Blog
Why your B2B forecast is wrong
Ask a B2B owner mid-month how much will close by month end, and you get a number. Ask again on the last working day, and half of it has moved to next month. Same deals, same customers, new date.
The usual explanation is that the sales team is padding. In long-cycle B2B, project work, installation, factory equipment, the kind where a deal can pass through four hands before it closes, that’s almost never what’s happening. The reps are answering honestly. The question has no honest answer, because nobody ever defined what the words in the pipeline mean.
Why B2B sales forecasts are wrong
Most advice on sales forecasting is written for a different business: retail, stock levels, restaurant covers. Hundreds of transactions a month, where last year genuinely predicts this year. Fit a curve to history and it works.
A pipeline of 20–40 open deals, each different, each taking two to nine months and decided by a committee, is not that problem. There is no curve. The forecast is a pile of human judgments, and they break in three places:
- “Stage” means different things to different people. One rep moves a deal to “proposal” when the quotation is emailed. Another waits until the customer has been walked through it. Same label, two realities, added together.
- The probability attached to each stage is decoration. It was a default nobody ever checked against your own close rate.
- Nothing forces a dead deal out. A deal that stopped moving five months ago still sits there, counted as revenue.
Percentage probabilities set by feel are just guessing with a spreadsheet
Nearly every pipeline carries stage probabilities: 20%, 40%, 60%, 80%. Multiply value by probability, sum the column, and you have a weighted forecast. It looks like arithmetic, so it gets trusted like arithmetic.
Two things are wrong with it. First, the percentage rarely comes from your data. Count back: of the last 30 deals that reached “proposal sent,” how many closed? If the answer is nine, that stage is 30%, not the 60% on the screen. Every deal in it is overstated by a factor of two.
Second, even a correct percentage only behaves at volume. With 40 similar-sized deals a weighted average means something. With six, where one is four times the size of the rest, it describes an outcome that cannot occur. You win the big one or you don’t.
The percentage isn’t really the disease. The disease is that stage placement is a feeling.
How to write stage-exit criteria nobody can argue with (with real examples)
Here’s the test a criterion has to pass: an admin who has never met the customer can tick the box from what’s in the record, without asking the salesperson. If answering needs a conversation, it’s an opinion, not a criterion.
Weak versus written:
- “Customer is interested” → “We’ve spoken with someone who can approve, and they told us the budget exists and which quarter it sits in.”
- “Proposal sent” → “Proposal delivered and presented live; the meeting date is recorded.”
- “Negotiating” → “The customer has returned a revised scope, or asked for a specific commercial change, in writing.”
- “Verbal yes” → this is where installation-business forecasts go to die. A verbal yes is not a commitment. Criterion: “A purchase order, signed quotation, or written approval is attached.”
A four-stage set for an installation or equipment business:
- Qualified. Need, site, timeframe and the name of the decision-maker are recorded.
- Scoped. Site survey completed, requirements written down, survey date on record.
- Proposed. Priced proposal delivered and presented live.
- Committed. PO, signed quotation, or written approval attached.
Every criterion is an artifact or a dated event, not a mood. That’s the whole trick, and it costs nothing to adopt.
Set a max age per stage in advance, and force a decision when it’s exceeded
The other half of forecast error is deals that stopped moving but never left. A deal sitting in “proposed” for five months is not a 60% deal. It’s an unanswered question displayed as revenue.
For each stage, write how many days a healthy deal actually takes. Use your own history, not a benchmark. Then any deal past that number gets exactly one of three outcomes at the weekly review:
- Re-dated with a reason. The customer gave a new date and it’s recorded.
- Moved back a stage. A condition you thought you had isn’t there.
- Closed as no-decision. Allowed, normal, not a failure.
Teams resist the third, because stale deals keep the pipeline total looking healthy. That comfort is exactly what makes the month-end number wrong.
Starting point: take your median time from first contact to close, and treat any deal sitting in one stage longer than a third of it as overdue.
Three numbers to look at before answering how much will close this month
Stop summing weighted values. Look at three:
- Committed value. Only deals meeting the top stage-exit criterion. An artifact count, not a forecast. Smaller than you’d like, and the only number that’s real.
- Deals with a customer-confirmed date this month. The customer stated the date and it’s on record. Not the salesperson’s target.
- Share of open pipeline that’s overdue. Deals past max stage age, divided by all open deals. Above roughly 30%, your pipeline is a storage unit, not a forecast.
Now the answer becomes: committed is this, these deals have a date the customer confirmed, and this share is stale so I’m not counting it. Something a business can act on. Produced without asking a single rep.
Agree on the criteria first, then argue about the numbers
Most forecast meetings argue about the number. Wrong order: everyone is arguing over a figure derived from definitions nobody agreed on.
Reverse it. Write the exit criteria, agree what counts as overdue, and the number stops being an opinion and becomes an output. That’s a week of honest work with the people who actually sell. Not a software project, and not a forecasting module you have to buy. Software is only useful once the process it holds has been written down, and the criteria have to be written by the team, or they die with whoever wrote them.
If your pipeline is telling you a number you don’t believe, book a 45-minute call. We’ll walk your existing stages, and you’ll leave with the exit criteria written for at least one of them. They’re yours whether or not you go further with us. That’s the shape of a BUILD 30 install from B2B Sales System, a ZestMate Solution program: we design the pattern, your team installs it, and day 30 is measured against three criteria written on day one.