Originally Posted on 8/4/2009
Applications like Power Pivot, at least by now, sit inside the broader "workgroup intelligence" category. This is often criticized because of the lack of focus on delivering a single version of truth.
In my opinion, backed by a respectable number of disappointing projects, there's nothing like a single version of truth. In fact, enforcing it may do more evil than good.
Let's consider the most basic of all BI outcomes: sales.
More precisely let's talk about invoices. Nothing appears to be easier the defining the monthly invoiced amount; just sum all the invoices. Many senior managers will think this way and will be irritated by any further question, save being irritated as well by realizing that the accounting department provides different numbers than those in their sales reports. How are credit notes supposed to be taken into account? Should they decrease the invoiced amount? All of them? And returns? What about those occasional invoices not issued upon a sale but for other services? Are all the products that produce a revenue relevant to sale performance (ex.. paid merchandising)?
How many other particular cases lie inside a complex organization? Accounting has its own standards to comply with, and knows exactly what to do with every document, so they're out of the game.
The other offices will likely have an answer for each of the questions above (and probably will be surprised that the answer is not obvious to everyone), but they'll all have different answers. This is not misalignment, the point is that everyone has a different goal and measure her success on different metrics. So the sales manager should count in returns but not those caused by product quality issues; marketing will need to include the merchandising sales; production will be interested more in returns than anything else and the top management, by definition, should care about everything but should also be able to clearly discriminate each component. Everybody will have a different number, but it must be the number which actually measures her performance and must be confronted with. Sales are only the most immediate example but the same happens in every other analysis. Costs, for example, are even more complex.
Our single version of truth paradigm so becomes something different.
What's crucial is the ability to relate each metric to the others and decompose the various effects which account for variations. There must always be a calculus through which the sales according to group A are transformed to the sales according to group B.
I often found useful sharing this schema with all the senior managers involved in the project. This extinguishes potential arguments from the beginning and promotes alignment among all the particular goals toward a common economic result. This usually works well till the CEO abruptly asks for the sum of all invoices…