13. Selecting the Best Forecast

Selecting the best forecast.
We talked about the knife method, the cumulative mean, the moving average, and the exponential smoothing. As a forecaster, how do I know what to pick?
Well, you have to look at it’s accuracy and other things. But the first step you should take is look at the graph.
If you look at the forecast visually and see how it matches up against actual demand, there’s a lot you can learn, so that should be your first step. Then in the second step, take a look at the data. Take some measurements. Look at the mean. Look at the minimum. Look at the maximum. Calculate the standard deviation. Calculate the coefficient of variation.
That will tell you, how much uncertainly there is in your demand data. And then look at your forecast, how much variance does it explain? Now in general, when we are selecting a forecast, we’re looking at a couple of things. We want to look at the graph, make sure it’s visually closed. We want to make sure it’s unbiased, we don’t tend to over or under forecast. We want to be close. And most importantly, we want to avoid huge errors in our forecast.
So those four things are the critical issues when selecting a forecast. So now that we’ve identified the best forecasting method, all we need to do is hit go and sit back, right? Well not quite. Just because your forecast was the best method on past data, it doesn’t mean it’s going to work in the future. You can’t just sit back and relax now. You have to constantly monitor your forecast, and make sure that what you picked in the past is still the best method out there. It may change over time, and you have to be ready to switch methods at any moment’s notice. So your job is never done forecasting.

Jim Rohn Sứ mệnh khởi nghiệp