BI vs. Analytics and the Death of Excel
We at RSR have been scratching our heads of late about the difference between Business Intelligence (BI) and Analytics. As a group we seem to have come down primarily on the side of the definition that describes analytics as the tool and BI as the output. In that context, BI is something bigger than analytics - BI is the enterprise strategy and analytics is the tactic that enables the strategy.
I've been struggling with this concept ever since Steve & I wrote our latest edition of RSR's annual benchmark on BI (or analytics, if you prefer). There are a couple of issues at play here.
First off, respondents told us pretty clearly that mobile and web access to analytics, as well as new forms of data visualization and dashboards are high priorities for the future, and that integration to Excel is not. In looking at the latest iterations of vendor applications (we're getting some previews of things to be announced at NRF next month), it's clear that a lot of vendors have invested a lot of money in putting a layer of analytics on their applications. This isn't reporting, mind you. This is new ways of visualizing data, either web-accessible or designed specifically to be mobile-friendly. If you want any kind of sense for the trend, just Google the word "infographic" and take a look at the image results and you'll see what I mean. Most, if not all, of these are not designed to be analytics tools per se, but they are creating an expectation among consumers of information that all information will eventually be presented in a visual way. Edward Tufte would be so proud.
The problem is, which application should be used to create these information rich, visually appealing analytics? Is the transaction application that generates the data the best place to do so? Theoretically, the designers of these applications know their data the best and would have the best idea of what insights would be most meaningful to get out of that data. On the flip side, more often than not, the best insights come from connecting disparate data sources, so analytics on a transaction application would be limited to only the transactions that the application itself has visibility to.
That would argue for an enterprise analytics strategy, where all important data (note that there's already a qualifier in this statement: who decides what is "important"?) is sucked up into a giant data warehouse where people can examine relationships and generate insights at will. Oh wait - generally speaking, those kinds of data warehouses are insanely expensive, and require specially trained people who both know the analytics tools and also know the data well enough to know the right kinds of questions to ask. So it's not just "anybody" who can play around in enterprise data, and that's where this coin flips back on itself. If you're not a user with access to that kind of analytics power, where do you turn? You turn to the analytics that come with the transaction application - the stuff you already have access to.
This relationship is getting disrupted, however, by high performance analytics. Steve and I recently had an experience with this, when we worked on a social listening project that performed analytics on Black Friday and Cyber Monday social media (the results were shared on Chain Store Age). There's been a lot of talk about social media analytics, but seeing it action drove something home to me. Sure, you can talk about "natural language processing" and "learning systems" and all the geeky stuff that makes social media analytics sound like it's pre-Star Trek technology or something. But the reality is, a very powerful analytics machine - the combination of software and hardware - is making it possible to chunk through a much larger data set than has previously been possible. And the results are already very infographic-y - time series and other charts that take relatively little massaging (especially compared to the effort of getting that first level of information) and instantly convey an enormous amount of information - literally millions of tweets and posts - in a very concentrated space.
This is no job for Excel.
Don't get me wrong - Excel is not going away in the retail enterprise. But we may rapidly find ourselves in a place where Excel holds sway only over ad hoc analytics, and either enterprise BI or vastly improved transactional analytics takes the lion share of end user attention.
For me, the question becomes one of the future of BI, or analytics - or both. Between Big Data on one side and mobile access on the other, the discipline of BI has never been more disrupted. Will data truly be democratized - accessible, with powerful tools for understanding what the data means, to everyone in the enterprise? That seems more likely today than ever before.
Considering how retail has historically been not very measured - it's only been in the last decade or two that the industry really cared which customers specifically were walking through the door (or shopping online, or posting about them on Facebook), as opposed to which products they put in their shopping carts. I think retail, more than a lot of other industries, has the potential to be transformed by the waves of change hitting BI - not because they'll be doing anything different when it comes to retail execution, but because they'll be able to learn so much more from the things that they are doing.
More on that in 2013.