Forecast accuracy is often a subject of discussion. What are the issues met that can generate such level of discussion? 

Evidently, the first question is whether a forecast needs to be accurate (or not) or more exactly what level of inaccuracy can be tolerated? In other words, what is in this context the definition of an accurate forecast. Let's take a few somewhat divergent views that I met in my carrier.

• The forecast is accurate if the actuals tend to match the forecast within “small” variances. It would mean that we define small but let that issue aside for the moment.
• The forecast is accurate when all the assumptions and decisions are sound, correctly captured, sized and timed to. This definition accepts that actuals are different due to assumptions or decisions that failed, are not implemented or achieved or with wrong timing.  
• The forecast is accurate when it effectively captures all the “negotiated” objectives of the divisions, departments and their management. This tends to “personalise” the variances.
• The forecast is accurate when its structure constitutes a sound suite of assumptions and decisions to achieve a given strategy. 

Rather than trying to decide what is right and wrong, more precisely what creates or is the result of positive and negative behaviours inherent to human nature, let us look at the “values” underlining each definition. 

• The forecast is accurate if the actuals tend to match the forecast within “small” variances. Obviously, this gives “easy” time to management by giving visibility and predictability.

It minimizes surprises and the needs for corrective actions (which does not mean that it does not include changes and challenges). 

• The forecast is accurate when all the assumptions and decisions are sound, correctly captured, sized and timed to. A “technical” view of the forecast that insists on confidence and consistency on the forecast. 
• The forecast is accurate when it effectively captured all the “negotiated” objectives of the divisions, departments and their management. That view insists on ownership throughout the organization.
• The forecast is accurate when its structure constitutes a sound suite of assumptions and decisions to achieve a given strategy. That one insists the overall vision and goals of the company. 

If we accept all of those are valid then the forecast needs to be a tool that helps the company to achieve its overall goals through ownership thanks to a sound technical process that give visibility and predictability based on a set of assumptions and decisions.

This is probably not achieved in one but rather by a set of tools covering the short, the medium and the long-term (accepting that term has a different meaning depending on the industry or business type). The overall goals would be the domain of the plan, the ownership would be through budget (budget and reforecast), the visibility and predictability would be help through “forecasts” all those been supported by a coherent and consistent process. Each tool would have its own level of precision and as such its own set of variances.

Such should eliminate or limit the number of “inaccuracies” not to say errors, but not all. It shall permit a clear analysis of the variances and lead to decision making. Variances and variance analysis then become a true asset (i.e. not the forecast accuracy) and set the basis for decisions.

For more details you may want to refer to my previous articles:

• Watch Your FP&A Processes
• My FP&A Suite
• Evaluate Your FP&A Processes
• FP&A Analytics and Simpson's Paradox

Still, all this is based on the perception of the environment and its evolution. These “externalities” that will impact the accuracy of the whole process which depends on the quality of the business and competitive intelligence of the company. “Business is a democracy”, customers, competitors and few others have their say and a lot of the assumptions are based on those (in particular for the revenues and cost of sales which generally drive the rest).

At any moment or over a period of time, one or several of those externalities can generate variances or deviances that endanger more or less your forecasts. This is not inaccuracy as such, this is a lack of business and competitive intelligence. The issue is not in the forecast process it is in the intelligence gathering domain and shall be looked at and treated as such i.e. no point in looking at your forecast process, you must look at your intelligence gathering process. Obviously, if the intelligence was there but not integrated into the forecast that brings us back to the internal process. 

There may even not be a clear solution at that level. This brings us to the flexibility and agility necessary in the forecast process. It is not directly an accuracy issue (except if the accuracy issue is linked to lack of agility). 

In summary, forecast accuracy is not the important item and not worth the debate (as long as you have a sound process). Variance analysis is the key. 

It shall permit to identify precisely whether those variances are linked to:

• Understanding / interpretation of the environment and the competitive situation,
  • Prices evolution, currency impact, customers financial situation, economic trend changes, anything that impacts volume and price or profitability, etc…  
• Externalities not perceived at the time of the forecast,
  • New entrant, competitor product or technologic breakthrough or evolution, etc…
• Management ownership issues,
  • Productivity, program success or failures, things links to management decisions or lack of decisions, etc… 
• “Internalities” I.e. other not management issues. 

It shall drive to corrective or adaptive actions i.e. … one new forecast.