Dashboards are Dead! Long Live Dashboards!

Speaker 1 (00:00): Hello friends. Have you been seeing the phrase "dashboards are dead" on LinkedIn and elsewhere? It's pretty catchy. It has a good ring to it, even in its softer forms like, thinking beyond dashboards. It's almost perfectly provocative, isn't it? It's the perfect kind of engagement farming toxicity that worms its way into our subconscious, creating a discomfort that gets us to click through or sit still long enough to watch whatever video short is being thrust onto our eyeballs. I'm getting better and better at spotting this kind of manipulation lately, and it's a dirty trick that isn't going away and it's not going away because it works. Even though that I think this kind of thing is a pestilence of sorts, let's take a moment and just admire its perfection because that helps us better understand and defang it. Why do I think something like dashboards are dead, represents like the peak of toxic engagement farming?

(00:59): Well, first of all, it's short. Dashboards are dead. Three words, four syllables. There's even some alliteration in it. Dashboards dead. The way things sound in your head when you read them does matter. It has a significant impact both on your willingness to engage and on the likelihood of you remembering it. It's pithy. It sounds good, it's memorable.

(01:23): Next, it invokes something we all understand and are all invested in, dashboards. Even if you don't have good dashboards today, you still know intrinsically that you'd be better off if you did have them, better informed, better able to spot problems and diagnose their causes, better able to spot opportunities and formulate ways to capitalize on them and to do so through a graphical third grade level visual interface, who doesn't want that? Yes, please. Everyone understands what dashboards are, whether you're a data practitioner or a business leader and everyone believes they're valuable. In fact, in the vast majority of organizations, we're still striving to get good dashboards for everything we need.

(02:06): Okay, so the phrase, dashboards are dead. It invokes something we all understand and that we all kind of value dearly, and then it negs that thing by saying that thing is passe, that it's dumb, that it's yesterday's news and only losers are going to stay invested in it. That word, neg, by the way, is a word I've learned from Reddit. Thankfully, I'm a bit too old to have known this word in my own life, and instead I've learned it as an observer, like an anthropologist looking in on the generations that came after me. But the term neg, if you're not familiar with it, is a dating tactic employed by toxic individuals.

(02:44): For instance, if a guy finds a woman attractive and wants to win her attention rather than being nice to her, he finds something about her to insult or denigrate. And if this tactic works on her, she then feels a bit of a desperate need to win back his approval. It puts him in a position of power right from the start, and if it doesn't work and it just makes her angry as it should, he just moves on and tries the next person, spewing toxicity into the world, shrinking the size of the overall happiness pie so that selfishly he can achieve some sort of short-term exploitive win. Gross, right?

(03:20): Well, this LinkedIn meme, dashboards are dead, is exactly the same thing. It's just as gross in my opinion, maybe even grosser actually, because the dating version of it is seldom practiced in a broadcast form. We can now leverage digital platforms to spew toxicity at scale. Again, shrinking the size of the total happiness pie so that the producer of said toxicity can extract an exploitive win. Yuck.

(03:47): And let's be clear, it is the same fundamental tactic as the dating neg. Don't think for a moment that insulting dashboards rather than you directly lets them off the hook. By insulting something you believe in, they pull off the dirty trick of denigrating you without having to come out and say it. You connect the dots in your own subconscious. I believe dashboards are valuable, and I hold this belief so intrinsically and confidently, and here's someone telling me that belief is dumb. You better believe you make the connection and you feel the angst, you feel the insecurity.

(04:24): And then just like the negging dating strategy, your first instinct is to engage with the person making the insult because clearly they know what's up. Maybe if I read or otherwise engage with their content, I'll be made whole again. I'll feel better. Yuck, yuck, yuck, yuck. It's one of the reasons why I continually struggle to engage with LinkedIn, both being exposed to this kinds of toxic strategy all the time and knowing that if I want to get people to engage with me, I have to compete against people who are cheating. It's just a massive, massive turnoff.

