# Sales Forecasting Methods: Two Questions, Two Windows

> Sales forecasting methods start with two questions: what did a stretch that has run produce, and what does today's open pipeline lead you to expect?

Canonical: https://funneleconomics.ai/resources/whitepapers/build-vs-perform

A sales forecast answers two questions in two windows. PERFORM, the look back, asks what a stretch that has run produced. BUILD, the look at the open pipeline, asks what today’s stock leads you to expect. Answering both questions with the same number compares quantities that do not belong together. Among sales forecasting methods, this separation is the first step: every forecast gets the window in which its answer stands. A deviation can then be noticed earlier.

## One number is asked to answer two questions

In many forecast meetings one number is on the table, and it is asked to answer two questions at once. The first question is what a stretch that has run produced. The second question is what today’s stock of open deals leads you to expect. Both questions are legitimate. Each needs its own number from its own window.

The four roles at the table ask the two questions in different proportions. The [CEO looking at the revenue forecast](https://funneleconomics.ai/solutions/ceo) wants to know whether the pipeline supports the target. The [CFO who plans with the forecast](https://funneleconomics.ai/solutions/cfo) wants to know how earlier forecasts stood against the result.

The [CRO who owns the forecast](https://funneleconomics.ai/solutions/cro) has to stand behind both answers. The [VP Sales who steers the work of the team](https://funneleconomics.ai/solutions/vp-sales) needs the answer early enough to still change something.

In the end, a supervisory or advisory board reads the number in a report. A reader there does not know where the number comes from. The report therefore has to say which question the number answers.

Three everyday observations point to the problem:

- The same number appears in the review of the quarter and in the outlook for the next one.
- A forecast from the open pipeline is placed next to a revenue target as if both were the same quantity.
- Nobody in the meeting can say from memory how earlier forecasts stood against the result.

Many overviews of sales forecasting methods compare calculation procedures. This whitepaper starts one step earlier. Before a procedure is chosen, it is settled which question is being asked and in which window its answer stands.

Funnel Economics uses two model classes for this: PERFORM and BUILD. Both classes deliver a number for the same question. Only one of them answers the question.

The risk is inconspicuous. Hitting the wrong window produces no error message. Hitting the wrong window produces a plausible number: a look back where a forecast was asked for.

## Every forecast has two windows

Every period of a forecast model has two windows: the window in which work happens and the window in which deals close. Funnel Economics calls the first the activity window and the second the impact window. How the two windows relate to each other describes the model class, and with it the question a number answers.

### Where work happens and where deals close

The work happens in the activity window: deals come into being, advance to the next stage, the next step of the sales process, or are pushed. The closes stand in the impact window. A number without its impact window is therefore incomplete. Nobody can say when its result will be known.

With PERFORM, in its basic form, both windows have the same length and lie on top of each other. A PERFORM period is a stretch that has run: the pipeline that stood at its start, and what closed within the same stretch.

With BUILD, the windows lie one after the other. A BUILD period is one stretch of build-up and its harvest in the periods that follow. The activity window is the running stretch, and the impact window comes after it.

The position of the windows determines which stock the calculation uses. PERFORM calculates with the stock at the start of the period, because that stock is the starting point of the stretch. BUILD calculates with the stock at the end of the period, because that stock is handed forward into the future.

### The length of the window does not decide the question

The length of a window says nothing about the model class. “The next four quarters” names an impact window. A PERFORM model and a BUILD model can carry the same length of impact window.

The sales cycle of a line of business, the average time a won deal takes from creation to close, determines the length of the window, not its class. A long sales cycle calls for a long impact window. Whether the question looks back or forward is not settled by that.

Which windows an organisation keeps is its own configuration. Every organisation cuts its windows differently: a different number, different names, different durations. There is therefore no fixed time distance between the two windows.

A shorter impact window is not a calculation error. A shorter window answers a different question, and the most common one in sales: how the running period stands and how the next one is developing.

A window that diverges from the sales cycle does change the meaning of every number on it. In a window that is too short, closes that come later are missing. That divergence belongs next to the number, not in a footnote.

### The direction of the question decides

The selection rule comes before every calculation. If somebody asks what will close in future, that is a BUILD question. If somebody asks what a stretch produced, that is a PERFORM question.

An example of a BUILD question: what is the funnel worth, from today’s point of view, for the coming quarters? An example of a PERFORM question: what did the quarters that have run produce?

If the question does not name its direction, the direction is asked for and not guessed.

:::merksatz
The length of a window says nothing about the question. What decides is the direction: back at a stretch that has run, or forward from today’s stock.
:::

## What does PERFORM, the look back, measure?

PERFORM measures what a stretch that has run produced. At the start of the stretch there is a stock of open deals. At its end it is known what was won out of that stock and within that stretch. PERFORM places the two side by side: the expectation at the start and the result at the end.

