob Collie (00:47:54): Something silly that occurred to me is that your Austin Powers metaphor might even be more accurate than you realize. We are now farther away, in terms of time, from the premiere of that '70s show, than that '70s show was from the time it represented.
Shishir Mehrotra (00:48:09): I like that. Yeah.
Rob Collie (00:48:11): It's crazy. We passed that point six months ago. So, when did Austin Powers the first one come out? Sometime in the '90s?
Shishir Mehrotra (00:48:17): Yeah.
Rob Collie (00:48:18): Right? And it represented a time probably 35 years before it? Probably 1964, maybe 1999. Right?
Shishir Mehrotra (00:48:25): Yeah.
Rob Collie (00:48:25): So, we're almost reaching the point where we're close to the Austin Powers movie as Austin Powers was to the time. So, clearly, if we rewind 35 years, we are what? We're in the '80s, right?
Shishir Mehrotra (00:48:35): Yeah.
Rob Collie (00:48:36): You're right our documents basically look like that.
Shishir Mehrotra (00:48:39): Yeah. You and I can probably geek out on this. And I get asked a lot about why did that happen? Lots of industries saw a change. And the database industry is a great example, you wouldn't expect the database industry to change that much. Codd wrote his book in the 1970s that's still the book that every database engineer you can find will have the book up on the shelf for Codd's relational databases, and yet things like OLAP came out and cubes and it turned into a Power BI. I think what happened in the document industry ... Well, two things.
Shishir Mehrotra (00:49:05): One, every company that wanted to innovate in that space was a platform company whose primary interest was evangelizing a platform. Microsoft didn't really want to displace Lotus and so on with a new thing, they just wanted people to use Windows. It was very important that it actually be backwards compatible with everything at Soft. The other thing that happened is we live through what I think of as a period where we're beholden to file format. And so, one of my favorite examples is Steve Jobs and Apple. I've met a bunch of people that worked on the early iWork suite. And the iWork suite, Jobs came in with a bunch of new ideas. He's like, "This is dumb. We shouldn't have a spreadsheet that's one big universal grid. We should have a bunch of separate grids that are actually a little closer to tables."
Shishir Mehrotra (00:49:45): And so, that's how numbers worked, actually, it's not actually one universal grid, it's a bunch of separate ones. And the way he did it with pages was a little bit different. And then, Keynote, which is probably the most popular of the three is actually different from PowerPoint in those really critical ways and none of the three took off. And why didn't they take off? I mean, Jobs was pretty smart and [inaudible 00:50:02] were pretty good. I think it was really simple reason. If I build something in numbers and then I want to send it to you, I have to assume that you have a copy of numbers and that you run on a Mac and that's not a safe assumption. It hasn't really been a safe assumption for a long time. And then, Google Docs came out.
Rob Collie (00:50:16): Which, by the way, is fundamentally what YouTube did for video. Right?
Shishir Mehrotra (00:50:19): That's right.
Rob Collie (00:50:19): I had all these delivery ...
Shishir Mehrotra (00:50:21): Plugin.
Rob Collie (00:50:21): ... And Coda and pl ... I couldn't send you a video, trust that you'd be able to watch it.
Shishir Mehrotra (00:50:27): And assume you could play. That's right. That's right. I mean, in that case, it was hard to send the videos because-
Rob Collie (00:50:32): Yeah. There was a file size problem and there was also a software compatibility problem behind it.
Shishir Mehrotra (00:50:37): Yeah. I think but in the documents phase, it's very interesting. I think Google Docs is mostly credited for teaching us the value of real-time collaboration, which I think is amazing and in retrospect, obvious innovation. And I think we were both still working at Microsoft when it first came out and all of us were super skeptical like, "Why would anybody want to co-edit a spreadsheet? That seems weird. What if they make a mistake?" And it just turns out it actually works just fine. And people avoid each other well and versioning systems can handle things and you don't need to have a bunch of merge conflicts and all that stuff just works. But, actually, I think the thing that Google Docs did for the industry is they taught us that there's no such thing as a file format. And if I send you a link to a Google Doc, you don't need to worry about whether or not anybody has anything installed and it's all just you could open it and use it.
