Assumptions II

Originally Posted on January 18 2012


In our second anecdote we see the CEO of the Italian subsidiary of a large, American, corporation in the area of CPGs, complaining that “quarters are always on the rollercoaster, they can start good than drop; or they can start really bad and recover on the finish line“. Actually, given the modest seasonality and a large, consolidated, group of customers, quarters should have been rather simple to manage.
Sales force was told to start the quarter by visiting the “Gold” customers, to collect the largest orders in the first couples of weeks. This was a good commercial practice and had the collateral advantage to improve quarter predictability.
Despite this directive, many (not all) quarters kept being irregular. Orders profile by customer and volumes seemed to be coherent with the market segmentation, so I was tasked to find a rationale through the issue.

I ran many reports, observing the quarter build-up against the targets, and, indeed, they showed that bizarre behavior. I tried various segmentations, till when I sorted a single salesman’s orders by the first input date. That was different than receipt date (a salesman was allowed to “withhold” a customer order before sending it to the company, to make corrections, for example).
I spotted two patterns. First, he didn’t visit his Gold customers first; he just visited those closer to his residence. Second, he withheld some medium sized orders just to have a “reserve” in case the Gold and Silver customers were below the expected quantities. A non-negligible number of orders also, were originated in the quarter before the one being examined.
I checked the entire sales force (about 100 people, a tough job) and spotted similar patterns in a large majority. As in the previous case, handling this information was rather painful and my popularity suffered from that.

They could have spotted that before, but that tiny date field was never used in reports and overlooked. Though, we diligently included it into the data warehouse even without a clear perception of its usefulness.

Now, lesson learned.


1)      Assumptions must be verified, even if they originate from the CEO, the CFO or any other C level executive.

2)      Learn to recognize assumptions when you see them. Some seem just natural, so they appear not to require any further justification, but they do.

3)      Managers’ mental models are hardly accurate because they focus on what matters them most, costs the most, and consumes more time but, not necessarily, affects the most the bottom line. They’re also human, and their focus may be not entirely rational.

4)      Managers’ mental models are more accurate than mere drillable reports, which are designed along the lines of database design. Whatever they use to consume data, it should include their view, not the simple database view.

5)      Points 1, 2, 3 and 4 are the reason why people like me keep being paid. This is somewhat reassuring, isn’t it?