Not legal advice: The simulations below use stated assumptions (independent deals, a fixed win probability), not your data. They show how noisy a small pipeline is by nature; your own win rates will differ.
Most CRMs give you a weighted forecast: multiply each open deal by the close probability of its stage, add everything up, and read off a number. It looks like a prediction. It is an average, and the gap between those two things is the reason small businesses so often feel a forecast "missed" when nothing went wrong.
We worked through three questions. How often can a weighted number be right when you only have a handful of deals open (we computed it)? What do the few published surveys say about how much sales teams trust their forecasts? And what does a forecast need to look like to be useful to a business with ten or twenty deals, not two thousand? The short answer to the first: with ten equal deals at a 30% chance each, the weighted forecast of three wins lands within 20% of the real outcome only about 27% of the time.
Key takeaways
- A weighted forecast is the expected value of your pipeline, not a forecast of one outcome. Each deal is closed or lost, so the real result is always a spread around the number.
- We computed that spread. At a 30% win probability per deal, the chance that actual wins land within 20% of the weighted forecast is 0% with 5 deals, about 27% with 10, 54% with 20, 72% with 50, 85% with 100 and 95% with 200. Small pipelines are noisy by arithmetic, not by poor selling.
- Concentration matters more than count. On a $100,000 pipeline with every deal at 30%, one $50,000 deal plus ten $5,000 deals lands within 20% of the weighted $30,000 only about 10% of the time; twenty $5,000 deals manage about 53%.
- Overstated stage probabilities are the easiest error to avoid. If you assume 50% and the true rate is 30%, a 20-deal pipeline reaches its forecast about 5% of the time.
- In Gong's 2022 survey of 928 sales and revenue professionals, 52% of leaders said their organization regularly misses its forecast by more than 10%. Gartner's 2020 release put high confidence in forecast accuracy at 45%.
- The fix is a calibrated range, not a sharper point: take stage probabilities from your own closed deals, show a low, likely and high case, and treat your largest deals as separate line items.
What does a weighted sales forecast actually tell you?
The method is simple. Give each pipeline stage a close probability, multiply every open deal's value by its stage's probability, and sum. Ten equal $10,000 deals, each with a 30% chance, give a weighted forecast of $30,000.
But no deal closes 30%. Each one is won or lost. The weighted number is the average you would get if you could replay the same quarter thousands of times; in any single quarter the pipeline delivers one of many possible totals. Statisticians call the first thing an expected value and the second a realised outcome, and a forecast built only from the first will look wrong most of the time even when the underlying probabilities are right.
This is the same trap as a weather report. "30% chance of rain" is a perfectly good forecast on a day it stays dry. The problem only starts when the 30% is read as "it will drizzle a bit."
How often does a weighted forecast land on its number?
We modelled the simplest honest case: every open deal is the same size, wins independently, and has a 30% chance. The number of wins then follows a binomial distribution, which we computed exactly. We then counted how often the actual number of wins falls within 20% of the expected number.
The answer rises slowly with pipeline size. With 5 deals the expected result is 1.5 wins, and no whole number of wins falls inside the 20% band, so the forecast cannot be hit at all. With 10 deals it is about 27%, with 20 about 54%, and only around 100 deals does it pass 80%. The relative spread tells the same story: the standard deviation is about 68% of the expected wins at 5 deals, 48% at 10, 34% at 20 and 11% at 200.
The assumptions flatter reality. Real deals differ in size and in quality, and they are not independent (a bad month hits several at once), so real pipelines are noisier than this model. Treat these figures as a best case.
Why does one big deal wreck the forecast?
Count is half the problem; concentration is the other half. We compared three pipelines with the same $100,000 total and the same 30% chance per deal, so each has an identical weighted forecast of $30,000, and ran 200,000 simulated quarters of each.
The twenty-deal pipeline gives a likely outcome near the forecast. The four-deal one puts a quarter of its outcomes at zero (the chance that all four are lost is 0.7 to the fourth power, 24%). The pipeline with one $50,000 deal and ten $5,000 deals is the worst: its realistic range runs from $5,000 to $70,000, and the forecast is within 20% of the result only about one quarter in ten, because the big deal is either in or out.
| Pipeline ($100,000 at 30% each) | Weighted forecast | Range covering 80% of outcomes | Lands within 20% of forecast |
|---|---|---|---|
| 20 deals of $5,000 | $30,000 | $15,000 to $45,000 | 53% |
| 4 deals of $25,000 | $30,000 | $0 to $50,000 | 41% |
| 1 deal of $50,000 + 10 of $5,000 | $30,000 | $5,000 to $70,000 | 10% |
Ranges are the 10th to 90th percentile of 200,000 simulated quarters per row. Your own spread will differ, but the ranking will not: the more of the total sits in a single deal, the less the weighted number tells you.
What do surveys say about forecast confidence?
Two published numbers frame how common the problem is. Gartner's State of Sales Operations survey, reported in a February 2020 press release, found that 45% of sales leaders and sellers have high confidence in their organization's forecasting accuracy, and that 47% believe their organization has high-quality data, while 13% rate their data quality as poor. The release does not state the sample size.
