Forecasting Accuracy That Survives the QBR

Most B2B forecasts miss by ±20-40% per quarter. Reps aren't lying when this happens. They're optimizing the commit number to keep their manager off their back, which is a different problem and a worse one. If you own the forecast, you own that gap. The QBR is where the gap stops being an operational issue and becomes your performance review.

This is the operating system I run when I'm sitting in the RevOps Manager seat and the number going to the CFO has my fingerprints on it. Cadence, categories, deal review, the human triangle, and the diagnostics that catch a miss six weeks before it lands. None of it is glamorous. All of it works.

The Forecast Is a Process, Not a Number

Read that again, because it's the only thing in this guide that matters.

When the CRO says "what's the number," they're asking the wrong question. The right question is "what's the number, and what's the variance band around it, and what assumptions am I making, and what would have to be true for it to be wrong by 20%?" That's not a number. That's a process running every week, producing a number on demand, with the math exposed.

A RevOps Manager who can't answer all four parts of that question is going to get torn up at the QBR. Not because the number was wrong (numbers are wrong all the time), but because nobody could explain why it was wrong. The QBR doesn't test whether you hit the number. It tests whether you knew you weren't going to hit it, and when you knew it.

Forecast Cadence: The Week-Over-Week Roll-Up

Forecasts that hold up have a cadence. Forecasts that blow up have a vibe. Here's the cadence I run.

Four artifacts, four owners, four days. Friday is for closing deals, not forecasting.

The roll-up matters because each layer adds judgment and removes information. Reps know the buyer's org. Managers know the rep. The CRO knows the board. RevOps owns making sure those three views reconcile.

The reconciliation is where most teams fail. Without a Thursday RevOps step, "Wednesday's CRO commit" becomes the forecast of record, and nobody checks whether it ties back to anything the reps actually believe. Six weeks later, you've got a $4M number on the board and a $3.1M reality on the ground.

Track Commit Drift Religiously

Every week, take each commit deal and check whether it was commit last week. Build a simple table: deal name, weeks in commit, original close date, current close date, current value, value change.

A deal that's been "commit" for six weeks is not a commit. It's a wish. Real commit deals close in two-to-four weeks. If you've got a deal sitting in commit for an entire quarter, you've got a rep who's terrified to push it back, a manager who isn't scrubbing, and a forecast that's about to detonate.

I worked with a Series C SaaS company where 38% of their commit dollars on week 4 were deals that had been commit for two-plus quarters. I built the drift report on a Monday. By Wednesday, $1.6M of "commit" had been moved to best case, the CRO had a real conversation with the board, and the actual close on quarter end came in within 4% of the revised number. The original commit would have missed by 22%.

Drift is the single highest-signal metric in forecasting. If you do nothing else from this guide, build the drift report.

Forecast Categories That Actually Mean Something

Most categories are theater. "Commit" means "the rep's manager pressured them into committing." "Best case" means "the rep wants credit if it lands." "Pipeline" means "everything else, sorted by hope."

Real categories have hard definitions a new rep can apply on day one without asking.

Commit: "I will personally guarantee this closes this quarter, and if it slips I owe you an explanation." Not "should close." Not "looking good." Personal guarantee. If the rep can't say that out loud to their manager's face, it's not commit.

Best case: "Realistic upside if two specific things break right." Two specific things. Named. "If procurement signs off by Friday and the second-level economic buyer says yes on Tuesday." Not "if the stars align." Vague best case is fantasy best case.

Pipeline: Everything that isn't commit or best case. By definition, it shouldn't move into commit this quarter. If it does, your stages are wrong.

The benchmarks tell you whether the categories are working:

If your commit closes at 70%, your reps are sandbagging the definition. They're calling things "commit" that they don't actually believe. Fix the definition before you fix anything else.

The 10-Question Deal-by-Deal Review

For every deal over $50K in the commit category, the rep answers ten questions cleanly. Not "kind of." Cleanly, with names, dates, and specifics. If they can't answer eight of ten, the deal is best case at most. Period.

1. Who is the economic buyer? (Name, title, last contact date.)
2. Is there a mutual close plan signed by the buyer and the rep?
3. Has procurement been engaged, and what's their stated timeline?
4. Are legal redlines started, and on which paper?
5. What's the security review status (submitted, in progress, approved, or not started)?
6. Has the internal champion been stress-tested? (When did they last advocate for us in a meeting we weren't in?)
7. What's the competitor displacement plan, and which specific competitor are we displacing?
8. When did the exec sponsor last engage in the past 14 days, and on what?
9. Is the contract value locked or floating? (If floating, what's the range?)
10. What would cause this to slip, and what's the slip risk percentage?

Run this review weekly with managers, not reps. Reps prep the answers Monday. Managers stress-test them Tuesday. RevOps spot-checks the worst three each week and pulls the call recording or the email thread to verify.

The first month you run this review, you're going to lose 20-30% of your commit dollars. Don't flinch. The dollars weren't real. You're just surfacing the truth earlier than the quarter close would have surfaced it.

The AE-CRO-RevOps Triangle

Each role owns a different forecast lens. The triangle holds when each role does its job. It breaks when one role tries to do another's.

• AE owns the deal-level truth: what's actually happening in the buyer's org, what the procurement timeline really is, whether the champion is bluffing.
• CRO owns the number: what we're telling the board, the financing partners, the leadership team.
• RevOps owns the math: how those two reconcile, where the gaps are, what the variance looks like, what the drift is doing.

The triangle breaks in three predictable ways.

The AE optimizes for cover. They sandbag commit because their manager will yell at them less for sandbagging than for missing. The forecast comes in low and the company under-resources hiring, marketing spend, and inventory.

