Forecasting and Financial Modeling

All forecasts are wrong. What a statement to start with, but it is true. If we could depend on forecasts as a correct indication of certain future states, our world would be totally different.  We would know with certainty what the future would bring and this would have a profound effect on how we would behave. So when an entrepreneur approaches financial model forecasting it is a “best guess” rendering of what the company might look like in the months or years ahead.  So as a guess, we must therefore not rely on these forecasts as factual renderings.

What financial models do provide is the opportunity to paint, through numeric terms, what could be in store for the company and also assist in guiding decision making.

A financial model requires many assumptions to be made about the future.  Future demand, operating and administrative costs, interest rates, inflation, and inventory levels are a few items that are inputs to the financial models of the company. Therefore, a financial model can only be as good as the assumptions and forecasts that are put into the model for analysis.

Some interesting and famous forecasting statements provide an indication as to how difficult it is to forecast new or emerging technologies:

• “Theoretically, television may be feasible, but I consider it an impossibility–a development which we should waste little time dreaming about.”
  – Lee de Forest, 1926, inventor of the cathode ray tube
• “I think there is a world market for maybe five computers.”
  – Thomas J. Watson, 1943, Chairman of the Board of IBM
• “This ‘telephone’ has too many shortcomings to be seriously considered as a means of communication. The device is inherently of no value to us.”
  – Western Union internal memo, 1876
• “Computers in the future may weigh no more than 1.5 tons.”
  – Popular Mechanics, forecasting the relentless march of science, 1949
• An official of the White Star Line, speaking of the firm’s newly built flagship, the Titanic, launched in 1912, declared that the ship was unsinkable.

The last quote is a perfect example of relying too heavily on the assumptions. Since the future is almost always uncertain, relying on point forecasts and ignoring variability is the first and foremost mistake made in forecasting. A ‘point forecast’ for example is an assumption that sales for the next quarter will be exactly 400 units. How many of us would bet that this number is going to be correct particularly in fluctuating markets. All that can be done is to make the best effort in thinking about as many influencing variables as possible; but even then there are no assurances that a forecast will be correct.  Here is some “food for thought” that might help.

There are several ways to incorporate variability into your forecasts so as to arrive at better assumptions. An introduction to these methods follows.

Qualitative Methods

These methods are useful for when the future is highly uncertain. For example, forecasting demand for new and innovative products would benefit from following certain qualitative methods. A number of different qualitative forecasting methods are:

• Executive judgment
• Market research through customer visits, surveys, etc.
• Panel consensus – whereby managers, executives, key decision makers, and customers can reach consensus about future outcomes. A good tool that can facilitate this is the ‘Delphi Method’. This is a facilitated system that structures the anonymous contributions made by a panel of experts on a certain subject area. It usually results in consensus around the subject matter after a few rounds of contribution by the experts.

A major mistake is to use point forecasts, derived from any of the above methods, directly in the financial models. A measure of uncertainty has to be captured so as to mitigate the underlying risks of point forecasts. One way of going about and doing so would be to use some sort of probability distribution theory (such as the normal distribution). Going into the mechanics of these models and theories will be out of the scope of this article.

Quantitative Methods

These methods are most useful for products which have longer life-cycles and are less prone to uncertainty. The major assumption behind these methods is that future outcomes will be based on, or is a function of, historical outcomes.

The following are some of the most popular quantitative techniques used for forecasting:

• Time series analysis
  • Moving average – This method smoothens past data fluctuations and can be used to base future outcomes on the most recent past data.
  • Exponential smoothing – Very similar to moving averages but also automatically assigns weights to past data. Double exponential smoothing is an extension which is used when there are trends in the data.
  • Regression analysis – This method is useful for trying to find relationships in data, such as trying to explain dependent phenomena through independent ones. The major use of this method is to understand trend and seasonality.
• Causal forecasting
  • This method tries to understand and discover correlations between the outcome to be forecasted and other factors such as price discounts, promotions, economic factors (e.g. interest rates), marketing communication campaigns, etc.

Tips and Takeaways

Forecasting is mainly based on past observations and historical data and is therefore prone to inaccuracies. There are scientific and proven ways to improve forecasting, some of which were introduced above. The following is a list of useful tips for entrepreneurs relating to financial modeling forecasting:

1. Point forecasts are almost always incorrect and variability needs to be built around them. This can be done by building different scenarios and assuming different future outcomes for each respective scenario. Then assigning probabilities to each of the expected scenarios. This simple technique results in a probability adjusted forecast that can be much more accurate than the point forecast.
2. The longer the forecasting period, the more inaccurate the forecasting becomes. Therefore, do not put much weight behind forecasts three years or more.
3. Forecasting aggregated data (such as total sales of all branches of a company) is always more accurate than forecasting disaggregated data (such as sales of each of the company’s branches). The reason is the more micro you go the higher the probability of error because many more variables have to be assessed.
4. Always use sensitivity analysis to predict for the extreme future outcomes. This will allow for proper planning and preparation for the worst, best, and most likely future outcomes.

Comments are closed.