If it’s a “bread and butter” item with a high rate of sale, is 80% accuracy good enough? What level of accuracy should I expect on a seasonal product? Or a slow mover? Or an item that’s promoted frequently?
The Myth of Forecast Accuracy
The problem with this line of questioning is that there will always be random error in any forecast you produce. To try to quantify an “acceptable” level of accuracy on an item requires us to quantify random error, which by definition is unknowable.
You’re not in school anymore. There is no grading system for forecasts. Depending on the circumstances, sometimes 80% accuracy is a poor result and under a different set of circumstances 30% accuracy may be good enough.
Your achievable level of accuracy largely depends on your forecasting process. If you are consistently following a good process, then your accuracy will be as good as it possibly can be.
The Ten Commandments
The Ten Commandments are recognized by Judaism, Christianity and Islam as an important set of rules for guiding human behavior. No matter what your religion, if you followed the commandments, particularly the ones pertaining to how we interact among ourselves (Thou Shalt Not Kill, Thou Shalt Not Steal, Thou Shalt Not Bear False Witness, etc.), then you could safely say that you’re not an evil person.
Over the years came many more laws based on the Commandments – some of which
contradicted each other. It became more and more difficult to know how to be good.
Then, about 1,000 years after the original Commandments were given to Moses, a young carpenter from Galilee boiled them down nicely into a single Golden Rule:
“Do unto others as you would have them do unto you” (or, alternatively, “Treat others as you yourself would like to be treated”).
Follow this one rule, and you have six of the Ten Commandments covered.
The “Golden Rule” of Forecasting
It’s fairly well documented that good forecasting processes have the following
- Free of procedural bias – This means that there aren’t steps designed into the process itself that cause habitual over-forecasting or under-forecasting.
- Free of organizational bias – In other words, performance measures are aligned with the goal of your operational forecasting process, so that people aren’t given the incentive to intentionally “sandbag” the numbers one way or the other.
- Doesn’t carry a risk premium – This is often the toughest to detect, because it’s caused by the best of intentions. In a nutshell, this is the practice of forecasting “a little bit extra, just in case” to guard against service failures.
- Accounts for both timing and quantity – You can’t schedule your supply chain around an operational forecast that says you’re “going to sell 10,000 units over the next 8 weeks”. It needs to be very specific with respect to timing and location as well as quantity to flow product efficiently.
Like the Ten Commandments, volumes of material has been written in support of these rules. But if your organization follows one simple “Golden Rule of Forecasting”, you won’t have to worry about the quality of your forecasts again.
The $500 Rule
The $500 Rule states:
“The operational forecast that runs your supply chain is the one on which you would bet
$500 of your own money.”
(Note: We chose to make it the “$500 Rule” because $500 is a number that everyone can relate to. But it’s also big enough so as not to be considered “chump change”.)
Think about it. Suppose you were in a forecasting competition with a $500 entry fee. You can use any resources and information at your disposal. But for a particular item in a particular week, whoever comes closest, in absolute terms, to the actual number wins the pot.
If these were the rules of the contest, would you be inclined to take steps that inflate the number beyond what is realistic? Would you simply average a big number across a number of weeks? Would you forecast what the facts are telling you “plus a little bit extra”?
Like we said earlier, there’s no antidote for random error – it will always exist and you can do nothing about it. But if you spend your energy on coming up with forecasts that are as close to the “truth” as possible (i.e. you’d plunk down 500 bucks of your own money), then the battle is half won.
The great part is that, to your organization, the benefits of unbiased forecasting really will make $500 look like chump change.