With PERFORM, the close is the measure. Won and lost are facts once the stretch has run.

The expectation at the start is a calculated value. [bEFV, the structural expected value of the funnel](https://funneleconomics.ai/resources/glossary/befv), values the amount of each open opportunity with a historical rate, the stage win rate. [rEFV, the risk-adjusted revenue expectation](https://funneleconomics.ai/resources/glossary/refv), additionally weights bEFV by SLA Health with measured factors per band.

The metric of the look back is EFV Ratio. [EFV Ratio, the comparison of won revenue with rEFV](https://funneleconomics.ai/resources/glossary/efv-ratio), compares actual Closed Won revenue, what was won in the period, with rEFV at the start of the period.

EFV Ratio says how well the forecast was calibrated for the pipeline whose impact window has run. It says nothing about target attainment; it can stand beside a missed plan.

A single value of EFV Ratio describes one stretch. The metric becomes readable as a series across several stretches that have run, with the case count next to it.

The look back also supplies the yardstick for the look forward. The stage win rates come from windows that have run and are applied to today’s stock.

[Decided-row win rate, the share of decided deals that were won](https://funneleconomics.ai/resources/glossary/win-rate), is a metric of its own with a population of its own. The historical rates are the stage win rate, not the decided-row win rate.

What PERFORM cannot do follows from the definition. A PERFORM window is a stretch that has run. A PERFORM window cannot carry a close date in a future quarter.

Putting the question about the future to a PERFORM window therefore returns an empty answer or a retrospective one. Both look like a finding about the business. Both are a finding about the question.

## What does BUILD, the look at the open pipeline, measure?

BUILD measures what today’s stock leads you to expect for future windows. What is counted is what stands in the funnel at the end of the running stretch. That stock is handed forward into the future. A BUILD window carries no result within itself. The closes lie in the impact window that follows.

### BUILD is a progress model, not a snapshot

A BUILD model is **not a snapshot**. A BUILD model is a progress model. It measures activity, newly created deals and stage progress during the activity window.

A leadership task stands behind this. The hardest task in dealing with a sales organisation is getting it to do something in this quarter for a future period. That is exactly the work a BUILD model measures.

What becomes visible in a BUILD window is the movement of the stock. New deals arrive. Deals advance to a later stage. Deals are pushed to a later date.

A pushed deal is moved, not decided. A stretch can look good because a lot is advancing, and at the same time push a lot back. The two movements belong side by side.

Progress in the activity window is a steering quantity. Progress shows what was done. This whitepaper derives no statement about the later close from it. Forecast and actual stand side by side, with no claimed cause between them.

### The term for the quality of the stock

The method’s term for the look at the open pipeline is EFV Density. [EFV Density, the measure of pipeline quality](https://funneleconomics.ai/resources/glossary/efv-density), sets rEFV against total pipeline volume, both at the end of the period. No tool reports EFV Density.

Funnel Economics calls this look the quality dimension. The method assigns EFV Density to a BUILD model, because that is where today’s stock stands.

What is read is pipeline volume and rEFV as amounts, as a development across several periods.

### Why the two dimensions are read together

The look back through EFV Ratio and the quality of today’s stock are two dimensions, and one can hide the other. An organisation can win more than expected in the look back while the quality of today’s stock is falling. The short-term success then covers a development in the stock.

Keeping only the number of the look back does not show today’s stock. Keeping pipeline volume and rEFV of the stock next to it shows both dimensions side by side. Whether falling quality later arrives in the result is shown only by the actual of the window once it has run.

:::merksatz
A BUILD window is not a snapshot. BUILD measures what was done during build-up for a later stretch: new deals, stage progress, pushed dates.
:::

## Every BUILD forecast has a window in which its answer stands

BUILD and PERFORM are not a separate pair of worlds. Both classes see the same deals twice. BUILD looks forward at a stock. PERFORM looks at the stretch in which exactly that stock closed. Every BUILD period therefore has a PERFORM period in which its answer stands.

The link is simple to describe. The stock a BUILD window shows at its end is the stock with which the corresponding PERFORM window begins.

An obligation follows from this. Once the impact window of a BUILD period has run, its actual is known. A series of forecasts then places the actual next to it, as an amount and as a ratio of forecast to actual.

A forecast that was never held against its result is a construction, not a measurement.

Not every window of a series has run. The longer the impact window, the fewer periods already have an actual. The remaining periods are marked, as running or as future. They are not left out.

A window that is still running is no counter-check. Only pairs that have run count for the comparison of forecast and actual.

:::example
Illustrative example with invented numbers. A company keeps a series of four consecutive windows. Each window shows the BUILD forecast in euros and, where the window has run, the actual of the corresponding PERFORM window.

