The 1 to 1 blog recently reported that SAS’ CMO “unveiled the eight levels of analytics”:
1. Standard reports — It answers the question, “What happened?”
2. Ad Hoc Reports – how many, how often?
3. Query drilldown (or OLAP) – where exactly is the problem?
4. Alerts – what actions are needed now?
5. Statistical analysis — Why is it happening? What opps am I missing
6. Forecasting – what if these trends continue
7. Predictive modeling – what will happen next?
8. Optimization – how do we do things better? What is the best decision for a complex problem?He notes that the first four support reactive decision making, while the second four support proactive decision making. The goal is to use analytics to optimize data and actions to find the best solution for a specific business challenge.”
[Note: I cut and pasted the above from the blog. Don’t get on my case about the inconsistency in capitalization and use of the question mark.]
This model has a few flaws: 1) It isn’t labeled correctly; 2) It incorrectly characterizes the components as levels; and 3) It inappropriately assigns the elements to reactive and proactive decision making.
First off, why couldn’t a report that addresses “how many, how often” be a standard (vs. ad hoc) report? In fact, why couldn’t a “how many, how often” report be a query drilldown? And for that matter, why is a query drilldown about problems and not about successes? While we’re at it, a business that applies a trending formula to sales could easily make a forecast (“level” 6) a standard report, no?
Second of all, these aren’t levels. Levels implies an order or a hierarchy. From a capabilities perspective, there’s no reason why a firm couldn’t forecast something like sales (“level” 6) without knowing where exactly the “problem” with sales is (“level” 3). “Level” 4 is really problematic for me. How could I, as a manager, address the question of “what actions are needed now?” without some understanding of “what if these trends continue?” and “what will happen next?” (“levels” 6 and 7)?
Thirdly, I just don’t get the distinction between reactive and proactive decision making. Decisions aren’t that neatly compartmentalized. A decision I (as a manager) make is informed by, in varying degrees, data and input that cut across the eight “levels”. How can anyone distinguish between a reactive and a proactive decision?
The questions posed in the model, however, do provide a good framework for assessing a firm’s analytics capabilities. You have to be able to answer those questions. And you should be able to determine how well your firm does in answering them.