News Landing Page

APICS News

Accurate Forecasts? Precisely!

By Jonathan Thatcher, CSCP | 23 | 2 |

Tips for hitting your targets every time

Reader B.K. writes, “We struggle to reach the accuracy we need in our forecasting. Sometimes, our forecasts are almost perfect; other times, we’re not even close. I know there’s an old expression that says every forecast is wrong, but is there something fundamental we may be missing?”

Start by considering the two ideas of accuracy and precision. Accuracy and precision may sound like the same thing, but they are distinct entities in the science and business fields. Accuracy measures how close estimates are to the true or desired value. Precision measures the degree to which work arrives at the same results over many iterations with similar conditions. 

To illustrate the difference, consider the game of darts. A dart that lands close to the bull’s-eye is accurate, while darts that land near each other are precise. In darts, both accuracy and precision are desired, and the expert player achieves both. In forecasting, as with darts, practice improves performance; but, often, more in one area than the other. A dart player who develops precision more than accuracy will throw darts that cluster together, but not necessarily near the bull’s-eye. 

Imagine a highly precise but inaccurate weather forecaster. A good forecast still can be achieved, as long as a reliable bias is determined. On the other hand, a highly accurate but imprecise weather forecaster is not very useful, as it’s impossible to tell which days will be right. Notably, a highly accurate but imprecise technician may be able to repair a complex machine, even if it means making several attempts.

Consistency is key
To begin applying these ideas to your own forecasts, first consider if your organization’s needs favor accuracy or precision. Your current practices may be compensating for weaknesses in one area by making presumptions about strengths in the other. For example, forecasts with greater precision than accuracy may cause team members to presume hidden bias. Alternatively, forecasts with greater accuracy than precision may prime expectations for greater variability. 

The next step is to find the root causes of any bias and address them. Let’s look back at the dart game from earlier. What causes distance to form between darts or the bull’s-eye after each throw? In other words, for each cause, what is harmed more—accuracy or precision (or both)? Carefully evaluate the factors that stimulate or suppress each cause. If the root causes are variable, try to find a relationship between each cause and the degree of inaccuracy and imprecision. 

Look for consistency where you can find it. Identify relationships between the elements of forecast data. For example, does competitor activity tend to move in the same direction as market demand? Or are there areas of useful precision—whether accurate or not—in partner data? Business intelligence and analytics systems may be able to help. 

Once you have identified the troublesome area, consider the following strategies for enhancing either accuracy or precision in your forecasts:
  • Enhancing accuracy. Identify the variables and misperceptions of your forecast, as well as any spots where you lack awareness to properly account for these. Where accuracy is low but precision is high, look for persistent causes, such as bias or a lack of suitable technique.
  • Enhancing precision. Investigate a variety of causes and influences in your forecasts. Identify inconsistent data or measurements, hidden sources of uncertainty in key areas, and varying levels of confidence in input data.

In either case, it may be helpful to work with a trusted forecasting advisor to guide you through the process. These individuals take a dispassionate view of the situation and make recommendations for improvement—some of which may seem unnatural or nonintuitive. A prime candidate for a forecasting advisor might be a senior manager, consultant, gifted colleague from another department, or an instructor who understands your operating environment. 

There are many avenues to explore in improving accuracy and precision. Start now, and see what you find. Sometimes, applying simple, fundamental ideas to complex processes helps reveal the true causes of difficulty.

Jonathan Thatcher, CSCP, is director of research for the APICS professional development division. He may be contacted at askapics@apics.org.

2 Comments

  1. 1 Hong 10 Dec
    I am confusing, too. I tried to understand the gap between accuracy and precision from your article, but no idea. Could you help to explain more?
  2. 2 Miriam Fontaine 10 May

    Dear Jonathan,

    I have just read your article with interest as I am currently setting up and deploying a company-wide forecast process (distribution, not manufacturing). Unfortunately, I do not understand the difference you are trying to highlight between forecast accuracy and precision. It may be that I am not yet fully awake after a 2-day French public holiday and insufficient coffee, but I have read the article twice and still do not grasp what you mean by precision. Do you simply mean accuracy over time?

    Any additional explanation you can provide would be much appreciated.

    Regards,

    Miriam Fontaine

     

     

Comment

  1.