5 Quick Wins to Improve Forecasting Techniques

 

In this section, FD4Cast teams up with Steve Morlidge, author of 'Future Ready: How to Master Business Forecasting', in order to produce a brief 'Quick Win' guide for better forecasting techniques. For the PDF version please click here.

As any FD knows, forecasting could never be considered to be an exact science. The more complex a forecast is to build, the odds are that its usefulness may decrease. Similarly, forecasting should never be assumed to be just ‘plain common sense’.

1. Understand why you are doing it

The purpose of forecasting is not to predict the future. The purpose is to anticipate what might happen, given a reasonable set of assumptions about the world. Armed with this information, you can then take decisions, either to help bring that future about, or to change it to one that is more desirable.

Some Questions:

• What decisions will be made with the aid of your forecast? (this usually involves starting, stopping, rescheduling, changing or creating some form of activity)
• What information do you need to make these decisions and is it provided as part of the forecast process? (usually this involves understanding the incremental impact of any changes you might want to make to the portfolio of activities. Often this information is not readily accessible to decision makers)
• Can you eliminate information that is not relevant for decision making purposes? (if you can it will streamline the forecast process and help improve forecast accuracy)

2. Managing the Time Dimension

Forecasts are necessary only because we cannot react fast enough to events as they happen. That is why super tankers need radar but speedboats do not.

Some Questions:

• How long does it take to enact a decision? (the decision making lead-time will determine how long your forecast horizon should be)
• How frequently do elements of your forecast change in a way that impacts decision making? (this determines the frequency at which you need to refresh your forecast and is likely to involve updating some forecast elements more frequently than others)

3. Producing Forecasts

There is no mathematical ‘silver bullet’ that will allow you to produce a perfect forecast; forecasts are assumption-based, always involving the use of judgement in some way, but this carries a high risk of bias.

Some Questions:

• Which forecast variables tend to follow a clear historical trend? (These may best be forecast using statistical extrapolation techniques. Simple techniques usually give more accurate results than complex ones)
• If forecast variables do not follow a simple historical pattern do you have a good grasp of drivers of performance? (if so, it may be appropriate to use a driver based model to produce your forecast)
• Which variables are best forecast using judgment? (always make sure that your assumptions are clearly stated and communicated)

4. Check and Adjust

The only way to get good at anything is to learn from your mistakes. Why is it that in business, forecast performance is often not measured at all, and when it is, it is usually done badly? Why, all too often, are messengers shot?

Some Questions:

• Do you have a clear, well communicated forecast policy? (most businesses do not, so is not surprising that forecast quality is poor. As a minimum, this should state that forecasts should be unbiased - with no systematic error – and subject to acceptable levels of variation)
• Is forecast error routinely measured and appropriately followed up? (use error run charts to spot and eliminate bias in real time)
• Are forecasts ever overridden or adjusted? (The best advice is not to do it. If it is unavoidable, maintain records to ensure that any adjustments made consistently reduce bias)

5. Acknowledge and Manage Risk

Since perfect prediction is impossible it is important to understand how far you might be wrong, and what implications this has for decision making.

Some Questions:

• Does your organisation only produce single point forecasts? (if so, start producing range forecasts.
• Does your organisation produce multiple, competing forecasts? (if so, stop now. Use differences of opinion to build your range forecast)
• How quickly can you respond to unforecast changes? (be clear how you would react to the range of potential outcomes. Build contingency plans, particularly if a risk threatens the survival of the organisation)

Takeaway Tips & Quick Wins

If you begin by doing only the four things listed below, it’s still a great start to improving your forecasting:-

• START measuring forecast accuracy using run charts – look for 4 or more consecutive errors of the same sign / bias;
• START documenting your assumptions;
• STOP unilaterally adjusting forecasts;
• STOP praising people for coming back to / beating forecasts.

Content written by Steve Morlidge of Satori Partners, guide produced in association with FD4Cast.

Click here to find out more about FUTURE READY: 'How to Master Business Forecasting'

      - Fritz Roemer. The Hackett Group