Shishir Mehrotra (00:51:23): And I think that unshackled a whole industry that had been stuck in, you could call it the fax machine world where you just had to produce things that the other side had a receiver for. And so, I think that moment of opening ... I mean, in retrospect, we started crypto in 2014 and I think we felt it then but I don't think we quite understood how dramatic a change that was. I mean, I think right now our industry is going through probably the most fundamental rethink of productivity that we've seen in 40 or 50 years. And Coda, I think, obviously, doing super well and used by 25,000 teams all over the world and it's spreading very quickly, but there's now all these other players as well that are taking their own approaches at reinventing different parts of our activity.
Shishir Mehrotra (00:52:02): And for an industry that's seen very little change over that period of time, it's amazing how fas ... It's the renaissance of productivity. It's all these new things. I think it is mostly credited to, there's no more file formats. And that totally changes how you think about all these tools.
Rob Collie (00:52:16): A punctuated equilibrium theory of technology. Right?
Shishir Mehrotra (00:52:19): Right.
Rob Collie (00:52:20): Long periods of stagnation and then lots of diversity all at once. You know?
Shishir Mehrotra (00:52:23): Yeah. Something quick cages. Yeah. That's right.
Rob Collie (00:52:26): So, I almost needed a name for a new emotion when I visited you guys. We need a word for this where you guys were telling me about this grand plan reinventing, essentially, what it means to be a productivity suite and documents and all of that. On the one hand I'm sitting there going like, "Oh, come on. This is crazy what you guys are trying to do. This is never going to work. There's so much inertia in the market. How many things have tried to be the new Excel? Everything that's tried that is no longer here." And I was pretty clear about that with you guys. I wasn't hiding my feelings on that front, but at the same time though, and this is important, I was reading the room.
Rob Collie (00:53:05): I was reading you and going, "Yeah, I don't want to bet against these people though." I was really at a weird spot. So, 25,000 teams. Such an incredibly ambitious and gutsy mission, but six years later, you're not doing anything different. You're only growing. You've got people all over the place now. I basically met it was essentially the whole team back then.
Shishir Mehrotra (00:53:31): Yeah. You did.
Rob Collie (00:53:31): I think I've seen the news wire releases of valuations and things, which are always nuts. How has the product evolved? Is it still on that original mission? Is it everything? Is it everything that Office does?
Shishir Mehrotra (00:53:47): Yeah. So, it is still on that original mission. And I gave a talk to new hires. Now, we do it every month. We have a big enough set of people coming in every month. I always start with this big disclaimer. I say, "Sometimes you join a company and what they're doing six or seven years in is completely different than when they started." I generally walk people through our very first pitch deck, our very first [inaudible 00:54:06] and I say, "Look, we're doing almost exactly the same thing. You can attribute that to amazing impressions or amazing stubbornness, whichever way you want to think about it. But, yes, we're roughly doing the same thing as back then. Most customers would describe our product, they call it an all-in-one doc and say, "Oh, it blends the best parts of documents, spreadsheets, presentations, and applications into a new surface.
Shishir Mehrotra (00:54:27): Our deep believers will use that line that Alex and I started with and say, "Oh, Coda allows anyone to make a doc as powerful as an app." One of the things that I think wasn't obvious then that I think has become much more obvious now is initially people will use that frame. If you want to know why my wife is starting a new business, she runs the whole thing in Coda. If you asked her why, she said, "Oh, because I don't want to pick documents, spreadsheet presentation, I guess." It's like one of the most terrifying moments a user hits is like, "I'm about to do this thing. And I've got a new doc dialogue, which one is it? Am I ever going to have to write more than a few paragraphs of texts? Oh, I guess, I got to be over here. Am I ever going to need a formula ever? Oh I guess I got to be over here. Am I ever going to need to present this to anybody? I guess I need to be over here."
Shishir Mehrotra (00:55:09): And that just sucks. Right? It's a really hard way to think about it. That resonates very quickly with people. But I think when people really fall in love with Coda, they fall in love with all the choices we made on the core building blocks of the product. And I think one of the things that was hard to describe at the time, and I knew I had you come to talk to the team as one of the foremost spreadsheet experts I know inside and out, as a deep user, inventor, so on. And so, obviously, the frame of the discussion started with the spreadsheet. The interesting thing about Coda and our relation to spreadsheets is in some ways we're like a spreadsheet turned inside out, right?
