Sales forecasting: how to stop guessing your revenue

Sales forecasting in B2B companies usually looks like this: a salesperson enters an amount and a date into the CRM, the sales manager adds up the column, and at the end of the quarter it turns out that only 60 percent of that sum actually came in. Nobody is surprised, everyone is tired. The question is: what exactly failed, the people, the data, or the way we count?

Why sales forecasts fall apart every quarter

In most cases the team is not the culprit. The culprit is that the forecast is based on declarations rather than behaviors. At one company, early in the quarter a salesperson entered three opportunities worth 200k each into the CRM, each at the "negotiation" stage. At the end of the quarter one closed at 180k, and two were pushed to the next quarter. The salesperson said "I'm close" because they believed in the relationship. The client said "looks good" because they didn't want to close the door. Neither piece of information was data.

The second reason is the failure to distinguish between a pipeline stage and a probability. The "negotiation" stage means exchanging contract terms for one salesperson, and a third introductory meeting for another. If stages don't have defined exit criteria, the forecast is a sum of opinions, not numbers.

The third reason is less obvious: the forecast is built in isolation from marketing. Sales plans revenue, marketing plans campaigns, and nobody checks whether the number of leads in the pipeline can even deliver the target result. This is a classic planning mistake. Tomczak, Reinecke and Kuss in Strategic Marketing (2018) write that the starting point for marketing planning should be a quantitative identification of customer potential, based on both internal sources (e.g. sales representatives) and external ones. In other words, first count the potential, then promise the revenue.

How to forecast B2B sales: the data you already have

Before you reach for advanced models, check whether you're using what's already sitting in your systems. B2B companies usually have four categories of data available, and they're enough to build a first sensible forecast.

Historical CRM data

You need three things: the number of opportunities entering the pipeline each month, the conversion rate between stages, and the average time spent in a stage. If your CRM doesn't measure this, that's the first sign the forecast will be guesswork. It's also worth checking how many opportunities drop out of the pipeline without a decision, which is often the most underrated number in the whole calculation.

Data on customers and market potential

A sales forecast can't be based solely on what's already in the pipeline. You need to know how many companies in your segment even qualify to buy, and what share of them is within reach of your activities. This is exactly the quantitative identification of customer potential that Tomczak et al. (2018) write about. Without this step, the forecast is internally consistent but disconnected from the market.

Sales activity data

The number of calls, meetings, proposals sent, demos. This is the data that is easiest to falsify, but also the easiest to verify. If you know that with 40 qualified conversations per month the team closes an average of 6 deals, you have a benchmark. Any deviation from that relationship is a signal, not a coincidence.

Purchase cycle data

In B2B the cycle ranges from a few weeks to more than a dozen months. This means that sales in this quarter are largely the result of work from the previous two or three quarters. A forecast that does not account for this will always be late.

Sales forecasting methods: from the simplest to the most demanding

There is no single right method. There is a method matched to the amount of data and the maturity of the process. Below are four approaches ranked from least to most demanding.

If a company has a short data history, do not wait to forecast until you have collected two years of statistics. Start with conversions calculated at the level of the whole team over the last 3 to 6 months and use them as weights for the funnel stages. Instead of a single number, build a range, a pessimistic and an optimistic scenario, and treat them as a tool for discussing risks, not as a precise result. At the same time, make sure the CRM collects data on stages, conversions, and time in the funnel, because only after a few quarters will you have material for more demanding methods.

Method based on a weighted pipeline

You assign each funnel stage a weight equal to the historical conversion from that stage to close. If 30 percent of opportunities at the "proposal" stage close, the weight is 0.3. The forecast is the sum of the products of opportunity values and weights. Simple, fast, and it works if the stages have clear exit criteria. Without them, the weights are fiction.

Method based on conversion rates

You calculate the forecast from the bottom up: how many qualified leads will come in during a given period, times the lead-to-opportunity conversion, times the opportunity-to-win conversion, times the average deal value. This method is resistant to salespeople's optimism, because it relies on team-wide averages, not on individual declarations.

Method based on the sales cycle

You take the opportunities that entered the funnel in previous periods and shift them in time by the average cycle. This method is particularly useful in companies with a long cycle, where quarterly sales are largely already determined by work from months earlier.

Diffusion model and customer lifecycle analysis

In more mature organizations, especially with new products or when entering a new segment, it is worth reaching for diffusion models. The standard Bass model (1969) was created for consumer goods, however, and in B2B it requires adaptation, for example instead of consumer social influence you need to use other proxies, such as recommendations within the client's organization, staffing changes at the decision-maker, or competitor activity in the same segment. The model describes the pace of adoption of a new product through two parameters: the internal propensity to buy and social influence. In B2B practice this means that the pace of customer growth is not linear: first it rises, then it slows. A linear forecast in such a situation will always be wrong, usually in both directions, depending on the phase.

Three forecast levels: pessimistic, base, optimistic

A single number in a forecast is an invitation to argue. Three numbers are an invitation to decide. In practice, a model works well when the base forecast relies on average conversion rates, the pessimistic forecast assumes conversion rates 20 to 30 percent lower, and the optimistic one assumes rates 20 to 30 percent higher. These are indicative ranges worth calibrating against your own historical data. Companies with a short cycle and stable conversion tend to see a narrower spread, while those with large contracts and long cycles usually see a wider one. The ranges should be communicated openly to both leadership and the team.

What matters is that each level has its own set of assumptions attached. The pessimistic scenario is not a "worse mood", it is a scenario in which, for example, the two largest contracts slip and conversion from proposals drops by a quarter. That way, when the forecast does not materialize, you know which assumption failed.

How to connect the forecast to marketing planning and scaling

A sales forecast with no link to the marketing plan is a document that ends up in a drawer. To have operational value, it must answer the question: how many leads, and of what quality, must marketing deliver for the base forecast to materialize.

This works both ways. Marketing that knows it needs to deliver 120 qualified leads in the second quarter can plan its budget and channels in advance. Sales that knows there are too few opportunities in the pipeline to hit the forecast can raise the issue early instead of explaining it after the fact.

In companies going through a phase of intense growth, the sales forecast becomes the basis for decisions about scaling the business, whether to hire more salespeople or improve conversion first. Without a forecast, such decisions are made on gut feeling, and that is an expensive method.

It is also worth remembering that the forecast does not operate in isolation from the rest of the marketing funnel. Before you start forecasting revenue, check whether your input data is reliable at all. A marketing audit shows where the measurement gaps are in the funnel and which lead sources actually close in sales versus which only generate traffic.

What not to do so you don't break your sales forecast

The most common mistake is forecasting at the level of an individual salesperson and summing the results. People have different tendencies toward optimism, different definitions of stages, and different ways of entering data. A forecast should be calculated at the team level and verified against historical data, not against declarations.

The second mistake is not revising. A forecast made once at the start of the quarter and left untouched until the end is useless. It should be updated every two weeks, with a clear note of what changed and why.

The third mistake is ignoring seasonality and external factors. In B2B, seasonality can be strong, especially in companies selling to the public sector or to industries with a budget cycle. A forecast that doesn't account for this will always be caught off guard in January and June.

And the fourth, most costly one: treating the forecast as a tool for evaluating people. If a salesperson knows that an overly optimistic forecast will be held against them, they will start lowballing. If they know a lowballed one will too, they will stop entering anything meaningful. It's worth treating the forecast as a decision-making tool, not as a performance review of salespeople.

A good sales forecast doesn't require advanced tools. It requires three things: defined funnel stages, historical conversions, and regular revision. The rest is a matter of discipline, not technology.