 |
|
WORDS FROM WALLACE - MARCH 2007
HOW ACCURATE SHOULD
YOUR FORECASTS BE? - PART 2
Last month, we talked about why focusing on forecast accuracy was not an effective way to improve the forecasts.
We said that it's far better to go after forecast error. (If you missed
last month's newsletter, click here.) Forecast error gets us one step
closer to treating forecasting as a process, with inputs, a conversion
step, and outputs.
|
|
It's time for a three-part quiz. Here are two products, A and B. The forecast for
each is 20 per month.
The first question - which is more accurate - is the most difficult to
answer. Certainly Product A's actual sales are
closer to the forecast, and thus one
might say it's the more accurate. On the other hand, Product B's errors tend to cancel each other out over time, so perhaps it's the more accurate.
There is no good answer to this question. I suppose it's whatever you want it to be. (If you want to dig into this issue a bit deeper, click here to
see Chapter 4 of our book: Sales Forecasting: A New Approach.)
Let's move on to the next question: which is potentially more damaging?
Here the answer is very clear: Product A. It has far greater potential
to
disrupt the supply chain, because it can lead people to over produce,
under produce, over order, under order, over replenish, under
replenish, and make second-guessing the forecast a way of life. Product A's forecast is consistently wrong in the same
direction. We say it contains bias - a build-up of forecast error in
one or the other
direction. The technical term: Running Sum of Forecast Errors (RFSE).
|
|
Bias is the pits. It's the biggest villain in the forecasting world, because it leads people to do the
wrong things.
The third question: which is easier to fix? Here also the answer is
clear: it's Product A. To fix Product A in the short run, simply lower
the
forecast. Further, here's another opportunity to use your TQM/Six Sigma
training, specifically the Five Why's. Ask the question: "Why is the
forecast
biased?" Ask "why" five times and you'll almost always get to the root
cause: how people are rewarded, compensated, motivated, and so forth.
Bob Stahl says: "A biased forecast is one that's wrong on purpose."
What he means is that there are almost always factors within the
company that
cause people to make biased forecasts. So, once you find the root
cause, eliminate it. Biased forecasts are rarely caused by problems in the forecasting system; they're a management problem, and that where the
correction must occur.
Product B may be more difficult to fix, because the variability we see
here may be inherent in the demand itself; in some months we get more
orders
than others. It's the way the real world is.
An effective way to deal with some Product Bs - usually the important
ones - is through "demand shaping," motivating customers to order more
linearly, and this can be an excellent approach for important items.
For many others, often the best approach is to not chase the forecast.
Product B's forecast of 20 per month may be as good as it's going to
get (until demand shaping or other major process change occurs). In
cases such
as these which are make-to-stock, we can calculate the variability of
demand - standard deviation (sigma), mean absolute deviation (MAD) -
and base
safety stock on them to insure a given level of customer service.
|
|
And in those cases, you folks on the supply side of the business should
be aware that the shorter the lead time, the less safety stock is
required.
There's a rule of thumb here:
As lead time approaches zero, so does
forecast error.
As forecast error approaches zero, so does the need for safety stock.
Cutting lead
times -
the actual time, not merely the value in your planning system - can
work wonders. This is why companies are reporting such great results
from an
approach called Postponement and for
more
on that, click here.
The question "How accurate should our forecasts be?" implies
dissatisfaction with the forecasts. So my answer is: "What are you
going to do about
it?" I believe you should focus on forecast error, dust off your TQM
tools, eliminate bias, and shorten your lead times. If you do that,
questions
about forecast accuracy will go away. Count on it.
Thanks for listening,
Tom
|
|
Last month we said that effective implementation of Executive S&OP created discomfort. This is
because it's different from past experience. There are, however, several more issues to deal with: a) risk and b) uncertainty.
Most top management teams don't like risk and uncertainly. These must be minimized in order to get top
management to authorize and effectively engage in an implementation effort.
Here's the most effective way to deal with these issues:
- Get a potential Champion from the ranks top management to learn about Executive S&OP - via
workshops, books, or videos.
- The Champion should schedule an Executive Briefing - a facilitated business meeting - to be
led by a person having hands-on experience with successful Executive S&OP, and attended by the top management team, including the leader of the
business.
- From the Briefing gain a commitment to dedicate limited resources to bring up a live pilot
of Executive S&OP - one or two product families - within 90 days or less.
- Following the live pilot, make a go/no-go decision to implement full Executive S&OP: all families
and full financial integration within the next six months.
This approach reduces both risk and uncertainty to near zero. The risk is miniscule: some people's time for a three month period to bring up the pilot, and no one full time on the
project.
Following the live pilot, in which top management participates hands-on, the uncertainty
is zero. The executive group has seen how the process works in their company. They have not only seen it; they've done it. They are
now in a position to make a completely informed decision whether or not Executive S&OP will be helpful to their business.
The low-risk 90-day pilot is one of the key elements in bringing the top management team fully on board with Executive S&OP.
|
|
|
 |
|
| |
|