Shishir Mehrotra (00:55:42): It's like the core of Coda is not a grid. And so, a lot of the things you're used to in spreadsheets don't exist, but there is a deep re-calc engine sitting underneath Coda. And you can go in the Coda and anywhere you want, you can hit equals and type a formula and you get back a result and you can refer to things. We have a different thing we call tables, which back to our nouns you point at the world, is very importantly not a grid. It's columns have names, not letters. And there is a feature in Excel now called tables that is most people don't know about but you know very, very well. It's amazing how they squirreled it ... I don't know if that happened before you left or not.
Rob Collie (00:56:15): Yeah. So, the original version it was called Lists and that was done in 2003. And then we upgraded it into the full tables feature in 2007. That happened on my team. A guy named Joe [Cheralove 00:56:28] reported to me, was in charge of it. I was just a clipboard holder for that release, in a way, right? I can't claim responsibility for the table feature. Joe and company did that. It's cool.
Shishir Mehrotra (00:56:39): It's a great example. It's a cool feature and people deep in Excel, they discover this thing. And then, the people that really believe in Excel table to say, "Actually, I would prefer that you never use the grid. Please just do this because this is a better way to think about it." And so, there's a bunch of choices we've made that work along those lines that people just fall deeply in love with. Interestingly, the hardest audience for us is the deep spreadsheet lover, but that's actually ... If you are that person that is building models and the difference between INDEX-MATCH and VLOOKUP, then you're probably actually going to have the hardest time getting into Coda, which is interesting, but we completely opened up the market. And so, now, all of a sudden, the metaphors make tons of sense. And people come into Coda.
Shishir Mehrotra (00:57:19): The vast majority of Coda docs are just docs. We have this notion we call pages which, interestingly, they're tabs, spreadsheet tabs turned on the side. We just took a metaphor and we pulled it out, made it more approachable. So, the vast majority of Coda docs have no formulas, no data, no anything. They're just better documents. And then, when people discover tables in Coda, they discover it with a fresh eye and they don't even think about spreadsheets and the people that do, think about it with a completely different metaphor and then, gradually, discover that, "Oh, I can do these things." And actually it feels a lot more like an application and I can add ... Some of our most popular features are buttons and controls and we have these things we call packs that allow you to integrate Coda with all these different services and these automations. And you can build anything in Coda.
Shishir Mehrotra (00:57:56): We have many companies that run entirely on Coda. One really fun example, in the Coda gallery you'll find lots of example docs that people publish. There's one from a company called Squared Away. I think they're 150-person company and they wrote this doc about how they saved a $100,000 in software by rebuilding everything in Coda. And not only saved a bunch of money, but actually built a set of things that actually run the way they want to run. And they run their billing system in it, and how they do their virtual agent service, and they run their time-tracking system in it, their CRM, everything ends up in Coda. So, it's an interesting product that I like to call it a product with a very low floor and very high ceiling that you can come in. It's a blinking cursor, a blank screen.
Shishir Mehrotra (00:58:35): If you know how to use Google Doc, you know how to use Coda and most people, that's what they do. We're just a fancy document that has pages and a few delightful things that allow your documents to look a little bit better. And then, gradually, you learn a new set of building blocks and work your way up into this ecosystem. One of the reasons I think your initial impressions of, will you be able to replace Excel? And I think we don't try. I never walk in and say like, "Please get rid of spreadsheets." We will often get rid of a spreadsheet. The Pinterest team is a good example. Pinterest runs almost entirely on Coda now and the thing we replaced there, we started with Ben and his team started using Coda to run their meetings. Just a better way to run meeting notes.
Shishir Mehrotra (00:59:10): And we have this special type of button we call voting. And so, you can easily set up an agenda for a meeting and everybody votes on what they want to talk about or you can do these very simple, little, we call it pulse check where you can say, "How's everybody feeling about this decision?" And you can click a little button and hide everybody else's responses and just see your own. And then you can click it again and you can see everybody else's. And these are all such simple, little building blocks to go create. And then, they were about to transition from ... They do all their planning at Pinterest in these big spreadsheets like many companies do. And they were about to deploy this big package application for doing it. And the night before, they were doing the evaluation in the big package application, many people said, "I bet I can do that in Coda."