Gong's "Reality of Forecasting" report (a July 2022 survey of 928 sales and revenue professionals) is more detailed. 52% of leaders said their organization regularly misses its forecast by more than 10%, only 24% trust the forecast commitment from their reps (58% trust their own), and just 27% of sales professionals agree that their forecasting process delivers accurate results. Gong sells revenue-intelligence software, so read its recommendations with that in mind, but the survey figures are its own data.
52%
of sales leaders said their organization regularly misses its forecast by more than 10%, in a survey of 928 sales and revenue professionals.
Source: Gong, "Reality of Forecasting" report, July 2022 (vendor survey)
Both are surveys of confidence and self-reported accuracy, not audits of actual forecasts, and both come from vendors or vendor-adjacent research. They tell you the problem is common in teams far larger than yours. They do not tell you which method is best.
How do you calibrate stage probabilities from your own deals?
The weak point of most weighted forecasts is the probability column. CRMs ship with round defaults such as 10%, 25%, 50%, 75%, and teams rarely revisit them. A better method needs nothing more than your own history: for each stage, count the closed deals that ever reached it and what share of those were eventually won.
Use closed deals only. Open deals have not resolved yet, and counting them as neither won nor lost biases every stage's rate. The table below shows the mechanics with a hypothetical set of 60 closed deals; the numbers are invented for illustration.
- Recompute every quarter, because win rates move with season, offer and team.
- Split by pipeline or deal type once each has enough history. A pipeline for a recurring service and one for a one-off project rarely share a win rate.
- With fewer than roughly 20 closed deals per stage the percentage is itself noisy. Start with round numbers you can defend, and let real history take over as it accumulates.
| Stage reached (hypothetical) | Closed deals that reached it | Of those, won | Calibrated probability |
|---|---|---|---|
| New lead | 60 | 15 | 25% |
| Qualified | 48 | 15 | 31% |
| Quote sent | 33 | 15 | 45% |
| Negotiation | 21 | 15 | 71% |
What does the cost of a wrong probability look like?
Optimism is cheap to type and expensive to plan around. Suppose a 20-deal pipeline is booked at 50% per deal, so the forecast is 10 wins, but the true rate is 30%, which means 6 wins are expected. The forecast overstates the result by two thirds, and the chance of actually reaching 10 wins is about 5%.
The reverse error, sandbagging, is less visible but also costly: a team that books everything at 10% will cheerfully beat its forecast and under-hire, under-stock or under-spend on the leads that were about to close.
4.8%
the chance a 20-deal pipeline reaches its forecast of 10 wins when deals were booked at 50% but the true win rate is 30%.
Source: Our calculation: exact binomial probability P(wins ≥ 10), n = 20, p = 0.30
How should a small business present a forecast?
- 1Show three numbers, not one: a low case (what you would collect if only your most advanced deals close), the weighted likely case, and a high case. The low and high can come from stage rules, such as only deals in Negotiation for the low case.
- 2List deals above a size threshold, for example more than 15% of the pipeline, as their own lines. Their outcome decides the quarter, and averaging them into a probability hides that.
- 3Track the forecast against the result: snapshot the weighted number at the start of each month and compare it with what actually closed. A pattern of misses in one direction means the probabilities are miscalibrated; random misses mean you are seeing the noise described above.
- 4Keep the stage definitions strict. A deal moves to "Quote sent" when a quote has gone out, not when it feels warm. Probabilities only mean something if stages mean the same thing every time.
- 5Update stages as deals move, not at the end of the month. A forecast built on three-week-old stages mixes your best information with your oldest.
In oToK, every deal sits in a pipeline stage, and the deals board shows a weighted forecast by each stage's close probability, months and quarters ahead, with sales goals and live progress next to it. Deals and pipelines are unlimited on every plan, so you can keep a separate pipeline, and calibrated probabilities, for each type of sale.
Frequently asked questions
- What is a weighted sales forecast?
- It is the sum of each open deal's value multiplied by the close probability of the stage it is in. Ten $10,000 deals at 30% give a weighted forecast of $30,000. It is an expected value across many possible outcomes, not a prediction of one result.
- Why is my weighted forecast always wrong?
- Because each real deal is won or lost, so a single quarter lands somewhere in a spread around the average. In our model at a 30% win rate, a 10-deal pipeline lands within 20% of its weighted forecast only about 27% of the time, even when the probabilities are exactly right.
- How many deals do I need for a reliable forecast?
- In our best-case model (independent, equal-size deals at 30%), actual wins land within 20% of the forecast about 54% of the time with 20 deals, 72% with 50 and 85% with 100. Below ten deals, treat the number as a rough guide and show a range instead.
- How do I set the probability for each stage?
- Use your own closed deals. For each stage, divide the closed deals that reached it and were eventually won by all closed deals that reached it. Exclude open deals, recompute quarterly, and use round estimates until you have roughly 20 closed deals per stage.
- How accurate are sales forecasts in practice?
- Surveys suggest confidence is low. Gartner reported in 2020 that 45% of sales leaders and sellers have high confidence in their forecasting accuracy, and in Gong's 2022 survey of 928 professionals, 52% of leaders said their organization regularly misses its forecast by more than 10%. Both are survey figures, not audits.
- Should I forecast by deal count or by deal value?
- Both. Value drives revenue, but count tells you how noisy the number is. A pipeline with a few large deals can have the same weighted total as one with many small deals and be far less predictable, so list large deals separately.