The CRO optimizes for narrative. They want a number that tells a good story to the board. RevOps gets pressured to "find the missing $2M" in pipeline that doesn't exist. The narrative carries one quarter and explodes the next.

RevOps becomes the messenger getting shot. The AE tells RevOps the deal is at 50%, RevOps reports 50% to the CRO, the CRO doesn't like 50%, and somehow it's RevOps's fault.

You hold the triangle by surfacing the gap with data, not opinions. Don't say "I think the number is high." Say "the rep aggregate is $3.6M, the manager aggregate is $4.4M, the AI scoring is $3.8M, and we're publishing $4.4M. The variance is wider than any prior quarter at this point." Then let the room decide what to do.

Manager Forecast vs Aggregated Rep Forecast

This is the single most useful diagnostic in forecasting, and almost nobody runs it.

Take every rep's commit number and sum them up. Then ask each frontline manager, independently, without seeing the rep aggregate, for their commit number. Compare the two.

Worked example: 8 reps on a team. Rep aggregate is $4.2M. Manager independent aggregate is $3.6M. Delta is -$600K, or -14%.

That delta is the sandbag-or-stretch indicator.

If rep aggregate > manager aggregate, your managers don't trust their reps. The reps are stretching, the managers are deflating, and the truth is probably in the middle but closer to the manager number.

If rep aggregate < manager aggregate, your managers are stretching to please the CRO. The reps are giving honest commit, the managers are adding "stretch deals" that shouldn't be in commit, and the truth is closer to the rep number.

Both are problems. Neither is fixable with more forecasting tools. They're fixable with conversations the CRO probably doesn't want to have.

Track the delta weekly. Plot it. The pattern over a quarter tells you exactly where the cultural problem is.

AI and Clari Inputs: What They Help With, What They Don't

Tools like Clari, Gong, BoostUp, and Aviso are good at one thing and bad at another. Be specific about which.

What they help with: engagement signals. Email volume into the buyer org. Multi-threading depth. Late-stage activity decay. Meeting cadence dropping off. The kind of signals that show up in CRM and email metadata, summed across deals at scale.

What they don't help with: buyer org changes (a champion just quit), budget reallocations (CFO froze SaaS spend yesterday), competitor cycles (the incumbent just dropped their renewal price), legal cycles (their GC is on vacation for two weeks).

The biggest mistake I see RevOps Managers make: shipping a forecast number that's just the Clari prediction. The AI scoring is a find tool, not an answer tool. It tells you which deals to question. The answers come from the rep's mouth and the manager's judgment, not from the model.

Manager judgment plus AI signal beats either one alone. Manager judgment alone misses deals that look fine but have decaying engagement. AI alone misses deals that look engaged but have a champion who just got passed over for promotion.

Use AI to surface the questions. Don't let it answer them for you.

The "We're Going to Make It" Denial Diagnostic

Every quarter has the same conversation around week 9. Sales leadership says "we're going to make it" with no math behind it. The room nods. Two weeks later, you miss by 18%.

The RevOps Manager owns calling that out. Specific tells:

• Pipeline coverage dropped below 3x and nobody's flagging it.
• Commit is unchanged from week 6 to week 9. (Impossible. Deals move. Commit not moving means nobody's actually scrubbing.)
• Best case suddenly grew $2M. Where did it come from? New deals don't enter best case at week 9.
• No deals slipped out. (Also impossible. Deals always slip.)
• Drift report shows the same five deals in commit for the past five weeks.

Give yourself the language to ask the uncomfortable question without becoming the villain. My favorite version: "I want to make sure I'm reading the data right. Can someone walk me through what's changed in the last three weeks that increased our confidence?" If the answer is hand-waving, you have your answer.

±10% Accuracy by Week 8: The Actual Goal

Forecast accuracy at week 1 of the quarter is fiction. Don't sweat it. Don't even publish it externally.

By week 4 it should be ±20%. By week 8 it should be ±10%. By week 12 (close) it should be ±5%.

If you're still ±20% at week 8, one of three things is wrong:

1. Your categories are wrong. Commit isn't closing at 90%, best case isn't closing at 50%, the definitions don't hold.
2. Your deal review is rubber-stamping. Reps fill in the 10 questions with "yes/yes/yes" and managers don't push back.
3. Your managers aren't scrubbing. The Tuesday step is happening on paper but not in practice.

Instrument it. Track forecast vs actual at week 4, week 8, and week 12 across the last four quarters. Plot the variance. The pattern will tell you which of the three is broken. Usually it's all three, but in different proportions.

A team that gets to ±10% by week 8 has a forecast process that survives any QBR. A team that's still ±20% at week 8 has a vibe.

What the QBR Actually Tests

Back to the opening point. The QBR isn't asking "did you hit the number." Misses happen to the best teams. The board knows that.

The QBR is asking "did you know you weren't going to hit it, and when did you know it?" A RevOps Manager who flagged the gap at week 6 looks competent even on a 12% miss. A RevOps Manager who was surprised on the last day of the quarter looks like they weren't doing their job.

The forecast process is your insurance policy. The cadence, the categories, the 10-question review, the rep-vs-manager delta, the drift report: all of it exists so that when the miss comes (and one will, eventually), you can walk into the QBR with the data showing exactly when you saw it, what you flagged, and what was decided. That's a competent RevOps Manager.

The number is a number. The process is the job.

If you're stepping into this seat for the first time, the companion JD walks through the day-to-day reality of the role. The forecast process is the spine of that role.

Learn More

• The First 90 Days as a RevOps Manager
• Pipeline Hygiene That Finance Trusts
• RevOps Metrics: Coverage, Accuracy, Win Rate
• Companion JD: Revenue Operations Manager