Inaccurate sales forecasts are the bane of many managers’ lives. And they can be costly too, with disappointed customers who will never do business with you again or warehouses stuffed with goods that no one wants. Back in 2001 Nike allegedly suffered $100 million in lost sales when their forecasts told them to order $90 million worth of shoes that sold badly and to cut back on orders for popular sneakers like the Airforce 1. We can seldom predict sales with perfect accuracy, but there are common mistakes that company forecasters can avoid.
If you feed computers with good quality software they will crunch through mountains of data, producing forecasts that make optimal use of information buried in sales histories and market research data. But, like Henry Ford, who believed history is ‘more or less bunk’, many managers cannot resist the temptation to jettison data they think is past its use-by date. ‘Back then the trends were different,’ I’ve heard them say when ‘back then’ is a mere couple of months or years ago. Modern statistical forecasting techniques can adapt to changing trends, while also exploiting relevant past patterns that have remained stable over time. Mistake 1 is to dump most of your data.
Despite the power of computers, there are times when judgmental intervention is justified. A forthcoming sales promotion perhaps, a hike in VAT or a new competitor – they are all things the computer might be unaware of so it makes sense to adjust its forecasts up or down. However, people can seem addicted to overriding computer forecasts. In a food company I visited over 90% were changed for no apparent reason, except perhaps to justify the forecasters’ role. All this effort served only to damage accuracy. Mistake 2 is over adjusting forecasts. Reserve your interventions for important future events.
Politics is a more insidious reason why people change forecasts. Managers may deliberately keep their forecasts low so they can proudly announce each month that they have exceeded their forecast yet again. Alternatively, a high sales forecast can bring kudos because it will please the boss or attract a chunkier budget to one’s department. These machinations have little to do with genuine expectations of future sales. It can be hard to avoid, but mistake 3 is mixing politics with forecasts.
Sometimes there is confusion about what a forecast actually is. It’s not a target -that’s a sales figure you set to motivate people -not necessarily what you think will happen. Nor is it a decision. Staff working for one major retailer were confusing the sales forecast –the computer’s estimate of the most likely level of sales -with their decision on how much stock to hold. If the computer forecast sales of 200 units they might decide to stock 220 units in case of unexpectedly high demand. But they then referred to the 220 as the forecast. Other managers thought that this was the most likely level of demand and complained how awful these apparent forecasts were. Mistake 4 is to confuse forecasts with targets and decisions.
In some companies, groups of managers meet to approve computer predictions or pool their judgments on the future prospects for sales. But psychologists tell us that group dynamics can result in strange outcomes. In cohesive groups, or where the boss is at the helm, members may be reluctant to ‘rock the boat’. As a result, judgments can coalesce around highly implausible forecasts, as everyone rushes to support the prevailing view. Allowing this phenomenon, known as groupthink, to distort your forecasts is mistake number five.
The final mistake is being deceived by randomness. Consumers’ whims, accidently broken products that need replacing, advertisements that by chance catch a person’s eye and a host of other factors mean that a proportion of sales will be unpredictable. Despite this, we can be too willing to junk forecasts that are as accurate as possible because they have not been spot on. We see illusory patterns in sales graphs we think the computer has missed and wrongly think a freak sales figure is a sign of fundamental change. Even the best forecasts will usually differ from actual sales. But they will perform much better than forecasts that are constantly revised in a futile attempt to capture every random twist and turn in sales.
Avoiding these mistakes will not guarantee perfect forecasts. But you should see accuracy improve -and that’s a confident forecast.