How to Run a Pipeline Review That Reduces Forecast Variance to Under 5%

Every month this meeting produced a number. Every quarter the number was wrong by somewhere between 30 and 60 per cent.

I have sat in a lot of these rooms. Different companies, different industries, the same meeting.

A CRO. Four regional heads. Eleven deals on a screen, colour coded, a percentage beside each one. Ninety minutes in the diary. The commit figure at the bottom of the slide had already been read out to the board.

They reach deal four. The biggest in the quarter.

The CRO asks a mild question. Not a clever one. "When did we last speak to them?"

The rep checks his notes. "We had a really good call with them last month."

"Last month when?"

A pause. "The ninth."

It was the fourteenth. Five weeks of silence on the largest committed deal of the quarter. Still sitting at 90 per cent, because until that moment nobody had asked.

Here is the part people get wrong about this scene. Nobody in that room was lying.

The rep believed the deal. The regional head believed the rep. The CRO believed the regional head. The board believed the CRO. The number arrived upstairs with four layers of confidence attached to it and not one layer of evidence underneath.

That meeting was not badly run. It was well run, on the wrong currency.

It ran on confidence. Confidence compounds as it travels up a hierarchy, which is exactly why forecasts miss by 40 per cent and everyone is surprised.

A pipeline review that holds a forecast inside 5 per cent runs on something else entirely. It runs on evidence, and it is a genuinely different meeting. Different questions, different sequence, different things banned from the room.

This edition is that meeting.

The Thesis

Your forecast is not a prediction problem. It is a meeting design problem.

Most sales leaders try to fix forecast accuracy by changing the maths. Weighted probabilities. Stage-based percentages. A new dashboard. An AI tool that scores deals on engagement signals.

None of it works, because the input is the problem. If eleven reps walk into a room and describe eleven deals in the language of confidence, no model applied to that data produces truth. You are averaging opinions and calling it a forecast.

The pipeline review is where the input gets made. It is the only hour in the month where a deal's status is set, defended or corrected. Whatever that meeting rewards is what reps will bring to it.

Reward confidence and you get confidence. Reward evidence and you get evidence.

That is the whole edition in two lines. Everything below is the mechanism.

Why Does a Committed Forecast Miss by 30 to 60 Per Cent?

Quick Answer: A committed forecast misses by 30 to 60 per cent because confidence compounds as it travels upward while evidence does not. Each layer of management adds belief without adding proof, so a rep's optimism becomes a regional head's assurance and then a board commitment. The gap is not caused by poor forecasting skill. It is caused by a review meeting that accepts a story where it should demand an artefact.

Watch what actually happens to a single deal on its way up.

The rep is optimistic, and reasonably so. He has had good calls. The buyer has been warm. He has invested four months and he wants the deal to be real, which is a human condition and not a character flaw.

The regional head has fifteen deals to cover in ninety minutes. That is six minutes each. Six minutes is enough to hear a story and nowhere near enough to test one. So she takes the rep's word, because the alternative is a meeting that runs to four hours.

The CRO sees a rolled-up number, not eleven conversations. By the time it reaches him, the deal is a row in a spreadsheet with a percentage beside it. The percentage looks like data. It is a feeling that has been through two rounds of formatting.

The board hears a commitment.

At no point did anyone lie. At no point did anyone ask for evidence either.

Median B2B forecast accuracy sits at around 71 per cent in 2026. That is the market average, and most leaders quietly assume they are better than average. Very few are, and the ones who genuinely are have almost always changed the meeting rather than the model.

Here is the test. Pull your last three quarters. For each deal that slipped or died after being committed, find the moment someone in a review asked for proof and did not get it. You will find it nearly every time, and it will usually be earlier than you expect.

What Should Reps Complete Before a Pipeline Review?

Quick Answer: Reps should complete a pre-review scorecard for every committed and best-case deal, submitted at least 24 hours before the meeting. The scorecard tests ten binary conditions covering the buyer's stated problem, a number they calculated, the cost of inaction, the trigger, recency of contact, access to the economic buyer, the signature path, written confirmation, the buyer's own decision steps, and a scheduled next meeting. Deals score out of ten. The score, not the rep's opinion, decides which forecast category the deal sits in.

The point of the scorecard is not the scoring. It is that the argument happens before the meeting rather than during it.

A rep filling this in on Tuesday afternoon discovers on his own, in private, that he cannot answer four of the ten. He now has 24 hours to go and get the answers, which is exactly what you wanted him doing anyway. The deals that arrive in the room have already been improved by the act of preparing for it.

Ten conditions. One point each. No half marks, because half marks are where confidence sneaks back in.

INSERT IMAGE HERE: baseline-ed8-scorecard.png, saved in CLAUDE OUTPUTS. 1600px wide, BigLeaps palette. LinkedIn articles do not render tables, so the ten-point scorecard goes in as an image. Same approach as Editions 1, 3 and 7.

