You know the situation when you ask your sales manager what your future sales will be. It doesn’t matter whether it’s next week, next month, or next quarter. These estimates will always be higher than reality. We know that very well, we’ve heard it a thousand times. Every time we hear it, we compare the sales forecast with the real sales numbers. And in almost every case, the sales manager has predicted higher numbers than the reality. These are what we call biased forecasts. For a wide variety of reasons, it’s human nature to overestimate sales performance, especially if you’re responsible for sales. It’s a repetitive, systematic error. The end result is that companies keep more inventory than necessary, which besides the obvious loss incurred also brings higher logistics costs.
To help you understand what a biased forecast is we can use a dartboard, which is really the best way to understand forecast bias. When you play darts, you have several attempts to throw those darts to hit the bullseye consistently. It’s the same with forecasting future sales. In the analogy, the bullseye represents your company’s real sales, so hitting the bullseye means 100% sales forecast accuracy. Now, take a look at the diagram below. When you see a pattern like this, you can immediately recognize some systematic bias in your attempts. There is a systematic error that is causing your darts, your forecasts, to be off, either slightly or more. Every throw, every prediction, is somewhere above and to the right of the bullseye.
The main difference between biased forecasts and unbiased forecasts is that the ‘dart pattern’ of an unbiased forecast shows dart throws spread equally around the bullseye, as seen in the diagram below. When you think about it, if you have to be ‘off’ slightly, this is a more ideal bias scheme, because if you sum the differences of the individual attempts, you get an average which is pretty close to your sales reality.
Forecast bias is one of those pesky issues that can really throw off supply chain planning. Even with the best predictive analytics and forecasting methods, bias still creeps in. But what exactly is forecast bias and how can you calculate it to correct planning mistakes? I’ll demystify forecast bias calculations so you can take control and improve supply chain forecasting.
What is Forecast Bias?
Forecast bias occurs when there is a consistent difference between the forecasted demand and actual demand. This bias can cause you to either consistently under-forecast or over-forecast demand across products, locations, customers, or time periods.
Some common patterns of forecast bias include:
- Optimism bias – Sales teams consistently over-forecast to be confident in hitting targets
- Sandbagging bias – Teams under-forecast to easily beat targets and maximize bonuses
- Recency bias – Putting too much weight on recent demand patterns
- Anecdotal bias – Relying too much on stories vs. data to shape the forecast
These biases creep in due to common human heuristics and incentives. But being aware of the patterns can help you spot and address them
Calculating Forecast Bias
To calculate forecast bias, you need to compare forecasted demand to actual demand over a historical period. Let’s walk through some options for calculating bias:
Simple Forecast Bias Formula
The simplest way is to calculate the forecast bias for each product/location/time period with this formula
Forecast Bias = Forecasted Demand – Actual Demand
If the result is positive, it indicates over-forecasting. If negative, it indicates under-forecasting
You can sum the values across products/locations to reveal overall bias direction and magnitude. Or calculate a % value relative to actuals.
This lets you quickly eyeball periods with significant over or under-forecasting.
Mean Absolute Percent Error
A very common statistical approach is Mean Absolute Percent Error (MAPE). This calculates forecast accuracy as a percentage of actual demand:
MAPE = (Σ Abs(Forecast – Actual) / Actual) / count of periods
MAPE helps normalize bias across products/locations with vastly different demand volumes. You can apply thresholds like green (under 10%), yellow (10-20%), red (over 20%).
Tracking Signal
Some supply chain experts recommend Tracking Signal to identify bias. Thisapproach sums the errors across time periods:
Tracking Signal = Σ (Forecast – Actual) for each period
Results above +4.5 or below -4.5 indicate an out of control forecast. 0 represents no bias.
This method highlights bias that accumulates over time.
Normalized Forecast Bias
My company Arkieva utilizes a Normalized Forecast Bias metric:
Normalized Forecast Bias = (Forecast – Actual) / Actual
With this approach, results stay bounded between -1 and 1. Values below -0.2 or above 0.2 indicate potential bias issues.
I like this simplicity of interpreting the normalized range.
Pick Your Preferred Method
The various options provide slightly different insights into bias. Pick one core methodology to stick with across your organization for consistency.
Also consider if you want to calculate bias for every SKU/location separately. Or aggregate across categories, brands, regions, etc. Different approaches can reveal bias at varying product/geography levels.
Addressing Forecast Bias
Once you’ve calculated forecast bias, here are tips for addressing it:
- Review metrics – Have planners review bias metrics and investigate root causes
- Adjust algorithms – Tune forecasting algorithms to account for bias patterns
- Realign incentives – Ensure forecasts aren’t skewed by management incentives
- Improve processes – Standardize procedures to limit individual bias
- Augment judgment – Leverage demand sensing/shaping between forecast releases
- Enhance visibility – Provide bias metrics on reporting dashboards
Taking steps to measure, understand, and correct forecast bias leads to major gains in forecast accuracy and demand planning. Quantifying the bias is an important first step.
Key Takeaways
Here are some key tips to remember:
- Forecast bias exists when forecasts consistently differ from actual demand
- Recognizing bias patterns like optimism, sandbagging, recency is key
- Simple formulas can quantify bias by comparing forecast vs. actual demand
- Pick a consistent methodology to calculate bias across products/locations
- Address root causes of bias through improvements across people, process, and technology
- Continuously monitoring bias metrics leads to better supply chain forecasts
While forecast bias is unavoidable, proactively measuring and correcting it can markedly improve demand planning. Incorporating bias calculations into your supply chain analytics toolkit helps keep plans on track.
With the right awareness and diligence, you can master forecast bias for a competitive edge. Now you have techniques to decode forecast bias and enhance supply chain performance.
Biased and unbiased vs. accurate and less accurate forecasts
While it is helpful to have some knowledge of biased and unbiased forecasts, that’s not enough. It’s also important to understand the difference between precise and less precise forecasts. Precision of forecast measures how much spread you will have between the forecast and your real sales. When you combine bias and precision, you can make a matrix. Obviously, the best result is an unbiased, highly precise forecast. The worst case is a biased, less precise forecast.
If you are tired of missing sales targets and holding unsold inventory for months, even years, it’s time for you and your company to try sales forecasting that is custom-made to be as unbiased and precise as possible. Try Inventoro for 14 days for free, and you will see the difference between your sales forecasts and those created by 21st century technology with more than 103 different forecasting methods including machine learning and deep learning and the latest in advanced AI technology.
Forecast Bias
How do you calculate bias?
Rick Glover on LinkedIn described his calculation of BIAS this way: Calculate the BIAS at the lowest level (for example, by product, by location) as follows: BIAS = Historical Forecast Units (Two-months frozen) minus Actual Demand Units. If the forecast is greater than actual demand than the bias is positive (indicates over-forecast).
What is the difference between precision and bias in forecasting?
The first distinction we have to make is the difference between the precision of a forecast and its bias: Bias represents the historical average error. Basically, will your forecasts be, on average, too high (i.e., you overshot the demand) or too low (i.e., you undershot the demand)? This will give you the overall direction of the error.
What are the different types of bias in forecasting?
A forecast can exhibit different types of bias. For instance, a positively biased forecast consistently predicts values that are higher than the actual outcomes, while a negatively biased forecast predicts values that are consistently lower.
How do you determine forecast bias?
One only needs the positive or negative per period of the forecast versus the actuals and then a metric of scale and frequency of the differential. Forecast bias can always be determined regardless of a report’s forecasting application.
