The board deck said the quarter would land. It did not. Someone points to two deals that pushed, someone else says the pipeline was never real, don't forget about that delayed marketing campaign. And... the room agrees to watch it more closely next month. Three months later you are having the same conversation with different deal names. Nobody in the room is lying. Nobody can explain the gap either.
Sales pipeline and revenue forecasting is arithmetic on facts you already own: stage conversion, deal size, cycle length and win rate applied to the deals in front of you. That is the whole mechanism. You do not need a data science function to check it, and you do not need to be a mathematician. You need the last four quarters of closed business and an afternoon.
What follows is how to run that audit. The four numbers that matter, where the data lives, the questions that separate a forecast from a story, and what to do with the answers.
Four metrics carry the entire forecast. Stage-to-stage conversion: the percentage of deals that advance from each stage to the next. Average deal size, measured on closed-won only. Sales cycle length, measured from qualified opportunity to signature. Win rate, measured against qualified opportunities rather than against every lead marketing ever touched.
They multiply. Qualified Opportunities X Win Rate X Average Deal Size = Expected Revenue. Cycle length tells you which quarter it lands in. If any one of the four is an estimate someone typed in a spreadsheet, the forecast is an estimate too.
This is also where coverage ratios get abused. A team with a 25 percent win rate needs roughly 4x pipeline coverage. Claiming 3x coverage is healthy when your history says 4x is how a forecast quietly becomes a wish. Ranges we commonly see in mid-market B2B run 3x to 5x coverage and 90 to 180 day cycles, and those are ranges, not benchmarks. Your own numbers overrule them.
Most forecast arguments are really data arguments in disguise. Before anyone defends a number, agree on where the facts live.
Four sources cover almost everything you need. CRM stage history tells you when each deal entered and left every stage, which is how you get real conversion and real cycle length. Closed-won and closed-lost records over the last 12 to 24 months give you win rate and deal size with enough volume to trust. Calendar and email activity show whether deals are actually moving or just sitting. Invoicing and order data confirm what closed and at what value, which is the only version finance will accept.
A forecast built on rep opinion is not a forecast. It is a confidence poll. Ask a seller what closes this quarter and you get their optimism, their quota pressure and their memory, blended.
A clean trail is simple. Every deal has a dated stage history, a written close reason and a value that matches the invoice.
Sort last year's closed-won revenue by rep, largest to smallest. In most mid-market teams we work with, the top 20 to 30 percent of sellers produce somewhere between half and three quarters of the revenue. That is a range, and yours will land inside it. Run the same sort on stage-to-stage conversion and cycle length. The pattern usually holds.
Then look at what the top group does differently. It is rarely charisma. It is usually a repeatable motion: they qualify harder and earlier, they multi-thread into two or three contacts, they lose faster, and they work fewer deals with more discipline, and they are persistent as all get out.
If the motion lives in three people's heads, the process was never written down. Document it, build it into your stage definitions, coach to it and recruit with it. Replacing the bottom half restarts the ramp clock and leaves the same gap behind.
In your next revenue pipeline review, ask five questions and listen for hesitation.
Where is the written definition of each stage, and who signed off on it? If nobody can produce the document, every conversion rate in the forecast is an opinion. Which deals have been in stage longer than your average sales cycle? Those are not late, they are lost and nobody has said so yet. Does the quarter close without the single largest deal? If the answer is no, you have one deal and a story around it. What percentage of marketing-sourced pipeline reached closed-won last year, and what happened to the rest? A steady flow of leads that never converts is usually a handoff problem between marketing and sales not a lead quality problem.
Last question: what changed in the forecast since last week, and why? A leader who can answer that in specifics is managing the plan. A leader who cannot is reporting on how they are missing it.
A forecast is only as accountable as the plans behind it. Every sales leader should have a written plan, one or two pages, that you could read in five minutes. It names the accounts and segments they are responsible for. It shows the coverage math: pipeline required to hit the number at their own conversion rate, not a blanket three-times rule. It states ramp assumptions for any new hire, including how long until first close. And it lists the weekly leading indicators they will manage, such as first meetings booked, proposals out, and deals moved past qualification.
When that plan exists, the monthly review changes character. You stop discussing the number and start discussing variance against a stated assumption. Coverage short by 20 percent has a cause you can name. A rep behind ramp has a coaching answer, not an excuse.
That is what it means to own the number.
AI earns its place in three spots. It scores patterns across your closed-won and closed-lost history faster than any analyst, surfacing the deal attributes that actually correlate with a win. It flags deals aging past your average cycle before a rep volunteers the bad news. And it does CRM hygiene work nobody wants: missing contacts, stale next steps, activity that never got logged.
What it cannot do is invent facts. If your stage definitions live in three people's heads, a model trained on that data will produce confident nonsense. If half your closed-lost records have no reason code, no algorithm recovers it. Fix the definitions and the data trail first, then let AI work on clean inputs. We cover the mechanics of forecasting with AI in more depth separately.
And it does not replace the deal review. A model can tell you a deal looks stalled. Only your sales leader knows the buyer went quiet because their CFO changed.
Most of this work starts with Fractional CRO leadership inside your business, sitting in the pipeline reviews and reading the same data your team reads. The first few weeks are a diagnostic on the actual records: 12 to 24 months of closed-won and closed-lost, stage history by rep, cycle length by segment, and the gap between what the CRM says and what the forecast assumes. From there we rebuild stage definitions in writing, set conversion baselines everyone agrees to, and rebuild the coverage math on those baselines instead of a habit. Each sales leader ends up with a written plan you can hold them to, with named accounts, ramp assumptions, and weekly leading indicators. Marketing gets tied to the same definitions so sourced pipeline means one thing. The target is a forecast you trust and measurable results in 90 days, run by your team, not by us.
Export every opportunity that closed in the last 90 days, won and lost. Include the rep, the stage history, the create date, the close date, and the amount. One spreadsheet, no dashboard required.
Then do three calculations. Win rate by rep on that quarter's closings. Stage-to-stage conversion by rep, counting how many deals entered each stage and how many advanced. Average days from first stage to close, by rep and by segment.
Now put those three numbers next to the assumptions inside your current forecast. If the forecast assumes a 30 percent win rate and the quarter says 18, you have found the gap and you did not need a consultant to find it. If the numbers vary widely by rep, you have found where the process is undefined rather than where the talent is missing.
Bring that single page to your next pipeline review and ask your sales leader to reconcile the difference.