- Window 1 (has run): BUILD forecast €1,840,000, actual €2,430,000.
- Window 2 (has run): BUILD forecast €2,260,000, actual €2,020,000.
- Window 3 (running): BUILD forecast €2,110,000. There is no actual yet.
- Window 4 (future): BUILD forecast €2,480,000. There is no actual yet.

In one of the windows that have run, the forecast was below the actual; in the other, above it. The numbers are invented and are not guide values. No rule can be derived from two windows that have run.
:::

The example shows the form of the series, not a result. For windows three and four there is only a forecast. Anyone reading those forecasts sees next to them how the forecasts of the windows that have run stood against the actual.

### A forecast on the stock is not a revenue target

When a BUILD forecast is held against a revenue target, two different quantities stand side by side. The BUILD forecast is calculated on the stock at the end of the build-up stretch. A revenue target means the revenue of a period, including deals that only come into being and close within that period.

As long as the forecast has never been held against the actual, a distance to target cannot be interpreted. The distance can describe the state of the business. The distance can equally be a property of the valuation.

How forecast, plan and actual are kept as a series is described in a separate article: [Forecast Integrity, forecast vs actual in sales](https://funneleconomics.ai/resources/patterns/forecast-integrity). This whitepaper starts before that point and does not carry out the comparison against plan.

## Which question gets which form of answer?

The position of the window determines the form of the answer. A window not yet begun gets a range of three named cases. The running quarter gets three standings. A window that has run gets the forecast next to the actual. A single number without its origin is the answer in none of the three positions.

| Question about | Form of answer |
|---|---|
| a window not yet begun | range of three named cases |
| the running quarter | three standings, each with its kind |
| a window that has run | forecast against actual |

These rules belong to the optional add-on Funnel Economics Strategist. Strategist is designed to give a forecast answer for the executive level in the form that fits the position of the window. Funnel Economics Core calculates bEFV and a simplified rEFV by health status in your Salesforce org. rEFV with measured factors per health band, EFV Ratio and cEFV are calculated by Strategist.

### A window not yet begun

For a window not yet begun, the answer is a range of three named cases: Worst Case, Realistic Case and Best Case. The executive level needs a range it can defend, not a single number.

Worst and Best are named business assumptions, not statistical bounds. Both cases have to be defensible without talking about mathematics.

The three cases are not sorted by size. The label hangs on the assumption of a case, not on its height. An order of the three amounts is therefore not guaranteed.

[cEFV, the calibrated revenue forecast for the end of a period](https://funneleconomics.ai/resources/glossary/cefv), is a metric of its own that Strategist calculates. cEFV is the best case of the forecast range, not the realistic one.

### The running quarter

The running quarter is neither a stretch that has run nor a window not yet begun. The range does not apply to the running quarter. The three cases describe a whole window from its first day. In the middle of the quarter, part of the window is already fact.

The running quarter therefore gets three standings, each with its kind:

- **Won so far.** This standing is measured.
- Expected. This standing is an upper bound, not a point.
- The bound of today’s stock. This standing counts everything open today as won, unweighted.

The three standings and their kinds come from the methodology for the running quarter.

The bound of today’s stock is not an absolute ceiling. Deals from the following quarter can be pulled forward, and new deals are still coming into being.

Every number from a running window comes with a caveat. A running window has no end yet. Until then, the stock keeps moving.

### A window that has run

For a window that has run, the answer is the comparison: the forecast next to the actual. That is the series from the previous chapter.

If the series shows a deviation, the next step is not to rebuild the forecasting procedure. The next step is to break down the cause of the deviation and place a measure next to it.

A cause is measured and not read off the direction of the deviation. A label from a single comparison does not hold.

## What does this mean for leadership?

For leadership, the separation means that every forecast number gets its question, its window and its form of answer. The benefit is noticing a deviation earlier. The benefit is not a sharper number. A forecast describes the past and carries it forward. What matters is whether somebody notices in time that the direction is changing.

### Keeping both windows in view

The two windows call for a balance. Concentrating only on the closes of the running period means the future impact windows are not built up. The consequence is a hole in the pipeline that becomes visible only later.

The same holds the other way round. Building up only and neglecting the closes hits the revenue of the running period. In both cases leadership trades one window for the other.

For the VP Sales this is a question of steering. The work for a future window competes with today’s close. A BUILD window makes that work visible before its result is known.

### Four steps before every forecast number

This whitepaper recommends four steps, in this order:

1. **Name the question.** Does the question look back at a stretch that has run, or forward from today’s stock?
2. Name the window, with its start and end.
3. Choose the form of answer: range, three standings, or forecast against actual.
4. Add the actual as soon as a window has run.

The steps do not change how a forecast is calculated. The steps change what stands next to the number.

### What the paper for the leadership body contains

The paper for the leadership body contains no term from the methodology. It contains the origin of the number: what the number was checked against. Whoever defends the number does not have to be able to explain the calculation. Whoever defends the number has to know what it was checked against.