Shishir Mehrotra (00:59:48): And she went and built that same system they were about to buy and spend millions of dollars on, she built it in Coda and said, "Oh, this is actually pretty good." And the goal-tracking system is a very natural ... Table metaphors make much more sense than a grid metaphor. And you probably spend a lot of time trying to teach people, "Don't use the grid, use tables. You'll end up in a much better spot if you do it this way." And then just rebuilt it. And not only did they save a bunch of money, but they got to build it in a way that now works with all the rest of their systems. So, now, if you're a product manager at Pinterest and you say, "I'm writing a [inaudible 01:00:17] and I want to go pull in the goals for the team." It's literally one click and you can pull in the goals of the team that are relevant to your project, and then you can update it. And it flows right back to the main system.
Shishir Mehrotra (01:00:28): And so, what I find is, we start with the very simple metaphor. Use this to run meetings, write your project brief, write proposal for clients. Those are very common use cases. And then, gradually, people learn these new building blocks. And if I ever get caught in the compared to spreadsheets, probably the wrong spot to be, but you gradually figure out that actually my spreadsheet was not that good for that. And I was wrestling with, how do I produce a view for this thing that actually allows multiple people to see a different thing? Or how do I produce something that integrates with these four other systems? And actually spreadsheets are pretty bad at all that stuff. It reminds me to use a different analogy that a lot of people will describe this as a fourth file format.
Shishir Mehrotra (01:01:04): There's doc sheets, slides, and you can technically put a table on a document, but it's a pretty limited table compared to a spreadsheet. And the fact that documents support tables doesn't actually compete with spreadsheets because you quickly realize the wrong thing to do. The same thing, people still use spreadsheets, all the clients. I don't think people stop using doc sheets and slides, but we end up taking a set of processes that are a little bit different and they end up moving over to Coda and it goes very quickly. And I do think this Austin Powers analogy of start from scratch, don't use terminology, people don't understand, oh, I understand pages in the document, I understand what a table is, usually, the first concept people have to understand in Coda that's new to them is a view. That you can create a table and you can create a view at the same table.
Shishir Mehrotra (01:01:46): And we actually see the same thing and they're both editable. For database people, that's a very understandable thing but, for normal people, that's not a common thing. And it's total moment of magic where it's like, "Hey, we're going to build a list of tasks and I'm going to divide it up. And this is the one the developers are going to look at and this is the one that product managers are going to look at and they show four different columns and that's okay." It's total mind change or that's a different way to think about it. And we work our way through people's learning experience that way. And the metaphors we currently see have been there for 40 or 50 years. So, it'll take time for them to change. But I think it's fairly obvious to me that the world is ready for it. And I think we're seeing that change happen as we work through this.
Rob Collie (01:02:27): My picture of what you're doing is really changing quite a bit. Not that what you're doing has changed, it's my picture. This discussion about processes and so, the company you mentioned that runs their billing and they typically think of as a CRM, all of that, all of that happening in Coda is fascinating. The things you're talking about with Excel, it's not necessarily that Coda is aiming at the spreadsheet workload, but spreadsheets are used, misused for a jillion different things. Even what you were talking about, YouTube, if someone flagged a video, it would be a line on a spreadsheet. Right?
Shishir Mehrotra (01:03:06): Yeah. Right.
Rob Collie (01:03:07): Okay. That's an example of where you, "Ah, spreadsheet."
Shishir Mehrotra (01:03:09): [crosstalk 01:03:09] spreadsheet.
Rob Collie (01:03:11): It's just the most convenient thing. The way that our company operates, the processes that run our company ... And our company has turned out to be very heavily dependent on excellent processes. The average consulting firm, the old business model is you land a client and you try to never leave. You park as many people there, you drag that one project out as long as you can. That's the business model. And so, utilization isn't a tough challenge when you're working in that industry. But our whole goal is to be really good for the customer. The best possible thing for the customer is to burn through that project in an amazing way, not too fast, as quickly as possible, as high quality as possible and then if we get to the end of it, maybe they hire us for another one. But that creates what we call the Tetris problem for utilization, right? It's just so hard. And so, we run on all these processes and we have all of these line of business systems, "best of breed", AKA, whatever shit you bought at that point in time and then you bought another one later.
