How to reduce detail in your forecasts

Rolling Forecasts are quite popular today. But to implement them properly it is usually imperative to reduce the detail in the forecasting models. Less detail speeds up the process and helps to increase the accuracy.  A recent post on this blog looked at some of the problems with too much detail. The big question though is to where and how to cut detail. While people tend to look at the chart of accounts first, many organizations actually have great success with making a few modifications to their timescale.

THE BIG SCALE

Take a look at the photo below. It symbolizes one of the key issues with forecasting: the further out we look the more diffuse our view gets. While we might have a good idea of what is going to happen next month, it is usually more difficult to do the same for the months after. That’s just the way it is.

Rolling Forecasts - The time horizonUnfortunately, most forecasting templates do not reflect this fact of life. Take a look at the original time-scale from a customer that I used to work with. The organization wanted to look beyond fiscal year end. However, all months were treated equally:

A traditional time-scale (208 data points)

Notice how much detail is being generated. And detail requires effort. As a business person, I will have to sit down and try to provide an amazing amount of detail. This could take a while. The basic assumption of this template is that business people are able to precisely quantify when something is going to happen no matter if it’s tomorrow or next year. That is dangerous and it’s simply not possible. Here is an example: I might know that a certain customer will purchase my product next month. But I will most likely not be able to precisely identify the same thing for next year. The forecast will therefore most likely be wrong from a timing perspective. Why the detail then?

A DIFFERENT TIMESCALE

How about changing the timescale? Take a look at the final redesign in IBM Cognos TM1:

Rolling Forecast Model
Less detail. Probably more accurate (112 data points)

The new version reduces the detail by almost 50%. And this approach pays tribute to the fact that the further out we look the more diffuse our view of the future becomes. Overall, we could argue that this template will produce more accurate forecasts while also making it easier for the business. This is a lot easier to work with! My client implemented a similar timescale with excellent results.

YOUR MODELS

Take a look at your current models. Is there an opportunity to alter the timescale? How much detail could you get rid of? If you want to embark on implementing a Rolling Forecast, you should most definitely look at this approach. Please let me know your thoughts and experiences.