The word ‘feedback’ is in common usage, and people often talk about ‘closing the loop,’ but few people realise what these terms really mean…and how important they are.
Moreover in the world of forecasting there is a conspicuous failure to close the feedback loop, with the result that the performance of most forecast processes is way below managers have the right to expect; particularly if they have just paid a king’s ransom for a fancy new piece of software.
The best way to explain this is by describing what feedback isn’t using a simple example.
Imagine driving a car with no instruments. At the end of every day, you receive a piece of paper telling you how many miles you have driven, your average speed and engine revolutions and the total amount of fuel consumed.
Would you feel well informed?
I’m guessing not, because you really need to know whether you are breaking the speed limit, about to run out of fuel and whether you are thrashing the engine. Receiving a high-level summary when it is already too late to do anything about it, is useless. And yet this is the form that most forecast process feedback takes; high-level, paper reports at the end of a period. Worse still, the drivers don’t have any of the contextual information that a car driver has; noise from a struggling engine, the scenery flashing past and so on.
The problem here is that, while there is management information, there is no feedback.
The classic definition of feedback is information about the state of a system that is fed back into the system in a way that can be used to change the state of the system. In layman’s terms it is information that can be used to take action that will bring the forecast process (the system) back into line with its goals.
No system or process can possibly perform consistently in the desired way without adequate feedback - much of the improvement we have seen in everyday machines that now perform better than they ever used to, is because feedback has been built into the system. So automobile engines now have engine management systems to regulate their performance and fuel efficiency, and street lights turn on when it gets dark, saving both money and lives. Even bombs have become smart by detecting when they are off target and taking corrective action.
Building feedback into forecasting processes means that we need very frequent, very granular information about the quality of forecasts.We need to be able to tell the difference between random noise and evidence of a real problem and to compensate for the inherent difficulty of forecasting a highly volatile data series compared to a very stable one. Finally, it should be able to focus attention on those areas that really matter; where the gap between actual performance and that which is achievable is the most significant.
If we want to accelerate the returns on our investments in forecasting software and ‘best practice’ processes, we need to learn how to ‘close the loop’ on forecasting. Without it forecasting will remain one of the biggest sources of waste in business.