Rob Collie (01:04:10): And the glue between them has really opened my eyes to this second age of middleware, this integration point between all of these things. I've really been hot on this lately. And to hear you echoing this from a different angle, a lot of it is you are integrating with different systems but a lot of it is also happening. A lot of things we might typically think of as line of business software that we would buy and then, of course, infinitely configure forever like Salesforce. The idea that you could do that from scratch in Coda is really fascinating.
Shishir Mehrotra (01:04:43): Yeah. I think it's also related to one of our most popular features is called packs. So, packs is how you integrate Coda with other systems. And we built a few dozen of these now. We're working right now on opening it up so anybody can build their own and you can connect to any system you want. But it's interesting so you can connect Coda to Salesforce or Jira or so on. And you can do some of that with spreadsheets. I mean, you can connect them together sometimes. And, certainly, for analytics, you can. I mean, you can go do the Power BI type stuff and you can go connect to some of these systems. What happens in spreadsheets is that the metaphor is actually very different, right? None of these things have a grid-based data model. It's actually not really what you want.
Shishir Mehrotra (01:05:18): You want something that's a slightly different interface. The other thing is the whole model of a spreadsheet is designed around one way access. The idea is you got one sheet that's all your data and then everything else is [inaudible 01:05:31]. Formulas are not editable and so on. The value proposition of putting it into the spreadsheet this is mostly about reporting and analysis and it's not really about taking action. So, in Coda's case, you pull these things in and the data model makes sense and you can actually like, "Oh, that's a table of customers and that makes sense." You expect to be able to edit it and take action on it. We have this feature we call buttons where you can actually take action on these things. And so, we have a bunch of packs.
Shishir Mehrotra (01:05:53): For example, one of our most popular packs is the Slack pack, Gmail pack, so on, are very popular because in Coda you set up this thing and it says, "All right, I'm going to pull in this stuff for billing data and then I'm going to pull around it like what we did with the project. And then, I'm going to create this button that just emails the client every week with, "Here's the update." You could spend forever trying to pull together four different systems of like, "Here's one view for the client, what's going on with your project?" And you can obviously share the doc with them if you want but in a lot of cases, no, I just want an email with what's the last update. A really popular pack in Coda is Twilio. So, you can actually send text messages from Coda.
Shishir Mehrotra (01:06:28): So, you can go, "I've got a list of people coming to the event and I just want to be able to set a little drip text campaign that says, one week to go. Did you sign up your registration form? Did you fill out your COVID test? And then, oh, two days ago, here's what the weather's going to be. Now, event's starting. Reply to this if you want." And all that stuff that you'd have to be a developer to do, all of a sudden, literally, looks like a document and you can take it and you can build whatever you like in it. And so, we've seen people build just amazing stuff for this. One of my other favorite examples, there's compost farms outside St. Louis and this family runs it and they do it as a volunteer thing. And they somehow scrounged up enough volunteer money to pay a person to go collect compost from everybody's house to be able to put it and it's called the Soil Farm up St. Louis.
Shishir Mehrotra (01:07:11): And it's cool. The community really loves it [inaudible 01:07:13]. And they used to run in a big Google sheet and it's like, "Okay, these are the houses that have their compost set outside." And this person would try to wiggle through it on their phone and you can't really see anything. And he's always wrong. And he's turned this into this Coda doc and it texts people and they text back. There's a good writeup on it, actually. If you search for Soil Farm in Coda, you'll find it. They built this whole workflow and, all of a sudden, this thing just works. So, from that perspective, I would think of what Excel does is a weird detour off of this set, right? It's like sometimes you use Excel to run applications but it's actually pretty inappropriate for most of those things. I think you know I'm one of the deepest lovers of Excel in the planet.