(04:57): Okay, that aside, now that I'm done railing against this specific idea, dashboards are dead, and also against the state of professional social media at large, I'm going to surprise you and say that deep down under the hood, there is something worth learning here from this, dashboards are dead thing. Dashboards are not dead, but there is a change coming and I think it's pretty exciting. I've actually talked about this a bit on previous episodes, so if you've been tuning in lately, the contents of this won't necessarily be a surprise, which is that generative AI chatbot interfaces are coming and they're going to change our relationship with dashboards. They're going to change our relationship with semantic data models, with Power BI data models.

(05:37): But before I revisit that topic, let me say definitively that dashboards are not going away. They are not going to become less valuable. You know that hole a picture is worth a thousand words thing? That's actually basically a fact. Our eyes digest well-formatted, visual graphical information at just a massive. Our ability to quickly compare the length of bars, spot outliers via color shading and digest trends in line chart form, this is all leveraging our biological hardware in a way that can't be replicated by anything else, at least not until we all get brain implants, I guess. So to suggest that graphical and interactive dashboards are dead is not only a toxic professional dating strategy, but it's also just dumb and false on its own and dumb and false in some really obvious ways. Maybe I should start posting things on LinkedIn like, only dumb people think dashboards are dead, like playing the Uno reverse card on those who neg. Tempting, but I think I'm going to stick to the high road.

(06:40): Okay. Dashboards are super valuable and they aren't going away. So what is the shred of truth lurking under the toxicity? It turns out that the way dashboards get built and consumed has always been a bit broken. As I've mentioned in previous episodes, I think it's super obvious that generative AI chatbot interfaces are going to change a lot of the workflows around dashboards.

(07:03): First of all, even finding the right dashboard, even if you know it exists, there's often a lot of cognitive overload there. Even I struggle with this on a regular basis with our internal dashboards at P3. Where is that one dashboard I've seen before? The one that I don't need all the time but suddenly need today? Which workspace is it in? What was it even called? I know what's in it, what it contains, the metrics that I need to see, but the names of the reports and workspaces are completely lost to me.

(07:33): Imagine instead of hunting around our Power BI tenant and not even recognizing the names of the things I want when I see them, I instead get to go to a Power BI Copilot chatbot and ask it, "Hey, you know that thing that shows me XYZ broken out by ABC," and it not only takes me to the right report, but to the right page of the report, that's going to be a game changer, a search interface that understands my choppy vague description and then shows me precisely where that's displayed. And that's going to help in cases where I don't know whether the dashboard I want even exists.

(08:05): That's the use case that confronted me recently with the dashboards I made for our recreational hockey league. Someone smart and savvy was asking me if we had a dashboard that answered question X. and I was like, "Yeah, that's on the very first dashboard of the set." I think this is a massively underestimated problem, not being able to find the thing that you want, not knowing that it exists, even when it does. Imagine if instead he had a chatbot, he could grow to trust over time, like a hockey oracle of sorts, instead of having to navigate around hoping that the thing he's looking for even exists much less knowing where it is. And as another one of the hockey guys told me today, "Yeah, the more dashboards you make, the more questions you successfully address." Well, that is a downside where each additional dashboard obscures the others and makes them all harder to find, which is 100% true.

(08:57): So I think in the not too distant future, we're going to look back on the go hunt for the things you need interface and see it as Stone Age level primitive. Relying on people to know that something exists and where it exists in a tedious choose the right tab interface is not going to be something history is going to judge kindly. And good riddance. This problem is a huge impediment to adoption and traction that we never really like to admit to ourselves. We think of it as have the data model, we built the dashboard, problem solved. But if it's not meeting the world and it very often isn't, we really haven't succeeded. So generative AI chatbots are going to help us find things, things we know exist and things we hope exist. But that brings me to probably the most powerful use case, which is when stakeholders are going to be using bot interfaces to build dashboards that did not exist.

