Dashboards are dead? Really?

I saw this nice discussion on LinkedIn last week on the state of dashboarding. And yes, I was as surprised as you to see a relevant, professional post on this platform in 2022, as LinkedIn has turned into some sort of Facebook lately, but okay… 😅  

The premise of the post was that dashboarding has gone completely out of hand: everybody now creates his own version of the truth with all these modern easy-to-use tools. An interesting memo by ThoughtSpot called Dashboards are dead was included with some thoughts on a different approach for data analysis. The ThoughtSpot solution is driven by natural language search (NLS) and artificial intelligence (AI) and it actually ranks pretty high on the Gartner Quadrant for Analytics and BI. This is actually a very current topic as SAP acquired Askdata last week! Askdata also operates in this area and can be expected to further extend the “Search to Insight” capabilities in SAP Analytics Cloud.  

So will this Askdata finally elevate SAC Search to Insight to something that will actually be used (and not only in demos)??? #SAPAnalyticsCloud

— Xavier Hacking (@xjhacking) July 21, 2022

Time wasted

Although this “Dashboards are dead” document by ThoughtSpot is obviously created for marketing, it does give some interesting insights on how dashboarding and reporting is approached, especially in the world outside of our SAP bubble. Some comments made me scratch my head though. One of the narratives here is that although we have all these fancy dashboards now, still a lot of time is wasted in keeping them up to date with proper data (chapter 3). I actually do recognize this from some projects where we were asked to rebuild/develop some PowerBI solution that had grown out of hand and indeed couldn’t be maintained anymore. But, I absolutely don’t agree that this is because of the tools that are used. 

A common reflex is to blame the software and switch to another solution and try again, instead of properly using the tools that you already have. Nothing here that can’t be solved by using your SAP BW/HANA/DWC backend with an SAP Analytics Cloud front-end, and I’m pretty sure that’s also the case for solutions from competitors. Let’s start by using these tools in a more structured way, without all these quick shortcuts and workarounds that I often see in these standalone PowerBI solutions. And yes, that is boring to setup and will take some time and effort upfront, I know, but the benefits are huge in the long run. 

Whenever we deliver a dashboard solution (end to end), the standard should be that it runs completely hands-off and doesn’t need any periodical maintenance to keep it working. This means no manual data loads and transformations, no manual data cleansing, no “current year” filters that have to be changed on the first day of the new year, but also no wacky formulas or logic in the front-end to “fix” some data that was already incorrect in the source system. On top of that, our dashboards should be live and make use of a proper (data) authorization concept, on a platform that is setup in such a way that user can quickly find their stuff (for example using the SAC Catalog).

Long development cycles

Another topic that is discussed is the lead time for dashboard development (chapter 5). It takes a lot of time to implement changes. Yeah, true, but I believe we are making huge steps in this area. With a combination of an agile project approach and fast(er) development tools as Data Warehouse Cloud and SAC we are definitely able to deliver working solutions and implement changes in a much shorter time than in the HANA/BW + BOBJ days.

The ThoughtSpot document also states that most dashboards are way too generic and fail to deliver personalized insights. I think however that it is actually a good approach to keep a dashboard focussed on just a few topics and KPIs, and in addition offer some options for end-users to further explore the details and create their own ad-hoc insights. Again, this shouldn’t be a problem when your datamodel is setup in a proper way. In SAC we have tools as the Explorer or Data Analyzer for this and don’t forget good old Analysis for Office here. This works fine without the need to open your system and let everybody create their own stories (which could end up in a mess if not managed properly). 

A new paradigm?

What I see for NLS looks still very much similar to the current “Search to Insight” feature in SAC, which I don’t believe anybody is actively using. Also the examples that I see in these videos are so simple that it is almost embarrassing. Do we really expect data consumers to type these long queries to get a revenue per month chart? And what problem does this exactly solve? I can do that with a few clicks in a SAC story (or probably even faster with a Data Analyzer/Explorer view). Also, the basic idea of having a dashboard is to have a standard set of KPIs that need to be tracked on a recurring base. The standard is important here, as we want to be sure that we are looking at the same definitions throughout the whole organization, over time. 

My idea for AI analytics would be that a tool eventually generates a complete analytical dashboard for me, based on what I’ve analyzed in the past, supplemented with new insights on major changes in the data whenever they occur.

I must say that I like the idea of ThoughSpot’s SpotIQ, where the tool looks into the dataset to find insights that you might not have thought of. SAC has these “Smart Insights” for some time as well, but this often only shows the obvious things that an analyst already knows (and there are still all kind of technical limitations to use this). I saw that SpotIQ also has a like/dislike feature for each chart/insight to help to train the tool. Nice! 

Searching for a problem?

Still, looking at this document I get the feeling that ThoughtSpot, and probably NLS/AI in general, is a solution that is still a bit looking for a problem to solve. AI will eventually be huge and the developments in this area probably will go super fast at some point. However, I’m really curious how they will solve the “data modeling” part, as somebody still has to define all these company-specific definitions somewhere, right? Applying AI on nicely structured datasets doesn’t sound that shocking.

NLS is nice for a demo but I don’t see this being adopted on a wide scale. But, let’s follow these developments and see for example what for example an Askdata will do for SAC in the coming year! Please surprise me! 😃

What do you think about NLS and AI in analytics? Can we already ditch our dashboards?

HackingSAP.com - Jul 28, 2022 | SAP Analytics Cloud
Tagged | AI, Artificial Intelligence, Askdata, Natural Language Search, NLS, ThoughtSpot