Shishir Mehrotra (01:07:50): This is not meant as a jab at the creators of what I think is the lingua franca of how the entire world operates. This is amazing accomplishment, but these metaphors were literally done 50 years ago. And you can only bolt on so many additions to it before realizing time for a new metaphor. That's where tools like Coda come in. And I think we're not the only one. We see companies running on Airtable and Quip and Notion. And there's other ways that people are approaching this too. And it's very clear, at this point, that all of these are making pretty fast progress. We see more and more people get started with them. And these days it's rare to start a company and not pick one of these tools and say, "I'm going to run based off of one of these."
Shishir Mehrotra (01:08:32): And I'm a big fan of crossing the chasm as a way of thinking about how products grow. And it's just very clear to me that in that analogy of crossing the chasm that Excel, Office [inaudible 01:08:41] is that late majority. It's like, I'm hanging on for dear life. I'm not letting it go. And you see these early adopters. And this whole industry I think in this next couple of years is jumping that chasm. And we're seeing people that feel a lot more normal, early majority folks pick up these tools and say, "Actually, that's quite better. So, I don't really need that old one." For Coda, I think one of our insights is the easiest starting point is a document. It's amazing how quickly kids learn what a document is and blinking cursor, blank screen, what do you do with it?
Shishir Mehrotra (01:09:11): And I think if you get people at that moment and say, "Oh, this is a better blinking cursor." Call it the battle for the blinking cursor. This is a better blinking cursor then you won't think otherwise when you're ready for that next tool. I just thought of a really fun story. Before when I was starting Coda, what was called Krypton, I had this formative experience with my daughter who at the time was, she was probably seven or eight years old. And the way my office laid out, her desk is right over here. Actually, both of them have desks right behind mine. And she was watching daddy work and she said, "Hey, daddy, what's that?" And I was in a spreadsheet and I'm trying to explain to her what a spreadsheet is. And, first off, trying to explain to an eight-year-old what a spreadsheet is pretty hard, pretty hard to explain to a 30-year-old what a spreadsheet is. I'm trying to describe and it's like, "All right, forget it. I'm not going to describe, I'm just going to show you an example."
Shishir Mehrotra (01:09:51): Trying to come up with a use case now. What's a use case for an eight-year-old and their six-year-old sister? And so, I said, "Okay, they've been playing this card game." And I said, "Well, why don't we keep score?" And so, I build this spreadsheet and I write Anika and Ria, that's their names, and I build a column for score and then Anika and Ria were columns. And then there's another column for who won. And that the formula is if this is greater than that then Anika won, otherwise, Ria won. And it's like, "Okay, that's cool." And then I said, "By the way, don't forget to copy and paste that formula all the way down the column." And she says, "Why? What does that do?"
Shishir Mehrotra (01:10:18): And I was like, "I don't know how to explain." So, we've got this tool where you can't describe it to anybody. You can't come up with a use case for an eight-year-old that makes any sense. And the core mechanic of the tool, is it possible to understand and explain? Why is he copying a formula? I'm going to try to explain R1C1 notation to an eight-year-old? It's not the right way to explain what is happening here. And so, in my mind, these building blocks are good and they've taken us a long ways, but we can probably do better. And that's where we started with Coda. And I think so far that's working pretty well.
Rob Collie (01:10:49): My goodness. As a side note, some point, just for fun, just for fun, I'd love to, without recording, just sit down with you and Kellan, our president and COO, and just walk you through how we run our business. I think we're probably a little too deep in to be thinking about switching the way that our whole DNA works maybe. Right? But I think it would be a really interesting cross-pollination because our internal software, our internal automation and systems is like life support. We're like this moon base. We've set up business in a corner of the market that is inhospitable to human life, but there we are. We're thriving. The colony's growing and it's because of this internal system that keeps us alive. I think if we'd started from the beginning, we probably could have used something like what you're talking about. Probably could have used Coda for all of this. We wouldn't need our super overwrought and expensive Salesforce implementation. In the end, it's like simpleton language. The sum total of the whole thing is complex, but you can use small words for all of it.
