How to Forecast When There Is No Historical Data

In an earlier post, I quoted from the Wall Street Journal:

In March, Best Buy Co. said it could have sold more electronics equipment in the three months ended Feb. 28, but its suppliers' deep cuts made it tough to keep shelves stocked. Suppliers "all decided to build a lot less," says Best Buy merchandizing chief Michael Vitelli.

In that post, we looked at the problem from a risk behavior point of view. That is, the intensified risk-averse behavior of suppliers in the turbulence of an economic downturn, combined with the lack of visibility, could push suppliers toward cutting the production volume more than what is necessary.

Here, I want to return to the same problem, but from a different angle: the inability of firms to make proper demand forecasts in volatile market conditions. The traditional demand forecasting relies on historical data. That is, we look at the trends and seasonal factors of past data to predict the future demand.

So, what should we do when the available historical demand data does not match the current unprecedented market conditions, and therefore cannot be used to predict the future demand?

The lack of suppliers' ability to make proper demand forecasts, combined with their conservative attitude, could result in heavy production cuts.

This problem is not limited to economic crises and their unforeseen market conditions. When a company launches a new product, similar problem could arise: no historical demand data exists for the new product. This forecasting challenge is specifically more common in industries with short life-cycle products, such as consumer electronics and fashion.

In these situations, we are seemingly left with no choice but to rely on our market intuition, which, of course, could be quite risky. Below, I intend to talk about two alternatives which can be used for demand forecasting in the absence of proper historical data.

Paul Saffo in his Harvard Business Review article "Six Rules for Effective Forecasting" suggests when the level of future uncertainties is beyond a certain level, the best approach might be patience. That is, the forecaster should wait until the market conditions settle down and some of the uncertain parameters reveal themselves. Listen to Saffo talks about his article in an HBR IdeaCast here.

Rita McGrath and Ian MacMillan in their book "Discovery Driven Growth" suggest a more proactive approach. In their discussion on how to plan for future growth, they argue when a firm plans to walk into uncharted fields as a growth effort, it is usually difficult to find historical date to predict the future outcomes.

The authors prescribe a trial and error approach to deal with the problem. That is, firms should plan for a set of controlled failures. In other words, firms should make small steps in different directions and should be tolerant of the failure in many of them in order to find the direction which leads to success.

McGrath talks about her book in an interview with Harvard Business IdeaCast. You can watch this interview by clicking here.