How do top revenue leaders run a forecast call that prevents sandbagging without creating a blame culture in 2027?

The outcome you should expect

The measurable outcome of a well-run forecast call is not "more accurate reps." It's a narrower, more stable distribution of error at the team level, achieved without the meeting becoming an interrogation. Those two goals sound like they fight each other. They don't, once you stop asking reps to be prophets and start asking them to be reporters.

Sandbagging is a rational response to an incentive structure, not a character flaw. A rep who is measured against their own call, and punished when they miss it, will systematically under-call. That's not deceit — it's the only defensible strategy available to someone whose downside for over-calling is a difficult conversation and whose downside for under-calling is a pleasant surprise. Every seasoned seller learns this within two quarters. The same rep on a different team, with a different incentive structure, will call honestly. The person didn't change; the payoff matrix did.

So the first outcome to expect from a redesigned forecast call is a change in the shape of error, not just the size. Before the change, a sandbagging team's error is skewed one direction: the team consistently lands above its own call, and leadership silently applies a "multiplier" — everyone knows the number is inflated downward, so leadership adds 15–25% back and calls it experience. That's a broken system with a duct-tape patch on top. The patch itself is corrosive, because it means the rep's call carries no information. It's a ritual number.

After the change, you should see three things. First, the *directional bias* flattens — the team lands both above and below its call, roughly symmetrically. That's the single cleanest signal that sandbagging has stopped, and it's easy to measure: track signed error (not absolute error) by rep over eight quarters and look for a mean near zero rather than persistently negative.

Second, the *variance* tightens more slowly than the bias flattens. Bias correction is a behavior change and can land in one or two quarters. Variance reduction is a data-quality and process-maturity change and typically takes three to six quarters, because it depends on the underlying deal inspection getting sharper — better next-step hygiene, real mutual action plans, procurement timelines captured before the last two weeks of the quarter.

Third, and least discussed: *meeting duration drops*. A blame-culture forecast call is long because it's adversarial — reps defend, managers probe, everyone performs. An evidence-based call is short because the evidence either exists or it doesn't. Teams that make this transition well often cut a 90-minute weekly call to 35–45 minutes, and push the deal-level surgery into separate, smaller deal reviews where it belongs. Mixing forecast roll-up with deal coaching in one meeting is one of the most common structural mistakes, and it's the mechanism by which the call becomes a blame venue: you're inspecting a person's judgment in front of their peers while simultaneously asking them for a number.

There's an adjacent outcome worth naming, because it shows up downstream and surprises people. When forecast calls stop being punitive, CRM hygiene improves without a hygiene mandate. Reps stop hiding deals until they're safe. Early-stage pipeline that used to appear out of nowhere at 60% probability starts appearing at creation, in stage one, with a real close date. That has a second-order benefit for marketing attribution and capacity planning that nobody puts in the business case, and it's often the biggest ROI of the change.

What drives that outcome

Three mechanisms drive it, and they have to move together. Change one alone and the system snaps back within a quarter.

Mechanism one: decouple the call from the comp and review cycle. As long as the forecast number is an input to a performance conversation, it will be gamed. The rep's committed number should be treated as a *report on the state of the pipeline*, not a *promise about their effort*. Concretely: quota attainment is measured on closed-won revenue. Forecast accuracy is measured separately, as a process metric, and it is measured *symmetrically* — a rep who calls $400K and closes $600K is exactly as inaccurate as one who calls $600K and closes $400K. The moment you score over-delivery as a win and under-delivery as a failure, you have re-created the sandbagging incentive with extra steps.

Mechanism two: move from adjectives to artifacts. "Feeling good about it," "champion is strong," "verbal yes" — these are adjectives. They cannot be verified, so the conversation about them degenerates into a contest of confidence, which is exactly the terrain where blame culture grows. Artifacts are different: a calendar invite for the next meeting with the economic buyer, a written mutual action plan the buyer has edited, a security questionnaire in flight, a procurement contact name, a signed order form draft in the redline stage. An artifact-based call asks one question per deal — "what's the evidence for the close date?" — and the answer is either a thing that exists or it isn't. Nobody's character is on the line.

Mechanism three: give the rep a range, and let the system produce the point estimate. This is the piece most teams skip. Asking a human for a single number forces them to embed their risk tolerance in it, and risk tolerance is where sandbagging lives. Ask instead for a commit / most-likely / upside triplet, or a low-high band. The rep supplies judgment about their deals; RevOps supplies the aggregation math. When the roll-up is computed rather than declared, the rep isn't defending a number — they're describing a distribution, which is a much less threatening act.

The loop matters more than any single box. The reason blame culture forms is that a miss has no designated destination — so it lands on a person. Give the miss somewhere to go (a process review that updates stage criteria) and the same event that used to produce blame now produces a system improvement. This is straight out of operational quality practice, and it transfers cleanly from manufacturing and incident response into revenue: blameless post-mortems work in RevOps for the same reason they work in engineering — the goal is to find the flawed signal, not the flawed human.