Shishir Mehrotra (01:11:55): Yeah. I can take a look. Agencies and consulting firms as well are very common clients to Coda. I actually think software is not really built for that market or the software that it is, is pretty poor. Salesforce is not really designed for that market as an example. And one of the way I like to think about Coda is I picture the world as like you have Office and docs, sheets, slides, and so, on one end of the spectrum, and then the other end, you have all these packaged applications, which you call best of breed, right? And in the middle, you had this valley. Everybody goes through this point where your spreadsheet and docs run out and then you have to jump across the valley and like, "Oh crap, I really wanted some of what I had back there and now I'm stuck." And I think of what we're doing with Coda and this whole generation of productivity platforms is we're filling in that valley a bit, or at least extending the bridge across.
Shishir Mehrotra (01:12:40): And it's not meant to say that many of our clients still use Salesforce and Jira and so on. Small companies often can do without them but as you get larger, there's a lot of reasons why people end up with those things. And then you want to connect to them and make sure that you can get a seamless integration together in some. But this idea of this gulf between these things is very high. And I think what happens with these, you buy these pieces of package software that are best of breed and one size fits all. I can say one size fits all over one size fits none. And you end up with this thing that does 1,000 things that you don't need and it's missing the three things that you do.
Shishir Mehrotra (01:13:13): And then you're working a way around it. And maybe I can customize it and then I build a script that does some other thing. And you end up with this mess of not really what was ever intended to be done there. And then, you can't really take advantage of any other updates and so on because they don't really apply to you. And so, we find consulting firms are actually a pretty good target for us. So, yeah. I'd love to show you.
Rob Collie (01:13:31): And vice versa.
Shishir Mehrotra (01:13:32): Yeah.
Rob Collie (01:13:33): Here's the edgy question, the important question that everyone here wants to know, can I pull data from Coda into Power BI?
Shishir Mehrotra (01:13:39): The API works great. I actually haven't tested it myself, but I can't imagine why not.
Rob Collie (01:13:45): We need a Power BI pack.
Shishir Mehrotra (01:13:46): Yeah, totally doable.
Tom Larock (01:13:48): Do you mean that the other way though? You mean to pull data from Power BI into Coda?
Shishir Mehrotra (01:13:53): Both are probably interesting. Power BI into Coda, when we open to the pack ecosystem, we'll be able to do it. One of the hardest parts about pulling from systems like Power BI or Snowflake or so on is the schema changes. And so, when you pull from Jira the schema is mostly the same. They're custom fields, but it's mostly the same. And so, you engineer them all but differently. So, we're about to ship a set of updates that allow pulling from schema list or changing schema systems. But at that point, we'll be able to pull in not only what ... The key part of what happens when you pull into Coda is because we have native support for tables, you end up with relationships that actually match what your database does. So, you don't pull in one table, you pull in all of them and all the relationships are intact.
Shishir Mehrotra (01:14:30): You can navigate them and the formula language understands them and so on. The other way around of, can I take data that's in Coda and pull it in the other systems? That's all possible too, in some senses. We have a good rest head on Coda it'd be relatively straightforward any tool that can pull out of a well-formed, rest-based data API. You could probably do it today without really any work from us. But there's probably some work we could do to make it better.
Rob Collie (01:14:52): Here's the use case. And Tom's right to ask about it both directions.
Tom Larock (01:14:56): Oh, hold on. You said I was right. Hold on. Write it down.
Rob Collie (01:14:58): Yeah. Write it down. Write it down. Right. Yeah. We need to share to say you're right. That's going to carry more weight. Okay. So, if we had, let's say, instead built our whole company's digital nervous system around Coda rather than best of breed, blah, blah, blah, plus middleware, well, we would still need to be able to do some very, very sophisticated reporting and analysis on it that Power BI it is a hammer for that nail. It is really, really, really good, right? So, we'd still want to use Power BI for that, but then we would reach some conclusions. We would want to take some actions.
Shishir Mehrotra (01:15:31): Yeah. You pipe it back in.
Rob Collie (01:15:36): And so, there's two pieces of techno-integration that you could consider at Coda, make it a lot more relevant immediately to me, which is make it really straightforward for me to pull the data in. But then, also, the equivalent of a custom visual implemented in Power BI that allows me to take that action back in to Coda. Right? If my processes were running in Coda, I would do some analysis, reach a conclusion. I need to take some action. And then, I would want to trigger another Coda process. Well, Microsoft has made the Power BI surface pluggable. You can write a custom visual which, typically, you would use to render your own chart, but that same framework can be used for basically anything.