(09:53): Think of it this way, your stakeholders, your dashboard users, they have questions. Forcing them to map their question into the right corner of an expansive portfolio of dashboards stretching out across a Power BI tenant or a SharePoint site is kind of the wrong thing. It's completely backwards. But it's been the only thing they've needed to go find the location that answers that question. But what if they can just start with their question and not even bother to go looking in the first place? And yes, in cases where we do have a pre-built dashboard and someone goes looking for it via a question or whatever, the chatbot will take them to that dashboard. But when no such dashboard exists yet, do you think the chatbot is just going to go, "Nope. Sorry, no such dashboard." It might say that, but only if there isn't a good Power BI model built.

(10:45): If there is a good Power BI semantic model already built, one which contains the data and metrics required to answer the question, I don't think the new chatbot interfaces are going to give up. Instead, I think they're going to build a dashboard for you. And just to be clear, the result is going to be a dashboard. It's going to use visuals to take advantage of that biological hardware I was talking about earlier. It's going to be interactive. You'll still be able to slice and drill down and select single bars in a chart to see it cross-filter other visuals. It's just that no human being operated the field list to create it. So in this case, the chatbot will be a replacement for the field list. Drag and drop field list interfaces are kryptonite to the average user, to the average stakeholder. Heck, they're often kryptonite even to savvy data professionals when those professionals aren't the ones who built the model in the first place.

(11:38): And I think we're going to look back someday soon and also see the field list as a bit of a primitive relic. Maybe we'll still have the field list as a backup for the data pros, but most of the time even for us, it's going to be faster to describe what we than it is to be dragging, dropping and formatting forever.

(11:57): Well, for any of this to work the semantic model, the PIVX file will need to be very well-built. It will need to contain all the right data. Just like today, it will need to have all the right relationships, again, just like today. And it will need to contain all of the right semantic definitions of core metrics just like today. But it'll also need to contain some amount of extra information. Let's call it semantic instructions, which are specifically crafted to help the chatbot interface do its thing.

(12:28): For example, my hockey model contains multiple different measures for goals per game. There's a player version and a team version because there's subtleties in the denominator, a number of games played that aren't easy to get right for both. And if someone asks a chatbot interface in the future, which player scored the most goals per game in season 16? I'm pretty sure the chatbot Copilot interface will know which measure to use, the player one, and it will also know which version to use if I ask it, which team scored the most goals per game in a particular season. But if I ask it a more vague question like, does goals per game scoring go up in the summers, is it going to know that I want it to use the team version? Maybe not. So if I want to clear that up, I, the developer of the model, am going to have to tell the model to use the team version in cases like that.

(13:20): So that's a cool new thing for us data professionals to be doing, a new component of building these models, enriching them with semantic instructions with the chatbot in mind, something which greatly enhances the impact and utility of everything we do by making its value far more accessible.

(13:38): So yeah, the dashboards are dead crowd, they do have a point. In the near future, we're likely to see our stakeholders start with their trusted chatbot rather than go looking for a dashboard. There will still be pre-built dashboards because some needs are every day and predictable needs. And for that small set, having them on speed dial like the favorite list in your browser, yeah, you're not going to need the chatbot to find those. And building at least some dashboards is also an important part of validating and debugging the semantic models themselves. But we probably won't build as many dashboards as we used to because for our users and stakeholders, the line between using the chatbot to find a dashboard and using it to create one might get deliberately wonderfully blurry.

(14:26): What do they care? They have questions. The chatbot always gives them the right dashboard. Simple. And by the way, when you hear people say things like, "Well, if you're going to take advantage of AI, you need to get your data organized," building semantic models might be the prime example of getting your data organized. In many situations, semantic models might end up being the only significant organization that you need. Organized might just come to mean enriched with the right data, relationships, formulas, definitions, and semantic search instructions.

(14:57): So yeah, all you folks out there building and using Power BI models, you're prepping the ground for this new wave, a bright new future in which dashboards aren't dead, but are even more alive than they are today.