Humans have personal and political pressures that pull at them and, therefore, they are biased towards something.
As long as there are humans involved in making a forecast, the forecast will be biased.
The key to making a forecast unbiased is to find a method where humans have minimal influence on the outcome.
There should be only ONE forecast
The first rule of forecasting is to have a dialogue between Finance and business to lock one single forecast that is owned by both business and Finance. Having multiple forecasts is not an option.
Although Finance and business do not always agree on numbers, such an alignment can be supported through the use of unbiased forecasting when Finance lets the “data talk” and leaves bias out.
Unbiased Forecasting is a framework where Finance uses multiple methods to forecast which cannot be manipulated and as such are independent of personal opinions. These are the methods where historical data, market data, statistic or an industry index are examined, and forecast algorithms are applied to predict future outcome.
In this article, we will investigate five of those methods. Each of them has its own pros and cons.
Crowd forecasting does not rely on a single individual but a whole group of people providing their own individual views of the future, thus, reducing the bias component.
The variability of individual performance makes it hard to know which individual to trust. If you aggregate the forecasts of a crowd of people, you are much more likely to come up with a more accurate forecast on average.
With crowd forecasting, there are several pros and cons to consider:
- Pro: It is a relatively easy starting point towards creating forecast validation models.
- Con: It is not unbiased.
- Con: The outcome is only as good as the average of the crowd’s opinion.
Understanding the competitive landscape can be used as an unbiased gauge to assess a forecast. Using the quarterly announcement from public companies is a guide to forecasts for the following quarters. If Finance has noticed a pattern between its own performance and its competitor’s performance, those insight can be used to provide an unbiased forecast.
There are also some pros and cons with this method:
- Pro: It is a way for Finance to utilize external data sources to validate its own performance.
- Con: A perfect competitor doesn’t always exist.
A regression uses the historical relationship between an independent (often time) and a dependent variable such as sales, revenue, etc. to predict the future values of the dependent variable.
- Pro: Regression can be performed with most tools including Excel as a simple scatterplot and adding a regression line.
- Pro/Con: This is a good option if the data is relatively linear, exponential, logarithmical, etc., but the framework cannot be used if the dependent variable is seasonal.
- Con: Regression should only be used to forecast if the R-square is higher than 0.85.
Smoothing & Moving Average
Smoothing and Moving average covers a number of different methods including ARIMA, Holt- Winter, etc. These models are statistical techniques using historical time-series data and applying algorithms to predict future outcome.
- Con: Running these models in Excel can be very time consuming.
- Con: To select the best model, all models need to be run. This means that a biased human will select the one he or she deems best.
- Pro: Analytics tools have major advantages to Excel and can be used for this method. They utilize artificial intelligence to find the best method among all models that has the best predictive capabilities.
Leading indicators are industrial and economic metrics from which an indication of the value or direction of another variable (for example, a sales forecast) can be obtained.
They are called "leading" because their direction or magnitude historically "leads" the focal variable. For example, we may find that unemployment rate indicates (leads) the future of a company’s revenue.
- Con: Leading Indicators can be extremely difficult to find when using Excel as you need to search through volumes of industrial and economic statistics to find correlations with your performance.
- Con: If a relationship is found, it could still change over time, so the leading indicator will no longer be leading.
- Pro: This forecasting method works best with analytics tool with artificial intelligence.
Achieving Better Forecast Accuracy
If Finance would like to achieve a higher forecast accuracy, it needs to:
- Use unbiased forecasting together with the business forecasts. Where multiple forecasts agree within a planning tolerance, there is a higher confidence of outcome, and where they do not, there should follow a deeper discussion to sort through the risks.
- Use a variety of different unbiased methods. They are built from different data sources including crowds of people, competitive intelligence, statistics and mathematical algorithms.
- Expand its toolbox beyond Excel to include advanced analytics systems and techniques that can challenge and influence the business forecasts. The more advanced the tool in your toolkit the more advanced Unbiased Forecasting you can produce.
The ability for Finance to provide unbiased recommendations both improves forecast accuracy and helps Finance become a strategic partner to the business.