What is revenue forecasting?

Revenue forecasting is predicting the revenue your business will realise in a coming period, based on what is already secured, what is in the pipeline and how things went in the past. A good forecast is not a single number but a range, built from layers with different levels of certainty: contracted revenue, recurring revenue, open deals and new revenue that is not yet known. The goal is not to be exactly right, but to know early enough where you will land that you can still do something about it.

Why do you need a revenue forecast?

The managing director of a B2B company makes decisions every month that depend on future revenue. Do we hire someone? Can we make this investment? Do we need to adjust costs? How much working capital will we need in the third quarter? How much stock do we buy? What do we tell the bank?

Without a forecast, you make those decisions on instinct or on last year's revenue. With a bad forecast, you make them on a number that looks more certain than it is, and that is worse. A forecast that comes in ten percent too high every quarter leads to hires, investments and spending that should never have happened.

A forecast therefore serves three purposes:

1. Planning. Matching capacity, staffing, purchasing and investment to expected revenue.
2. Steering. Seeing early that you are behind, so sales, marketing or pricing can still intervene.
3. Accountability. Giving shareholders, the bank or the supervisory board a well-founded expectation.

Each purpose needs a different level of accuracy and a different horizon. A twelve-month forecast for workforce planning does not need to be right to the euro. A forecast for next month's working capital needs to be roughly right.

How does revenue forecasting work?

A forecast is not a prediction out of thin air. It is built up in layers, from certain to uncertain. The more certain a layer, the less you have to estimate.

Layer 1: contracted revenue

Revenue already secured in signed contracts, running subscriptions and framework agreements with committed volumes. A maintenance contract of EUR 3,000 a month with eighteen months left to run is EUR 36,000 of near-certain revenue for the next twelve months. Near-certain, because customers can go bankrupt, terminate within the terms or start a dispute.

The source here is the contract register, not the CRM. And this is the first trap: if contracts are not properly recorded, or invoicing does not match the contracts, even this most certain layer is unreliable. An indexation that is never invoiced is revenue the forecast expects and the accounts never see.

Layer 2: recurring and predictable revenue

Revenue without a fixed contract that does follow a pattern. A customer who has ordered every month for five years will probably do so next month too. For wholesalers and manufacturers this is often most of their revenue. The source is history: what did this customer order in recent months and years, and at what rhythm? The article forecasting from historical revenue explains how to turn that into a reliable baseline, including seasonality and customers who quietly start buying less.

Layer 3: pipeline

Deals in the CRM that have not yet been won. Each deal has an expected value, an expected close date and a stage. The classic way to turn this into a forecast is the weighted pipeline: deal value multiplied by a probability per stage. A EUR 100,000 deal in the stage "quote sent" with a 40 percent probability counts for EUR 40,000.

This is the layer where most forecasts go wrong. Stage probabilities are usually chosen, not measured. Close dates slip. Deal values are optimistic. Deals that died long ago are still there. More on this in forecasting from CRM data and in why pipeline is not revenue.

Layer 4: revenue not yet known

Revenue from deals that are not in the pipeline today but will be won and invoiced within the forecast period. For a forecast of next month, this layer is usually small. For a forecast of next year, it can be large. The source is history: how much revenue in the past came from deals that did not yet exist at the start of the period?

Adding it up, with uncertainty

The forecast is the sum of the four layers, but each layer carries its own uncertainty. Contracted revenue has a narrow range, pipeline a wide one. An honest forecast shows this: not "we will do EUR 9.2 million", but "we will probably do between EUR 8.8 and 9.5 million, and the difference sits mainly in five large deals". How to set and communicate that range is covered in forecast confidence explained.

Which forecasting methods are there?

There are several ways to build a forecast. Most companies combine a few of them.

No single method is good enough on its own. A strong forecast uses contract analysis for layer 1, historical trend for layer 2, a calibrated pipeline weighting for layer 3 and a historical average for layer 4. Sales judgement is then not the basis but a check: where does the salesperson's view differ from the model, and why?

The difference between the traditional approach and models that learn for themselves is covered in AI forecasting vs traditional forecasting. How such a model works and what you can expect from it is in AI revenue forecasting explained.

