Future BI: Bring on Synthesis
The next frontier in business intelligence and analytics might well be a synthetic one.
- By Stephen Swoyer
- October 24, 2012
The next frontier in business intelligence (BI) and analytics might well be a synthetic one.
That's "synthetic" as in synthesis -- i.e., the recombination and re-contexualization of (in this case) the results of BI, analytics, and other information practices.
According to Scott Davis, co-founder of upstart business intelligence (BI) player Lyzasoft Inc., synthesis is by far the most significant -- and the most intimidating -- problem in BI.
Although BI and analytics can't yet deliver turnkey synthesis, Davis says, they can help users to more effectively synthesize for themselves: chiefly, he argues, by making it easier to construct "narratives" -- stories that explain what a user thinks is going on.
"[S]ynthesis is taking the things that exist and ... combin[ing] them in a way that creates something new," Davis explains. "I have to have at least a loose idea of a non-existent target toward which I am synthesizing this stuff. This is what the guys with the ETL tools do. They have an idea of what they want to create and they create it from existing constituent elements. This is a lot like what [users are] doing when they create narratives. When I have a narrative, I have a conceptual plan for how I'm going to manage multiple interdependent concepts."
"Synthesis" isn't an off-the-shelf technology product; you can't "buy" it. More to the point, Davis argues, the concept itself is anathema to the tool-centric philosophy that he says dominates traditional BI. The site of synthesis -- both now and for the foreseeable future -- is the human brain. Tools can be used to augment this activity -- e.g., by permitting users to model or "test" their narratives, or (in a social context) by enabling other users to provide feedback on a narrative -- but they can't be made to perform it entirely.
"We have to be very careful that we do not conflate tool-related limitations with cognitive-related limitations. What we've been dealing with up to this point is a very tool-oriented view of the world," he says.
Synthesis isn't just a process of re-contextualizing reports, charts, KPIs, predictive analytic results, and so on. Instead, Davis indicates, it consists of pulling in information from a wide variety of contexts -- from charts and KPIs to social entries on Facebook or Twitter to multimedia content on YouTube to "reference" content on Wikipedia -- and using it to tell a cogent and coherent story.
This example gets at what Davis describes as one of the fundamental problems of BI and data warehousing: from the beginning, business has been (force) fed information in IT-packaged morsels. These morsels aren't particularly tasty, Davis argues; in fact, they're unhealthy. From the perspective of business users, they do not provide the kind of nourishment that can and should be used to inform decision-making.
"If you ask an IT specialist what constitutes a datum, they will tell you [that] a row in a database constitutes a datum, or a value in a cell. If you ask a user the equivalent of that question -- [i.e.,] what the fundamental atomic level piece of information is that you pass around, the thing that you have a handle on -- they're going to talk about something that is not atomic [from the perspective of] the IT guy. It's molecular: it's composed of many 'atoms.' A chart, [for example,] is a datum," he explains.
"To the IT guy ... synthesis is when [you] do a join, it's a two-key join, and it's a left-join. That's synthesis to the IT guy. Synthesis happens all of the time in the end-user community, and what they're synthesizing is 'I took this chart from over here, and I took this Wikipedia entry, and I took a whiteboard drawing ... and I put all of this stuff together.'"
The salient point, according to Davis, is that "users don't live at a datum layer -- they live at an artifact layer." This is where Davis and Lyzasoft have a dog in the race. Their offering, Lyza, claims to permit users to work with and manipulate information at this "artifact layer."
"Instead of forcing users to synthesize information at the row-column level, what if we allow them to synthesize information at a level that's meaningful to them? What if that's the chart, a dashboard, or a paragraph? Why do we continue to view business intelligence as fundamentally a row- or a column-[centric enterprise]? Where is the narrative?" he asks.
The goal, says Davis, is -- in effect -- to be all things to all potential users. In some cases, this means catering to datum-focused analysts or power users; in others, however, it's a matter of promoting collaboration and contextual diversity -- e.g., being able to pull in information from structured, semi-structured, and unstructured contexts -- in an environment (Lyza itself) that makes it possible to construct "narratives."
"We allow people to synthesize at the level of a traditional IT approach, so we have 'users' and they will do the equivalent of an ETL process without having to have IT people do it for them. Yes, that's cool and great and we're proud of that," he concludes.
"The point is that ... [n]ot everyone is able to deal with the sort of logical challenges that are involved in wading through database records. The point of Lyza is how do you allow every participant ... to interact with information and people so that they can understand what they need to understand in order to make better decisions?"