The monster called ‘Forecast’

Niels Henrik David Bohr - a Danish physicist one famously said – “Prediction is very difficult, especially about the future”.  A malapropism as it may seem; it inadvertently forces me to think about one of the most discussed topics in a company – the Forecast.

Forecast is a monster because of the sheer impact it has on almost all the segments of a company.  It is a monster that some companies manage to understand and tame better than others and draw direct results in the company’s bottom-line if not the top-line or both.

An inaccurate forecast can toss capacity plans, production plans, purchase plans, staffing plans, cost-working, working-capital requirements haywire overnight. An inaccurate forecast can lead to lost opportunities or huge losses. However some companies that build-to-order (start production after receiving a customer order, example-  Dell) are impacted less compared to those that build-to-stock (who produce in response to an anticipated forecast, example- Unilever).

It is important to note the following characteristics of forecasts:

1)      Forecasts always include two components: the expected value and the forecast error. The ‘expected value’ is simply put the most plausible estimate of future sales while the ‘forecast error’ is the deviation from the expected value (+ve or –ve) based on various factors such as seasonality, trend, market share movement, economic scenario etc.

2)      Long-term forecasts are usually less accurate than short-term forecasts; that is the long-term forecasts have a larger standard deviation of error related to the mean than short-term forecasts.  In this sense sales forecasts aren’t very different from other things in life where near term predictions are more reliable than long term predictions.

3)      Aggregate forecasts are more accurate than disaggregate forecasts; that is they have a smaller standard deviation of error relative to the mean.  It is easier to forecast the GDP of India than it is to forecast the revenues of a company. Likewise it is easier to forecast the revenue of a company than to forecast the sales of a single product from a portfolio of products that a company may have.

4)      The farther up a function is in the supply chain (or the farther it is from the customer), the greater is the distortion of information it receives. This is also called the Bullwhip effect where the order variation is amplified as the orders mover farther from the customer. As a result the farther up the function (example raw material sourcing) the larger is the forecast error.

There are numerous factors that a manager must take into account while forecasting and more importantly while choosing the forecasting methodology (maybe a topic for future discussions).  Some of these factors are

ü  Past Demand

ü  Past Forecast error

ü  Supply lead times for products

ü  Planned advertising campaigns

ü  Macro-economic environment

ü  Competition

As much as scientists would love to create a perpetual motion machine, they are hypothetical and would violate laws of thermodynamics. However the pursuit of these machines is popular in an effort to achieve the ‘as-near-as-possible’ to the ideal. Similarly there can be no such thing as a ‘Perfect Forecast’, but our pursuit of the same shouldn’t end. As a result we can achieve more accurate forecasts and more desirable bottom-lines.