Why do revenue forecasts go wrong?

Forecasts rarely go wrong because of a wrong formula. They go wrong because of the data that goes in and the behaviour of the people entering that data. The article why are sales forecasts so often wrong covers this in depth. The main causes:

Optimism in the pipeline. A salesperson who marks a deal as "lost" admits they will not win it. Nobody likes doing that, so deals stay put. The pipeline grows, the forecast grows with it, revenue does not.

Close dates that slip. A deal that was going to close "at the end of the quarter" moves to the next quarter, and then again. Each time it sits in the current quarter's forecast. A deal whose close date has moved three times has a different probability from one that has never moved, but most forecasts treat them the same.

Stage probabilities that were never measured. Someone once set "quote sent" to mean a 50 percent chance. Nobody has checked whether that is right. If in reality a quarter of quotes are won, the forecast overstates this stage by a factor of two.

CRM data that does not match reality. Deal values in the CRM differ from what is eventually signed and invoiced. The article CRM forecast vs actual revenue shows where that gap comes from and how to measure it.

Revenue leakage in the base. The forecast assumes what ought to be invoiced. If indexations are not applied, extra work is not invoiced and deals are not turned into orders, actual revenue sits structurally below the forecast, even if the pipeline is predicted perfectly.

Incomplete data. Fields are empty, deals are entered after the fact, some salespeople barely use the CRM. How to still build a usable forecast from that is covered in forecasting with incomplete CRM data.

How do you assess the pipeline?

Because layer 3 holds the most uncertainty, assessing the pipeline is a discipline of its own. Three concepts help.

Pipeline coverage is the ratio between the value of your pipeline and the revenue target for a period. If you need to close EUR 1 million of new revenue and your pipeline is EUR 3 million, your coverage is 3. How much coverage you need depends on your actual win rate. A fixed rule of thumb says little. See pipeline coverage explained.

Pipeline velocity is the speed at which value moves through the pipeline: number of opportunities times average deal value times win rate, divided by the average sales cycle. It tells you how much revenue your pipeline produces per day or per month, and which of the four levers you can pull. See pipeline velocity explained.

Pipeline hygiene is the question of whether what sits in the pipeline is still alive. Deals without activity, deals with slipped close dates, deals far beyond the usual cycle length. The article how do you spot an unreliable pipeline lists the signs.

A pipeline with high coverage and low velocity is a warning: there is a lot in it, but it is not moving. A pipeline with low coverage and high velocity can be perfectly fine, as long as new inflow keeps up.

How do you build a reliable forecast?

A reliable forecast is not a matter of a better model, but of a better process. The article how do you build a reliable revenue forecast sets out the full step-by-step plan. The essentials:

1. Split the forecast into layers. Contracted, recurring, pipeline and new. Forecast each layer with the method that suits it.
2. Make the contract layer reliable. Record contracts with term, amount and indexation, and check that invoicing matches. This is the foundation everything else rests on.
3. Calibrate the pipeline on your own history. Measure per stage how many deals were won in the past, how long it took and how far the final value differed from the value entered. Use those figures instead of default percentages.
4. Clean up the pipeline. Deals without activity for longer than your normal cycle are removed or given a lower probability.
5. Give a range. A floor (only contracted and very likely revenue), an expectation and a ceiling.
6. Measure your forecast afterwards. Record each month what you predicted and compare it with what happened. Without that measurement you do not know whether your forecast is improving.
7. Discuss deviations, not totals. The forecast meeting is about what has changed since last time and why, not about the total.

Scenarios instead of a single number

For decisions over the next six to eighteen months, one forecast is often not enough. What happens if the two largest deals do not close? If a large customer cancels? If the price increase from 1 January is or is not accepted? Scenario forecasting builds a separate forecast for each of those questions, so you know in advance what you will do if it happens. See scenario forecasting for B2B.

Scenarios are most useful when a small number of events determines a large part of the outcome. In a company with thousands of small orders, uncertainty averages out. In a company where five deals make up a third of new revenue, it does not.

Forecasting by type of business

The right approach depends heavily on how your revenue comes about.

