I’m a statistician by training, so numbers aren’t abstractions to me. They’re more than just digits—they represent real things.
In a training program at Ford in the late ’60s, I worked on model introduction. The program wasn’t about numbers; it was about selling cars to people. But we had to find the right numbers to incent them to buy cars and to incent the dealers to sell. We determined the optimal reduction in selling price below invoice, based on the model and certain specific options. The numbers were only vehicles, if you’ll pardon the pun. They weren’t the end goal, which is all that numbers often are to many techies. With results-oriented computing, you have to see the story those numbers can tell.
I took a job at an auditing firm in the early ’70s, in an era when there had been some major corporate defaults as a result of computer technology. One West Coast insurance company—not one of our clients, fortunately—had kept duplicate sets of books, and used computers to defraud their policyholders. Our auditing policies had to change thanks to the resulting backlash.
I was doing a review of a large firm where there was nothing wrong, but there was an appearance of wrongness: Two data-processing groups—one responsible for personal-trust accounting and the other for day-to-day processing—purposefully did not report to the same person. They didn’t really need to. Looking at it, though, we knew those numbers would appear to be an echo of what had happened on the West Coast. That was the first time I saw how clean numbers could seem dirty.
In the mid-1970s, I went to Acushnet, where the Titleist golf-ball division had a major business-process problem: Thanks to antiquated billing machines in the warehouses, there was a discrepancy between the number of balls shipped and the number we billed for. We didn’t even know how much we were underbilling—it could have been as much as 10%. We needed an immediate solution—otherwise, we couldn’t grow the business. The more we expanded, the worse the problem would get.
So we put in a new system, and we put it in fast. We took every shortcut possible and replaced more than a dozen separate systems. The first six weeks were an absolute disaster—the system worked, we were billing for everything we shipped, but it was god-awful slow. We needed 28 hours to process each day’s orders—so every day we got a little further behind. Rather than find an extra four hours in the day, we found faster hardware.
We took the correct approach—the expedient approach—but we didn’t sell it properly to our people. We didn’t take the time to look at other options. It took us three-plus years to get the system truly clean. But we got an immediate boost in revenue, since we were finally billing for what we shipped. That was the number we had decided was most important at the time, and by that measure, we were a success.
In retail, it’s been a challenge to deploy price-optimization systems, which are applications that determine when to reduce prices and by how much. Intuitively it goes against the merchant’s grain. The merchant thinks he knows better, based on his experience, rather than looking at the math and deciding simply because the system says, “Sales have hit the inflection point, take the markdown and take it deep, right now.” That’s a huge leap. But anyone can apply those mathematical systems—the algorithms and the statistics stay the same. It’s about choosing the right business rules for your company, the ones that suit your corporate strategy.
But numbers aren’t really absolute. They’re subject to interpretation. For paychecks, you want numbers to be 100% reliable. But for sales forecasting, the numbers can be an approximation. You just want to be as close as possible.
In the end, your success or failure is judged by your results. But people always forget that results are just another set of numbers.
Wollaston Morin recently retired from retailer Ann Taylor, where he was the top technology executive—a position he’s held at more than a half-dozen firms over the last 25 years.