Shishir Mehrotra (01:16:14): Yeah. [crosstalk 01:16:15].
Rob Collie (01:16:15): Now, we're really having fun.
Shishir Mehrotra (01:16:16): We're working on a bunch of the stuff the Microsoft team has asked us to prioritize, no big surprise, integrations in their teams. And so, we're working on some of those first. That's such a [inaudible 01:16:26] that's there. Every time I talk to them that's what they want. It's like, how come you don't have better teams? And we were blocked by a couple technical gaps to get that done, but that'll come soon. And we have actually a really good integration with Slack. And so, I think we'll end up mimicking a lot of that in the team's world. But I think some of what you're talking about about building Power BI visualization and so on is totally a reasonable extension project. And maybe we could hire you guys to do it.
Rob Collie (01:16:48): I mean, [Satya 01:16:51] asked you to do something. Rob asked you to do something. I mean ...
Shishir Mehrotra (01:16:55): All right, Tom, you can break the tie.
Rob Collie (01:16:56): Fine. Fine. 51-49, Satya. Fine. I can understand that prioritization.
Shishir Mehrotra (01:17:07): Satya has been a fun supporter of this ecosystem, actually. We all know he's a good guy. I think he has a pretty good view on, I think, courage innovation in this space.
Rob Collie (01:17:16): Well, I was going to ask you if ... Because I haven't heard you making Microsoft's radar. They were Google Docs obsessed on the Office team for a very, very, very long time. I wonder if they're starting to turn around and go, "Hmm, we'll have to save that for another day, won't we?"
Shishir Mehrotra (01:17:32): Yeah. I mean, I think that the closest thing they're working on is this thing called Fluid Framework, which is probably the closest answer to Coda. But it's pretty early. And then, Satya was pretty upfront about that, had the team present it to me pretty early on. So, I've been well aware.
Tom Larock (01:17:43): Shishir, I could listen to you talk all day. I haven't really said anything because I just wanted to hear more and more from you. Rob had sent me the link to the rituals for hypergrowth and I'm fascinated by all that. And I'd love to talk for another two hours just about that. I want to echo what you said though about the valley, the productivity tools and you talk to a person who feels that all the productivity tools on the market, essentially, make me less productive. If I have to spend more time using your tool than doing my actual job, then that makes it counterproductive. And I feel almost everything out there does it. Like you said, it has 100 things but not the three things you need. And it's a standard that your company will use now. I'm not going to name names.
Rob Collie (01:18:30): Yeah. That's software. That's software for you, right? That's why we hate-
Tom Larock (01:18:33): I'm not going to name names of the companies we use that are just shit but they are. And I just wanted to say, I may be oversimplifying it but I feel what you have because I've been looking at the gallery ... I might oversimplify it but I feel what you have created is the beginning of what I would call is a productivity notebook. That's what you have here is you have this ability to make people actually truly more efficient because they're trying to communicate ideas through disparate file systems, bring it all together and then they could pass it off to somebody else like a Jupyter Notebook where they can run things, look at things independently and you make it part of a process. And I'm really excited by all of that. So-
Shishir Mehrotra (01:19:13): That's a common analogy. I like that one.
Tom Larock (01:19:15): Okay. Can I get you to say that you're right?
Shishir Mehrotra (01:19:18): Tom, you're absolutely right. That's a great analogy.
Rob Collie (01:19:21): Write it down. Another ritual on the podcast. Hey, Shishir, I've really enjoyed this. Honestly, I feel like I get smarter when I talk to you or I want to get smarter anyway.
Tom Larock (01:19:34): I feel [crosstalk 01:19:34].
Rob Collie (01:19:34): It's one or the other. Thank you. Thank you so much for spending this time with us.
Tom Larock (01:19:37): Yes, thank you.
Shishir Mehrotra (01:19:37): Yeah.
Rob Collie (01:19:37): This was fascinating.
Announcer (01:19:38): Thanks for listening to the Raw Data by P3 Adaptive podcast. Let the experts at P3 Adaptive help your business. Just go to P3adaptive.com. Have a data day.
Sign up to receive email updates
Enter your name and email address below and I'll send you periodic updates about the podcast.