1. The buyer's problem is recorded in their own words, verbatim
2. There is a number attached, and the buyer calculated it
3. The buyer has stated what twelve more months of the problem costs them
4. A specific trigger event is named
5. Substantive contact with the buyer inside the last fourteen days
6. The economic buyer has been met at least once, by someone on our side
7. We know who signs the contract, and we know it from them
8. The buyer has confirmed our written summary of their situation
9. We have their decision steps and dates, in their words, not our assumptions
10. There is a next meeting in the diary with an actual date

8 to 10: commit. 5 to 7: best case, and name what is missing. Below 5: pipeline. Not forecast, whatever the rep feels about it.

The first month you run this, expect your commit number to fall by a third or more. That drop is not a crisis. That drop is the variance you were going to discover in week twelve, arriving in week one instead, while you can still do something about it.

What Is the Right Agenda for a Pipeline Review?

Quick Answer: A pipeline review should run 60 minutes with five fixed blocks: the number and last quarter's variance (5 minutes), scorecard roll-up and reclassification (5 minutes), inspection of mid-scoring deals only (30 minutes), slippage review of anything committed in a previous month (10 minutes), and named commitments with owners and dates (8 minutes). Strong deals get no airtime. Weak deals get no argument. The meeting spends its time exclusively on deals where the outcome is still movable.

Most reviews spend their hour in exactly the wrong places. Ten minutes celebrating the deal that was always going to close. Twenty minutes arguing about the deal that was never real. Four minutes on the one that could still be saved.

Invert it.

Minutes 0 to 5. The number. Read out the committed total. Then read out last quarter's variance against commit. Every month, without commentary. It takes 90 seconds and it sets the currency of the room before anyone speaks.

Minutes 5 to 10. Reclassification. Scorecards came in yesterday. Anything below 5 moves out of forecast now. Not discussed, not defended, just moved. Reps get one sentence to note anything the scorecard missed. This block is quiet and it is the most valuable five minutes in the meeting.

Minutes 10 to 40. Inspection. Only deals scoring 5 to 7. These are the deals where your intervention still changes the outcome. Roughly five deals at six minutes each. The rule inside each six minutes is strict: you discuss the missing conditions only. If a deal scored 6 out of 10, you talk about the four it failed. Nothing else.

Minutes 40 to 50. Slippage. Every deal committed in a previous month and still open. One question each: what has changed since you last committed this? A deal that slips twice is not a deal with a timing problem. It is a deal with a truth problem, and this is where you find it.

Minutes 50 to 58. Commitments. One action per inspected deal. An owner. A date. Written down where everyone can see it, and read back at the start of next month's slippage block.

Minutes 58 to 60. The new number. Restate it after the meeting has changed it. People need to hear that the number moved because the evidence moved.

Four things are banned from the room. Percentages. The word "feel". Updates on deals scoring 8 or above. Any discussion of a deal where nobody has contacted the buyer inside a fortnight, because there is nothing to discuss until somebody does.

What Questions Expose a Deal That Is Not Real?

Quick Answer: Twelve questions, three at each base, expose a fictional deal faster than any scoring model. First base tests whether the buyer has a reason to be talking to you. Second base tests whether a quantified problem exists in the buyer's own words. Third base tests whether you understand how the decision actually gets made. Home plate tests whether a signature is possible on the date you have forecast. A deal that cannot answer three questions at any single base does not belong in commit.

Ask these in the inspection block. The rep answers, not the manager.

First base. Is there a reason to be here?

1. Why is this buyer giving us their time? What changed for them?
2. Who else in their market are they talking to, and why us?
3. If they did nothing at all, who inside their business would notice?

Second base. Is there a case?

1. Read me their problem in their words. Not our summary of it.
2. What number did they calculate, and how did they arrive at it?
3. What did they say twelve more months of this costs them?

Third base. Do we understand the decision?

1. Walk me through their approval steps. Who is involved, in what order, and how do you know?
2. Who signs the contract, and who told you that?
3. What happens to our proposal in the room we are not in?

Home plate. Can this actually close?

1. What has to be true for the paperwork to be signed by your forecast date?
2. What is procurement going to ask for that we have not prepared?
3. If this slips one quarter, what is the reason it slips?

Question 12 is the sharpest of the twelve. A rep who cannot name the most likely reason his own deal slips has not thought about the deal seriously. A rep who names it instantly usually gives you the risk you needed to hear a month before it materialises.

The Four Questions to Stop Asking

Some questions actively make the forecast worse. These four are the most common.

"What percentage is this deal?" A percentage is a feeling wearing a number's clothes. It invites the rep to summarise rather than evidence, and it gives everyone above him something that looks like data. Replace it with: which of the ten conditions does this deal fail?

"Are they still interested?" Interest is not a forecast input. Buyers stay interested in things they never buy. Replace it with: what has the buyer done in the last fortnight? Actions forecast. Attitudes do not.

"What is your gut on this one?" You are explicitly asking for the thing that caused the 40 per cent miss. Every time a leader asks this, the room learns that instinct is acceptable currency. Replace it with: what would have to change for you to move this out of commit?

