When Everybody’s Playing Moneyball, Nobody Is
Forecasting Nightmares May Muddle the Analytics Revolution
Keith Riegert, Co-Editor-in-Chief
BUSINESS ANALYTICS IS THE FUTURE. That’s the refrain I hear constantly, along with companies asking how they can “get analytics for [their] data?” The problem with today’s data trend is that if everyone is chasing the same signals in an industry, any statistically gleaned advantage washes away—but firms will spend a lot of time and money figuring that out. However, if you want to find out more about business analytics it is worth a look.
Companies today are swimming in numbers; there have been enterprises hailed as “the Moneyball of” investment banking, movies, publishing, health care, advertising, etc. They’re all generally placing big bets on patterns emerging from meticulously collected (or expensively bought) data sets. But this explosion of analytics-driven business strategy has created a troubling reaction—accurate forecasting, crucial to everything from calculating consumer lifetime value and projecting future earnings to drafting growth strategies, is becoming an ever-more uncertain task that businesses are implementing intelligent automation software to sift through their massive data banks to try and make sense of the data they actually have.
Take Whole Foods as a prime example—the Washington Post recently covered the company’s announcement of their sixth-straight quarterly decline in same-store sales, forcing Whole Foods to slash its outlook for 2017. It was quite a turn around from October 2015, when the chain optimistically projected two-fold growth in its number of stores. Instead, it’s shuttering more locations than ever before.
One clear culprit in the decline: analytics-driven decisions made by competitors have taken an enormous bite out of its share of the organic food market. Companies like Walmart and Costco, names never before synonymous with “health food,” have captured billions of dollars in the organics niche.
An Analytics Time Bomb
The problem that Whole Foods is contending with should be one familiar to just about every industry around the globe. While the Austin-based market was successfully mining its wealth of proprietary customer data for a long-term game plan, every competing market chain in the country had access to the trends—the gist of the situation. With third-party firms providing entire industries with all the data needed to see what’s working (and to take action), billion-dollar insights can be reduced to commodity knowledge remarkably fast.
I first encountered this issue in the book publishing industry, where I found countless publishing houses were acting on the same trending topics simultaneously. Traditionally, a book’s sales-over-time trajectory follows a trend of exponential decay (a familiar pattern found across media industries). And taking the log of those sales allows you to produce a strong linear regression that is ideal for constructing a forecast for the book’s lifetime performance:
The problem is, once you move from, say, a unique, author-driven book to a title for which sales may be driven by a replicable niche topic, the sudden, data-driven entry of competitors can cause a reliable model to completely fall apart. Take this example of the logged sales pattern of a book plotted against the number of competing entrants into the topic niche. As the number of competing titles grows exponentially, the linear relationship collapses:
If you look around, you can see similar situations playing out in any industry. (For example, analytics-driven decision making is most likely a factor in why you can catch a superhero show this evening on Fox, Netflix, FX, NBC, ABC and the CW). Of course, chasing consumer trends is not a new strategy brought on by the analytics revolution—it just appears that the numbers boom may be leading to more rapid homogenization of products, market saturation, consumer exhaustion, and, ultimately, trend collapse.
Protecting Your Company from Bad Forecasts
Now, more than ever, building in uncertainty for competitive response in forecasts is critical. Here are the first five steps I always take in analyzing the viability of an analytics-driven decision or product:
1. “Proprietary” probably isn’t proprietary. Making decisions based off your own KPIs and a multi-million-dollar, protected data warehouse may appear to be providing unique insights, but always expect that your competitors are across the street analyzing similar data sets. If they are, chances may be you’ll both be catching the same patterns.
2. Prepare to adapt your forecasts. This is especially important if you’re experiencing a windfall-generating success—that boon will be picked up faster than you might normally anticipate, so build in a worst-case scenario for rapid market saturation. Weight heavily.
3. Mining a public data set? Join the infinite club. Analytics insights from professional firms like Nielsen are great for informing business strategy—just make sure that you and the decision makers at your firm go in assuming everyone else is drawing the same brilliant insights.
4. Make sure you’re working to a common goal. Without ensuring that everyone in your business is working towards the same target, you’re not going to be able to make the progress you want to see and that will be reflected in your forecasts. Make sure you have a good system to manage all the different goals you’re working on and use something like OKR to improve performances. So, What Is OKR? Well, following that link will tell you more about the system.
5. After the data, a heavy dose of creativity. Making your analytics-borne position as defensible as possible means relying just as heavily on innovative thought as strong data—a deliberate, meaningful twist to what the numbers are telling you can make all the difference between staying above the incoming crowd and getting sucked into the soup.
. . .
Find more from Keith at www.keithriegert.com