![[Banner Image]](images/hhome1.gif)
![[Dividing Line Image]](images/div1.gif){kind=small}
[Tip of the Month]
[Books]
[Events]
[Contact Information]
Like many engineers I love Dilbert cartoons. They brilliantly cut to the heart of an issue and their logic seems impervious to challenge. However, sometimes even the great Dilbert gets it wrong. Perhaps he unconsciously uses the logic of software in domains where it does not apply. In this case, he appears to assume that small forecasts are just a subclass of the larger class of forecasts and this subclass inherits the properties of the general class of forecasts. Thus, if a forecast is worthless, then a small forecast must also be worthless. He also tacitly implies that doing more forecasts takes proportionally more effort; for, if there was no additional effort involved, he probably wouldn't complain.
Let's examine these assumptions. What happens when we break the forecasting process into a series of small batches, instead of doing a single, larger, long-range forecast? Anyone who has done forecasting knows that forecasting errors rise exponentially with the distance to the forecast horizon. Detailed forecasts at short time horizons are much more accurate, which is what a rolling forecast exploits. Furthermore, every time we reforecast we gain the benefit of new information that has emerged since the previous forecast. Let's take a simple example. Draw the five cards of a poker hand in sequence and try to forecast the likelihood of winning before drawing each card. Does your accuracy improve as you get more information? Of course it does. Decomposing a single detailed long-range forecast into a series of short-range forecasts can easily turn a worthless forecast into one that is worthwhile.
Does it really take proportionally more effort to do more forecasts? You need to understand the concept of a rolling forecast. As we increase the frequency of our forecast we proportionally reduce the amount of detail in each individual forecast. For example, if we forecast once a year we might provide detail for 12 months. If we forecast monthly we only need precise detail for one month and a high level view of the remaining 11 months. We provide less detail and it is exponentially easier to provide this detail at a one month time horizon. The further ahead you peer in the fog, the harder it is to see, and the longer it takes to process what you see. Doing a sequence of short-range forecasts takes less effort than forecasting the same period in one large batch.
In essence, rolling forecasts are an application of batch size reduction, a
key method of lean product development.. By doing forecasts more frequently and in
small batches, we gain many benefits. We simultaneously improve the efficiency,
quality, and response time of our forecasting process. In reality, this exploits
the very ideas that underlie the shift from the large batch waterfall model to
today's agile software methods. While I do not expect Dilbert's pointy-haired boss
to understand this, I do expect more from my hero Dilbert -- after all, he is an
engineer. Smart developers no longer use the large batch size logic of the
waterfall process, and Dilbert shouldn't either.
Thrifty shoppers can find FLOW for $28.76 at: Barnes & Noble
Prepublication Reviews: Book Reviews
Overseas readers can order the new book from: Celeritas Publishing Order Page
UK Readers can order the new book from Dave Donelan (+44 (0)1926 336423) at Smallpeice Enterprises
For bulk orders contact: Celeritas Publishing: Special Sales
[Return to Top of Page]
Public events are shown below. Please contact us directly
for information about in-company courses. We have done such courses
in the United States and 19 other countries.