The same applies to the report for a supervisory or advisory board. The terms BUILD and PERFORM belong in the preparation, not in the report. The report says whether a number looks back or forward and which window it applies to.

Every number also comes with its series and its case count. The series places the periods that have run next to the current value. The case count says how many decisions carry the number.

A forecast additionally comes with the actual of the windows that have run. Where one of these details is missing, the number cannot be checked.

Funnel Economics is entering its beta phase. During the beta, you start by running it against a Salesforce sandbox.

Discover is meant for trying Funnel Economics before any installation. It is built on a demo org.

:::merksatz
The benefit of the separation is not a sharper number. The benefit is noticing a deviation earlier, because every forecast gets its window and its actual.
:::

## Methodology and data basis

This whitepaper describes the quantities of the two model classes: definition, basic formula, interpretation and delimitation. The statements about the method come from the Funnel Economics method documentation. The numbers in the example are invented. This whitepaper contains no measurements from any organisation and no external market figures.

### The quantities of the method

| Quantity | Definition | Basic formula | Window |
|---|---|---|---|
| bEFV | structural expected value of the funnel | amount of each open opportunity × stage win rate, summed | start for PERFORM, end for BUILD |
| rEFV | risk-adjusted revenue expectation of the open pipeline | bEFV, weighted by SLA Health with measured factors per band | start for PERFORM, end for BUILD |
| EFV Ratio | comparison of result and expectation | Closed Won ÷ rEFV at the start of the period | PERFORM |
| EFV Density | quality of the open pipeline | rEFV at the end of the period ÷ pipeline volume at the end of the period | BUILD |

The definitions and basic formulas come from the Funnel Economics method documentation. EFV Density is a term of the method; no tool reports EFV Density.

SLA Health classifies each open opportunity by its total age in the funnel. The yardstick is two SLA values the customer sets per stage and opportunity type.

cEFV is the calibrated revenue forecast for the end of a period. cEFV is a metric of its own; it appears here only as the best case of the forecast range, not as the realistic one.

### How the metrics are read

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

This whitepaper names no target value for EFV Ratio or EFV Density. EFV Ratio is read as a series across several periods. This whitepaper applies no label from a single value.

The stage win rates come from windows that have run and are applied to today’s stock.

### Delimitation

The name of the metric BPV (Build-to-Perform Variance) suggests a bridge between the two model classes. BPV measures something else. [BPV, the deviation of time in stage from the SLA](https://funneleconomics.ai/resources/glossary/bpv), compares the actual time opportunities spend in a stage with the SLA of that stage.

The forecast vs actual comparison with plan, forecast and actual is the subject of the article on Forecast Integrity. This whitepaper clarifies which question is being asked, in which window its answer stands and which form the answer takes.

### What the example shows and what it does not

The example in the chapter on the series uses invented amounts. The amounts show how forecast and actual stand side by side and how running and future windows are marked. How close a forecast comes to the actual cannot be derived from the example.

:::merksatz
Two questions, two windows: PERFORM shows what a stretch that has run produced, BUILD what today’s stock leads you to expect. Every number belongs in its window.
:::

## FAQ

### What is the difference between pipeline and forecast?

The pipeline is the stock of open deals with their amounts. A forecast is a statement about the revenue to expect in a specific window. The sum of open deals is therefore not a revenue forecast. Funnel Economics values the stock with the stage win rates from windows that have run and calls the result an expected value. Every forecast comes with the window in which its answer stands.

### What do BUILD and PERFORM mean?

BUILD and PERFORM are the two model classes Funnel Economics uses to form time windows. PERFORM, the look back, asks what a stretch that has run produced: the stock at its start and what closed within the same stretch. BUILD, the look at the open pipeline, asks what will close in future out of today's stock. The length of a window says nothing about the class. What decides is the direction of the question.

### Why can a pipeline forecast not be compared directly with the revenue target?

Because the two numbers measure different quantities. A forecast from the open pipeline is calculated on the stock that is open at the end of the build-up window. A revenue target means the revenue of a period, including deals that only come into being in that period. The comparison holds only once the actual of the windows that have run stands next to the forecast. Funnel Economics describes this forecast vs actual comparison as Forecast Integrity.

### Does separating BUILD and PERFORM make the forecast more accurate?

No, that is not the claim. The separation does not change how a forecast is calculated. The separation assigns each forecast the window in which its answer stands and places the actual next to it once the window has run. The benefit is noticing a deviation earlier. This procedure does not promise a hit rate.

### How do you check a forecast once the window has run?

Once the window has run, the result is known. The actual then stands next to the forecast, as an amount and as a ratio. Windows that are still running or lie in the future stay in the series and are marked. The full forecast vs actual comparison, with plan, forecast and actual as a series across several periods, is described in the article on Forecast Integrity.