"Can we pull it into this quarter?" This one is the most damaging, because it tells the rep what answer you want. He will find a way to say yes. Replace it with: what is the buyer's own timeline, in their words, and what drives it?

How Do You Install This Without the Whole Thing Collapsing in Month Two?

Quick Answer: Install it over five weeks rather than in one meeting. Run the scorecard for a month before changing the meeting, so you have a baseline. Reclassify in week two and expect the commit number to fall by a third. Protect that fall with the board before it happens, because the political risk to the sales leader is the single most common reason this initiative dies. Change the meeting format in weeks three and four, and expect regional heads rather than reps to be the source of resistance.

The mechanics are simple. The politics are not, and the politics are what kills it.

Week 1. Measure, do not change. Reps complete scorecards. The meeting runs exactly as it always has. Nobody is corrected. You are buying a baseline, and you are letting people see their own scores without consequence attached. Reps who discover their best deal scores 4 out of 10 tend to fix it quietly without being told to.

Week 2. Reclassify, and take the hit. This is the week the number falls. A third is normal. Half is not unusual in a team that has never been inspected.

Before this week happens, have the conversation with your CEO or your board. Frame it precisely: the number is not falling, the number was never there. You are moving from a reported figure to a real one, the gap will be visible for one quarter, and after that the forecast becomes something they can plan against. Leaders who skip this conversation get accused of sandbagging in week three and abandon the whole thing by week five. I have watched it happen more than once, and the process was never the problem.

Weeks 3 and 4. Change the meeting. New agenda, new questions, banned list on the wall. Expect the resistance to come from your regional heads, not your reps. Reps are relieved, because the scorecard gives them cover to say a deal is not ready. Regional heads have been rewarded for years on the confidence of the roll-up, and you have just removed the instrument they managed with.

Deal with that directly. Their job changes from defending a number to improving deals, and that is a better job. Say so out loud, and coach them on the twelve questions before you expect them to ask them well.

Week 5. The slip. Somebody will bring a deal to commit that scores 6, with a good story about why it is an exception. There are no exceptions in month one. How you handle this determines whether the system survives the quarter. Move the deal, name the reason without blame, and move on inside sixty seconds.

Month 3 to month 6. Variance closes. Single digits by the end of the second quarter is a realistic target for a team that holds the discipline. Under 5 per cent belongs to teams still running it in month six, where the scorecard has stopped being an imposition and started being how reps think about their own deals.

That is the honest timeline. Anyone promising you a fixed forecast in thirty days is selling you a dashboard.

FAQ

Q: How often should we run this? Weekly or monthly?

A: Monthly for the full commit review described here. Weekly for a shorter version covering only deals that moved, which should take 30 minutes. Weekly full reviews sound rigorous and produce the opposite, because there is not enough new evidence in seven days to justify the meeting, so people fill the time with opinion.

Q: A deal scored below 5 and then closed. Does that break the system?

A: No, and it will happen. The scorecard is not predicting outcomes, it is measuring what you can evidence. A deal that closes without evidence closed by luck, and luck is not a forecasting method. Track these, because if more than one in ten of your closed deals came from below 5, one of your ten conditions is wrong for your market and needs replacing.

Q: Our CRM cannot be changed. Can we still do this?

A: Yes. Run the scorecard in a shared spreadsheet for the first quarter. The CRM change makes it durable, but the meeting change is what produces the result, and the meeting change costs nothing. Do not let a CRM roadmap delay this by two quarters.

Q: Will reps not just learn to tick the boxes?

A: Some will try in month one. It stops for a simple reason. Every condition on the scorecard requires something from the buyer, not from the rep. You cannot fabricate a written confirmation from a buyer, or a next meeting in a shared diary, or the buyer's own decision steps in the buyer's language. Conditions that can be self-certified are the ones that get gamed, and there are none on this list.

Q: How does this sit with MEDDIC, if we already use it?

A: Comfortably. MEDDIC is a qualification checklist and the scorecard is a forecast gate, so they solve different problems. If your team already scores deals on MEDDIC, map your existing fields onto the ten conditions rather than running two systems. What matters is that the meeting demands evidence, not which vocabulary the evidence arrives in.

Q: How long before my board sees the difference?

A: They see the number fall in month one and they see variance close by the end of month six. The credibility gain arrives earlier than the accuracy gain, usually around month three, when you correct a forecast downward before the quarter ends rather than after. Boards forgive a miss they were warned about. They do not forgive a surprise.

Closing

If this edition was useful, three asks.

1. Subscribe so the next edition lands in your feed automatically.
2. Forward it to one Sales VP or CRO whose last quarter missed commit by more than 20 per cent.
3. Reply with the topic you want me to write about next.

Next Saturday, Edition 9: why sales transformation fails in Indian B2B tech, and what actually works. The four failure patterns I see repeatedly, and the one condition that separates the engagements that stick from the ones that fade by month three.

Until Saturday.