Cash flow forecasting is a major problem and opportunity for corporate treasurers to be able to understand and get control of their cash. However, it is horrendously difficult as companies' business models vary so much, see John Holmes. There are articles on cash flow forecasting on magazines and web-site all the time. These range from the those describing the basic disciplines and traditional forecasting systems, and the dynamic forecasters who are suggesting a new approach is needed.

Improving traditional forecasting

Michael Coveney, from STW Consulting, in an important article in gtnews wrote: "Cash flow forecasts should always be a range of values based on the subject of reason and backed up with some form of empirical evidence"..... And that, "with this in mind, forecasts can be greatly improved by: 

• involving more people who are relevant in the process.
• challenging assumptions on what drives performance.
• collecting the detail behind the numbers (workload, resources and outcomes) to give management a more comprehensive story.
• producing trends to see if submitted forecasts are logical.
• allowing managers to determine the variability of individual forecasts."

He then went on to take each point in turn and considered the role played by technology.

A new paradigm: Dynamic Forecasting

In 2009 Bjarte Bogsnes, Vice President, Performance Management Development, Statoil (a pioneering strategy management professional and a dedicated member of the Beyond Budgeting movement) wrote a mind changing paper on Dynamic Forecasting. He believes that the calendar is an arbitrary (and anachronistic) basis for forecasting. 

Bosgnes wrote, "The finance function may have its roots in statutory accounting, but performance management is about so much more. It is about enabling and supporting company performance by helping people and teams do their best. Traditional accounting has little to do with this aim. For the same reason that we at Statoil have largely decoupled performance management from our statutory accounting—by abolishing traditional budgeting altogether— we believe calendar-based planning must also go. In today’s accelerated (and volatile) business environment, it has become inflexible and increasingly irrelevant. The time is ripe for dynamic forecasting."

Dynamic forecasting represents an attractive alternative to calendar-based planning (including rolling forecasting) that is characterized by a fixed time horizon and frequency. Bogsnes believes that mixing target setting, resource allocation, and forecasting is guaranteed to destroy the quality of your forecast because target setting and resource allocation create systemic bias in forecasts. Just ask a sales manager about her best sales forecast, and then mention that this forecast will be next year’s target. Or ask a manager about his best cost or CapEx forecast, quickly adding that this number will also become next year’s budget. Those forecasts will all start to move, corrupted by the introduction of secondary  purposes.

By using "three separate processes to accomplish these very different purposes—allowing them to yield three separate numbers, not just “one” budget number— has enabled us to significantly improve the quality of each process."

Further improvement

Many Beyond Budgeting companies use rolling five-quarter forecasting. Rolling forecasting is much better than short-term forecasting, where the horizon keeps shrinking because the forecast stops at year-end. As a fixed frequency (typically every quarter) and fixed time horizon (perpetually five quarters ahead) method, fivequarter rolling forecasting still has too many calendar limitations for Statoil. Also different parts of the business have very different planning and business horizons.

So Statoil introduced Dynamic (Event based) Forecasting, in which:

• no fixed time frequency and no fixed horizons
• forecasts are updated when important new information arises
• forecast time horizons are relevant to the event/business.

This approach raises all sorts of questions, e.g.:

• What is an event? And who gets to decide? (Bosgnes believes that when to update a forecast and how far it should extend cannot be a corporate decision; it must be local.)
• How far ahead should a business unit look? (Bosgnes: That depends on the local entity’s needs—and on the visibility.)
• How often should we review our forecasts? (Bosgnes: Forecast input data will flow continuously into our systems, whereas output or reporting will take place on a regular schedule (as it does today), or more frequently, if necessary.)
• Will dynamic forecasting increase the workload? Does it mean more forecasting? (Bosgnes: Not necessarily, and in some cases it actually calls for less. It means forecasting at the right time, with the right time horizon.)
• What about data quality? Are all the numbers consistent and reconciled? (Bosgnes: Not necessarily in an accounting sense, but this, after all, is not accounting. Forecasting quality, as measured this way, might generally be somewhat lower than what we might reach in the much more intensive fall budgeting and planning period. But we believe it is much better to have something roughly right all the time than something that is precise only once a year.)

This approach to cash flow forecasting really is a new paradigm.

The ultimate: forecasting the dynamics of it all

MIT (Massachusetts Institute of Technology) in the 1960s began developed System Dynamic modelling techniques. This work is now very advanced and is being used to simulate dynamic behaviour in our economy, markets, our firms and our lives. The Dynamics Forecasting Group in the US are using this technology to understand and predict how our social systems are likely to behave over time -- the dynamics of it all.

---

CTMfile take: Many companies probably could not cope with Statoil's approach to cash flow forecasting, as they still have the traditional budgeting systems, but the central tenet that 'parts of the business have very different planning and business horizons' is spot on. The Dynamic Forecasting's work shows that many business situations can be modelled cost effectively for predicting market demand.