Lies, damn lies and sales figures

by Phil Cohen

How many business decisions are based on statistics? How important is an understanding of statistics to managers making those decisions?

When the government recently revised its estimates of business equipment spending for the past three years, throwing out capex survey figures and replacing them with harder data from tax returns, it invalidated a lot of other people's careful analyses of the lack of business investment; it turns out that the investment - whose absence the analysts had so worried about - was there all along.

The suddenly appearing investment upturn was a result of changing data sources; the upturn was there, but didn't appear in the capex figures because the data collection method was flawed. A better understanding of data collection and analysis methods may have improved some organisation's chances of cashing in on the investment rise, perhaps by investing in new plant themselves earlier.

Another problem caused by a lack of statistical knowledge is the habit of reading meaning into every rise and fall in every figure, particularly closely-watched figures such as sales volumes. It is not unknown for branch offices (and even individual salespeople) to hold back on reporting sales in good months or quarters, to even out the variations between reporting periods. The ideal is a set of sales figures that rise evenly - let them rise too fast this quarter, and you may have a drop in the figures to explain next quarter.

The problem is compounded if falls in production or sales figures are used as the basis for motivation. A salesperson has a certain amount of control of sales, of course. But can they control every aspect of an individual sale? Or is there an element which must be considered 'random'?

A large metasurvey carried out by researchers in the US showed that, while techniques such as incentives and commission were an effective way of increasing salesforce motivation, there was in fact little correlation between motivation and actual sales. So variations in sales figures from one period to another may not be due to motivation at all, and grilling sales staff about every fall in sales figures may actually reduce their effectiveness.

The manufacturing industry had the same kind of problem in understanding production figures, until W Edwards Deming and a number of other statisticians and engineers developed 'statistical process control' (SPC), which later formed the basis for some of the tools of Total Quality Management (TQM). SPC is a set of simple techniques for separating real changes from statistical 'noise'. In manufacturing, for example, managers and operators can use SPC to find out whether a change to the manufacturing process has improved the output of a process, or whether random variations are responsible for a one-off rise.

The same techniques can be applied to any sequence of numbers: prices, staff turnover or sales volumes. Researchers at Sydney University Graduate School of Business [Improving Business Performance: The Competencies You Need - paper presented at the Fourth Australian Financial Controller's conference (AIC), Sydney 15-16 August 1994] are currently looking at applying SPC techniques to business data.

If we are making decisions based on statistics, then it is important for every manager to have at least a basic grounding in the tools, techniques and - most importantly - pitfalls of statistical analysis.

9/9/94
------------------------------------------------------------------------


This is one of a series of articles written by Phil Cohen and Onno van Ewyk, HCi . Most of the articles were also published in the Australian Financial Review. This article may be reproduced only with the permission of HCi Consulting (email HCi ). Copyright HCi, 1993-1998.