Forecast vs Actual: How to Check Your Sales Forecast

Forecast Integrity checks a forecast against two yardsticks: the plan for the same window and the actual result of windows that have run, as a series across periods. “Forecast” here means rEFV, the risk-adjusted revenue expectation of the pipeline. cEFV is a separate metric that this article does not check. A single comparison of forecast and plan carries no verdict.

The forecast is below plan, and nobody knows what that means

Every finance leader knows the comparison of plan and actual. The plan is the target, the actual is the result, and the difference is the variance.

Sales adds a third number, and it arrives earlier. The forecast says what revenue the open pipeline leads you to expect before the period has ended. At that point there is no actual yet.

A typical situation: halfway through the quarter, the forecast is below plan. Two readings are in the room at once. One says the business is short of revenue. The other says the forecast has always been cautious.

Both readings can be right. The meeting cannot decide, because only two numbers are on the table.

Three everyday observations point to the pattern:

• The same variance is argued about anew every quarter.
• Nobody can say from memory how the forecast stood against the result in past quarters.
• The variance gets a name before anyone has done the calculation.

The usual comparison of plan and actual helps only in part. The usual comparison places two numbers of one period side by side. A sales forecast needs three numbers across several periods. Funnel Economics calls this procedure Forecast Integrity.

Why does a single comparison carry no verdict?

Plan and forecast come from different places. The plan is a number that was set. The plan comes from the company’s plan data and carries an origin and a version.

The forecast is a number that was calculated. rEFV is built from the open deals in the pipeline. In the CRM, a deal is called an opportunity.

The calculation starts from bEFV, the structural expected value of the funnel. bEFV multiplies each deal’s amount by its stage win rate. The stage win rate is, of the deals that stood in the stage and were decided, the share that was won.

rEFV additionally weights bEFV by SLA Health, with measured factors per band. SLA Health compares a deal’s total age in the funnel with two time targets per stage and opportunity type (line of business).

rEFV is therefore an expectation: a calculated value, not a result.

When a calculated expectation sits below a number that was set, there is more than one possible explanation. The distance can describe the state of the business. The distance can equally be a property of the valuation the forecast is calculated with.

The comparison itself does not tell these explanations apart.

What is missing is the history. Every period that has run has a result. A forecast that was never held against its result is a construction, not a measurement.

An average does not replace the history either. A mean across all periods hides the direction. The mean also hides how many cases stand behind each single value.

A forecast below plan is, at first, only a difference. Only the actual of the periods that have run shows what that difference means.

How do forecast, plan and actual become visible as a series?

Forecast Integrity places three quantities side by side and carries them across several periods.

Three quantities for the same window

The three quantities are the plan, the forecast and the actual. The forecast is rEFV for one window. The plan is the plan for the same window. The actual is added once the window has run.

A forecast for a quarter belongs next to the plan of that quarter. A value extrapolated to the year next to an annual target would be a different window.

The scope has to match as well. Plan and forecast apply to the same line of business and the same level of the company.

Forecast vs plan, one value per period

The first comparison is the ratio of forecast to plan. The ratio is kept as a series: one value per period, the current value next to the values of the periods that have run.

Each value comes with the case count of its period. The case count says how many decisions carry a value.

The procedure rules out two presentations: the ratio as a single value, and the ratio as an average across all periods.

Forecast vs actual, once the window has run

The second comparison is the ratio of forecast to actual. Once the window of a period has run, its result is known. The actual then stands next to the forecast, as an amount and as a ratio.

Periods whose window is still running or lies in the future stay in the series and are marked as running or future.

The second comparison yields a simple observation. Across the periods that have run, was the forecast above the actual every time, below it every time, or sometimes above and sometimes below?

This observation is a description, not a rule. The observation says how the forecast has stood against the result so far.

If no period has run yet, the forecast cannot be checked against an actual. The report then says exactly that.

What stays in the series

If the plan for a period is missing, the series says so, with the reason. The period is not calculated with zero and is not silently left out. A missing plan is not replaced by the prior year or by an estimate either.

There is no fixed tolerance value around the plan. The procedure knows no threshold beyond which a distance counts as critical.

The rows of a series have to be comparable. If the line of business or the level changes between periods, the series shows the change and not the course.

Taken on its own, the running quarter in the example shows only a distance to plan. The series places next to it how the forecast has stood against the actual so far. Only then can the distance to plan be put into context. A label from a single comparison does not hold.

What helps before a label is applied?

The rules of the procedure govern what may be concluded from the comparison. The series is calculated in every case, even where no conclusion holds in the end.

This article recommends four steps:

1. Keep the series. Forecast and plan stand side by side in the same way every period, with the case count.
2. Add the actual as soon as a window has run.
3. Apply no label from the comparison alone, neither “gap” nor “buffer”.
4. Measure a cause instead of reading it off the direction of the variance.

A label is not evidenced while the actual of the periods that have run is not next to it. An action cannot be derived from the direction of the comparison alone either.

The fourth step includes a counter-check. Anyone naming a cause checks the obvious other explanation on the same rows. A match is not evidence until the next-best explanation has been calculated.

For plan vs actual analysis in finance, this means the analysis starts with the series and not with the difference of a single period.

This article also recommends a clear division of tasks. Sales operations keeps the series and adds the actual. The CFO and the board receive the current value only together with the series.

Every figure in a paper for the leadership body comes with its series and its case count.

In the optional add-on Funnel Economics Strategist, a forecast statement goes out as a range of three named cases, not as a single number. That range is not the subject of this article.

How does the procedure differ from related metrics?

EFV Ratio, the comparison of won revenue with rEFV, is a metric. EFV Ratio divides actual Closed Won revenue by rEFV at the start of the period.

The example above calculates in the reverse direction, forecast to actual; where both use the same rEFV, the two values are reciprocals of each other. EFV Ratio says how well the forecast was calibrated, nothing about quota attainment; it can stand beside a missed plan.

Forecast Integrity is a procedure. The procedure sets the plan beside it and requires the series across several periods. The distance of the forecast to the plan is not quota attainment: quota attainment is the actual against the plan.

cEFV, the best case of the forecast range, is a separate metric for the end of a period. This article checks rEFV, not cEFV.

The article Customer Concentration Risk in the Sales Forecast asks a different question: on how few deals the forecast depends. Forecast Integrity asks how the forecast stands against the plan and against the result. The two views complement each other.