Don’t Hire An Accountant To Do Analytics

Moving from tracking results and bottom-up, siloed planning towards top-down opportunity analytics requires someone to crunch the numbers. But who? My advice would be to avoid using accountants.

Why? My accountant does a fantastic job…at being precise. I once asked him if our accounting software costs about $80 per month. His answer: “No, it’s $79.45 per month.” Now that’s precision.

But precision is a false god when it comes to strategic decision making. Often you can be highly precise and completely wrong. The goal in strategic decision making is not to get at the precise answer but to the right answer, or in other words, to be accurate. Accuracy and precision sound like the same thing but are in fact fundamentally different. Often they are at odds with each other, which can lead to poor decisions.

For example, most retailers measure their flyer performance using a metric called “flyer sales”. This is literally the sales volume of all items featured in a particular flyer, and is the metric about 70% of retailers in Canada use. Besides being a completely useless metric, it has the added downfall of leading to incredibly bad decisions. For example, you can sell 100 units of an item the week before the flyer, and then 110 units during the flyer. What are your flyer sales in this case? The report will say 110 units. However, only 10 units are incremental, often with no mention of the “baseline” 100 units. A merchant that wants to look good can game the system by simply putting their top-selling SKUs in the flyer, as more units sold before the flyer will mean more sales in the flyer, leading to fantastic productivity metrics. All of this is very precise – precisely wrong.

The right metric, which is tough to measure, is the incremental sales lift. And while some retailers attempt to incorporate this in, it can cause some apprehension in its ambiguity. Do you use lift vs. the pre-period? How many weeks should the pre-period be? What if you have a flyer in the pre-period? How do you factor in cannibalization of non-promoted SKUs? What about the sales the promotion is stealing from future weeks? While all of these questions make the analysis more accurate, they hurt the precision for none of these questions can be answered definitively as they each have an element of error. The fact is people hate ambiguity, and would rather have a precise but wrong number than a directional but accurate number. Overcoming this fear is the key to great analytics.

Ambiguity drives people crazy, so they just start to hide it and as a result start to fool themselves. More often than not, they start plugging the numbers by just making them up. For example, the CFO might say “we need an ROI analysis on this”. The intention is to get precise information on the return on investment of an initiative or campaign. But an ROI analysis consists of a number of inputs, many of which are not hard numbers. If there are ten inputs, eight of them may essentially be assumed, leaving only two based on hard facts. People start making stuff up just to get a precise number!

In some cases, they might even remove data points because they don’t fit into a precise picture, even though those points need to be triangulated to get a better picture. So, in the process of trying to be as precise as possible, the comfort of black and white thinking actually leads to less accuracy. Allowing some grey to enter the black and white zone is the only way to get information that isn’t misleading. This will require a change in thinking, such as when and how “estimates” should be applied. For many people dealing with financial analysis at retailers, the worst word in the English language is “estimate”.

Let’s take a weekly dashboard as an example. Even a simple report like a weekly dashboard is incomparable across departments. It’s not a perfect example because it’s a tracking report (and hence not about finding opportunity), but it is an example most retail executives can relate to. Here is what executives at a billion dollar retailer would typically see:

All three departments are doing worse than they should, according to plan, and the inventory team is even off by 10 percent. Listening in at dozens of weekly sales meetings, my consulting team found that the retail team would typically focus on the biggest issue: inventory. Thus, the main message coming out of the weekly team call is that inventory is dramatically off plan.

But is this fair? The report is designed to talk in each department’s “language.” This works fine for optimization within each department. Merchandising know they are off margin plan by 0.5% and operations know they are off the hours plan by 3%. This works for their silos. But how does the executive team know where to focus their additional time?

Our recommended approach to a weekly report would look like this:

1. Get every department talking the same language. For this retailer it meant using costs.
2. Start with margin data, and convert the 0.5% variance into an absolute number. Percents almost always have different denominators which can cause significant misunderstanding, so we need to get the data into the common language of absolute costs. This retailer did $19.2 million in sales that week, so a variance of 0.5% in margin was a cost impact of $96,000.
3. Convert inventory into a cost number. For inventory that is a variance of $17.5 million inventory dollars multiplied by the estimated weekly holding cost of 0.3% to equal $50,000. Please note the use of the word “estimated”. One of the major reasons why retailers don’t put everything in costs each week is that the actual costs aren’t precisely known. There is no precise way to measure holding costs, and they vary by category, season and geography.This lack of precision has led to grave errors at many retailers. They give up relevancy and accuracy to gain precision. 17.5 million dollars is a nice, precise number. But it also prevents the executives from comparing costs from one department to another, making the executive team read the report the wrong way and leading to an incorrect decision. It’s what we call “being precisely wrong instead of directionally accurate”. This retailer can be extremely wrong in its holding cost estimates and still have a more relevant and helpful report than before. That fact is often lost.
4. 1. Finally, we convert labour hours into cost. This retailer had planned for 100,000 in hours, but ran 103,000 hours, or an absolute variance of 3,000 hours. To get this into costs, we need to estimate again. Most retailers have a good idea of their hours per week but not their cost per hour. Hourly costs are always changing due to the mix of part time, full time and overtime.

   On Monday morning, when everyone is rushing to get the reports out, there is no time to know for sure what the precise hourly costs are, so instead of estimating them, the analysts release the number of hours, which is the only precise number they have.

   But there is another solution: to use an estimate of the labour costs, which will be only directional, not precise, but will lead to an overall view that is more accurate. In this case the client agreed to use $15 per hour, and so a 3,000 hour variance multiplied by $15 worked out to $45K.

   Now, with all three departments using the same metric and talking the same language of absolute cost, the executive team can see what is actually driving performance:

After going through the recommended approach, the insight is much clearer. Great analytics isn’t about precision; it is about capturing a more accurate picture by using a wider set of different lens on the problem and using a common language, in this case by removing percentages and moving to a common denominator. The result is a great report for the executives to quickly see the true issues driving their business performance.