Companies with recurring revenue, such as software companies, MSPs and maintenance businesses, have a large contract layer. Their forecast is driven mainly by churn, expansion and indexation. The pipeline drives growth, not the base.

Wholesalers and manufacturers have a lot of recurring revenue without contracts. History and seasonal patterns are the most important source here. The biggest risks are customers who quietly start buying less and price changes that affect volume.

Project businesses, such as construction, installation and engineering, have little recurring revenue and large, irregular contracts. Their forecast has to account for the difference between the value of a won contract and how revenue spreads over the duration of the project, for variation work and for delays. The article revenue forecasting for project businesses works this out.

SMEs often have no dedicated forecaster, no clean CRM and no data team. That need not be a problem: a simple forecast based on contracts, history and the ten largest deals is often more reliable than an elaborate model on poor data. See revenue forecasting for SMEs.

Worked example

Worked example: suppose an IT service provider with EUR 10 million in annual revenue builds a forecast for the next quarter.

The total range here is added up as floor plus floor and ceiling plus ceiling. That is cautious: in reality, not every layer disappoints at the same time, so the true range is narrower. For a managing director, the signal is clear all the same: well over two thirds of the expectation is almost certain, and the uncertainty sits mainly in the pipeline.

Now suppose a check shows that for 20 of the 140 contracts this year's indexation is not being invoiced. Then the contract layer is not EUR 1,650,000 but lower, and the forecast is already too high in its most certain layer. This is why forecasting and revenue leakage belong together: a forecast is only as reliable as the invoicing it rests on.

What a forecast is not

A target. A target is what you want to achieve. A forecast is what you expect to achieve. If the forecast is adjusted to hit the target, it is no longer a forecast but a wish. Keep them separate, in presentations too.

The sum of the pipeline. A EUR 5 million pipeline is not an expectation of EUR 5 million in revenue, not even weighted, as long as the weighting is not based on your own history.

A one-off document. A forecast made once a quarter is out of date after three weeks. Update it whenever something changes: a large deal won or lost, a customer cancelling, a contract being renewed.

A responsibility of sales alone. Sales knows the pipeline, finance knows the contracts and the invoicing, operations knows capacity. A forecast made by only one department misses two thirds of the information.

Which systems do you need?

For a first good forecast, you do not need a special system. Excel with an export from the CRM and the accounting system is enough to build the four layers. The problem with Excel is not computing power but maintenance: exporting again every month, matching again, explaining again why the number differs from last month.

If you want to go further, there are three routes. Forecasting features in the CRM itself, which usually only see the pipeline. A BI environment such as Power BI on a data warehouse, which can see everything but mainly shows what someone has built. Or a revenue intelligence platform that reads CRM, contracts and invoicing together and builds the forecast from them. RiOS has such a forecasting module, with a revenue forecast and a likely range, alongside your key figures compared with similar companies. It is described on the page about the system.

Whichever route you choose, the order is the same: first make the contract layer and invoicing reliable, then calibrate the pipeline, and only then automate.

Frequently asked questions

What is the difference between revenue forecasting and sales forecasting?

Sales forecasting predicts what sales will close, usually based on the pipeline. Revenue forecasting predicts total revenue, including running contracts, recurring customers and how won deals spread over time. A sales forecast is one component of a revenue forecast.

How accurate does a forecast need to be?

Accurate enough for the decision you base on it. A forecast that lands within a few percent of reality every quarter is good for most B2B companies. More important than accuracy at any one moment is that the deviation does not keep falling on the same side. A forecast that is structurally too high has a systematic error you can find.

How far ahead should I forecast?

For working capital and capacity, usually one to three months. For staffing and investment, six to eighteen months. The further ahead, the wider the range and the larger the share of layer 4, revenue not yet known.

Do I need AI for a good forecast?

Not to start with. The biggest improvements usually come from a reliable contract layer, a cleaned-up pipeline and win rates based on your own history. AI helps when you have enough history and want to weigh many signals at once, such as activity, cycle length and payment behaviour.

Who owns the forecast?

Ideally the person responsible for finance, with input from sales on the pipeline and from operations on capacity. A single owner prevents three forecasts from circulating, each with a different number.