Why Sales Forecast Accuracy is a Silly Goal

Here's a great interview [1] with Rob Hyndman, one of the world’s leading authorities on forecasting. Rob defines forecasting as:

An estimate of the probability distribution of a variable to be observed in the future. 

That’s worth pausing on.

Forecasting is not about accurately predicting a single number. It’s about understanding the probability distribution of possible outcomes.

A probability distribution shows the range of possible outcomes and their relative probabilities.

Hyndman’s Conditions of Forecast-ability

Hyndman highlights five conditions that make something forecastable:

1. You understand the drivers of variation.

2. You have lots of data.

3. The forecast doesn’t affect the thing we’re forecasting.

4. The future is similar to the past.

5. There’s relatively little unexplainable variation.

Of Comets and Exchange Rates

Hyndman gives two examples:

Comets

If you want to forecast the path of a comet, you can tick every item: the drivers of motion are well understood, data are abundant, the comet won’t change course because you made a forecast, the past is a good guide, and there’s little randomness. Easy to forecast. For most purposes, there is not much need for a distribution.

Comet trajectories are very predictable. Haley's comet in 1986. (Image credit: NASA).

Currency exchange rates

But try forecasting the USD to AUD exchange rate. You have a lot of data, but politics, human psychology, shifting global events—all come into play. And respected forecasters actually affect the rate. That fails on all but the second condition. Hard to forecast. Understanding the distribution is critical.

Currency exchange rates are hard to predict. By Emilian Robert Vicol from Com. Balanesti, Romania - Money-Euro-USD-LEI_53073-480x360, CC BY 2.0, https://commons.wikimedia.org/w/index.php?curid=38383407

Does Sales Forecasting Pass the Test?

Now, take Hyndman’s five conditions and hold them up against sales forecasting.

• Do we understand all the drivers of variation in a sale?

• Do we have deep, reliable datasets—or are they limited, sparse, or sketchy?

• Can salespeople and managers respond to the forecast and influence the outcome?

• Does the future always look like the past? Or can new products, competitors, shifting markets, and personnel changes make the past less relevant?

• Are there random factors that introduce variation? What if your key sponsor leaves the business, or your customer gets bought? Are these factors knowable or random?

  
  

Noodle on that. Sales forecasting is a lot closer to predicting a currency exchange rate than it is to predicting a comet. It is possible to predict sales, but there are limits. And understanding the distribution is critical.

What Most Forecasting Processes Get Wrong

Most forecasting processes aim to give you:

• A point forecast (e.g., “We’ll close $10.2M”).

• Or a handful of points—hi, mid, lo.

  
  

But what do those points mean? What’s the probability of hitting the “mid”? How likely is it you’ll beat the “hi”? How much risk is there of falling below the “lo”?

Without those probabilities, those figures are context-free and effectively meaningless.

A Better Way Forward

Instead of focusing on accuracy, focus on realistic forecasts based on statistical methods:

• Estimate the odds of exceeding any number.

• That means you can ask: “What’s the probability we beat plan?” or “How likely are we to hit $12M?”

• Don’t pretend forecasts are perfect. If your forecasts are grounded in classical statistical methods, they should be pretty good. But that’s not the point.

A forecast distribution showing the estimated possibilities of beating three different figures. The distribution can show the range and relative probabilities of any outcome.

Sales has too many uncontrolled factors to make it consistently forecastable. So it's critical to understand the forecast distribution and the effects of changes in the factors you control.

Forecast accuracy is not the objective. The forecast is the starting point for sales optimization. We use the forecast to improve outcomes by showing you where to focus, what risks to manage, and how to increase your odds of success.

The Bottom Line

Sales forecasting doesn’t meet Hyndman’s criteria for easy forecasting. Distributions are critical to making sense of the forecast, and accuracy is a silly goal. Most important, because you have some control of the outcome, accuracy is not even a desirable goal.

Instead, the right goal is to answer this question:How can we use forecasts to inform corrective action and improve what is likely to happen?

That’s the difference between contextless numbers—and a forecast that actually helps you sell more.

---

[1] https://www.buzzsprout.com/1641538/episodes/7708129This discussion starts at ~8:45.