A fourth driver deserves mention because it's increasingly the deciding factor in 2027: the model in the loop. Most CRM platforms now ship some form of probabilistic scoring built on engagement signals, historical stage-conversion, and deal-shape similarity. Used well, this is enormously helpful — it gives you an independent second opinion that has no career incentive. Used badly, it becomes a new blame vector: "the model says 80%, why did you call it 40%?" The discipline is to treat model output as a *third-party estimate to be reconciled*, not a verdict. When the rep and the model disagree, that disagreement is the most valuable minute of the entire call — one of them knows something the other doesn't, and surfacing which is the whole point.

Benchmarks and realistic ranges

Precise industry-wide forecast accuracy benchmarks are notoriously unreliable, because definitions vary wildly — accuracy measured against week-one call is a different number than accuracy measured against week-eleven call, and few published figures specify which. So treat everything here as directional and calibrate against your own history rather than against a headline stat.

Cadence. The dominant pattern in enterprise B2B is a weekly roll-up call plus a separate deal review cadence. For teams with a 90-day average sales cycle, weekly is right. For teams with a 30-day cycle in transactional SMB motion, weekly forecast calls are often too coarse to be useful — the whole quarter turns over three times — and daily or twice-weekly pipeline standups paired with a monthly roll-up work better. For teams with a 9–18 month enterprise cycle, weekly calls become theater by week four; biweekly with a monthly deep-dive fits the actual rate of information change.

Duration. A roll-up call should be 30–45 minutes for a team of 6–10 reps. If it runs past an hour consistently, you are doing deal coaching in the wrong room. The tell: a single deal consumes more than four minutes. That deal needs a separate 30-minute working session with the right people, not fifteen spectators.

Categories. Three to four categories is the practical ceiling: Commit, Best Case, Pipeline, and (optionally) Omitted/Excluded. Beyond four, category discipline collapses and everything drifts to the middle. Each category should have a *written definition tied to evidence*, not a probability percentage. "Commit = mutual action plan signed, economic buyer engaged in the last 14 days, procurement path identified" is a definition a rep can meet or not meet. "Commit = 90% likely" is a vibe.

Error tolerance. Rather than chasing a universal accuracy target, set an internally-derived tolerance band and tighten it. A reasonable practice: measure your team's historical absolute percentage error over the last eight quarters at the same point in the quarter, take the median, and set the target 20% tighter. If your week-six roll-up has historically landed within 18% of actual, target 14% and see whether the process changes move it. Teams that have never measured this are frequently shocked at how wide the band is — it is common for a team that believes it forecasts well to be running double-digit error late in the quarter.

Slip rate as the leading indicator. Absolute accuracy is a lagging measure; you only learn it at quarter end. Slip rate — the percentage of commit-category deals whose close date moves out in a given week — tells you the same story weeks earlier. Track it per rep and per segment. A healthy commit category has a low single-digit weekly slip rate. If 20% of commit deals are pushing every week, the category definition is being applied loosely and no amount of meeting discipline will fix it.

Coverage ratio, with a caveat. The familiar "3x pipeline coverage" heuristic is a blunt instrument that gets misapplied constantly. The correct coverage number is the inverse of your own stage-weighted conversion rate at the start of the period, and it differs enormously by segment — a high-velocity SMB team converting 40% of qualified pipeline needs roughly 2.5x, while an enterprise team converting 15% needs closer to 6–7x. Using a borrowed 3x across both segments guarantees one team is starved and the other is complacent.

Time-to-first-honest-call. The soft benchmark that predicts everything else: how long after the change do reps start calling deals *down* voluntarily, in the meeting, without being pushed? In teams that make the transition genuinely, this shows up within four to eight weeks. If three months in nobody has ever volunteered a downgrade, the psychological safety piece hasn't landed regardless of what the process document says.

Risks, edge cases, and failure modes

The sandbagging-to-happy-ears whiplash. The most common failure is over-correcting. You remove the penalty for missing a call, reps stop under-calling — and then some of them start over-calling, because optimism is now costless. The fix is symmetric scoring, applied visibly, from day one. Announce it before the first call under the new regime: over-call and under-call are the same error. If you only announce the removal of the downside penalty, you have simply moved the bias to the other side of zero.

Manager-level sandbagging. Reps aren't the only layer that hedges. A frontline manager who is measured on their team's forecast accuracy will shave the roll-up before it goes to the VP, and the VP will shave it before it goes to the CRO, and by the time it reaches the board it has been discounted three times. This compounding haircut is invisible unless you instrument it. The fix is to store and report the number *at every level* — rep call, manager call, VP call, actual — so the shave is visible as a delta rather than baked in silently. Most teams are startled the first time they see this chart.

The one rep who genuinely can't forecast. Symmetric, blameless scoring will still surface a rep whose calls are simply noise — high variance in both directions, quarter after quarter. This is a real coaching situation, and pretending otherwise is its own kind of dishonesty. Handle it in a one-on-one, framed as a skill gap in deal qualification (which is what it almost always is), never in the group call. The group call is for the system; the one-on-one is for the person. Keeping that boundary absolute is what makes the group call safe.

Evidence theater. Once "artifact present" becomes the gate, some reps will manufacture artifacts. A calendar invite sent to a contact who hasn't accepted. A mutual action plan the buyer has never opened. This is Goodhart's law arriving on schedule: the measure becomes a target and stops being a good measure. The counter is to require *buyer-side confirmation* on the artifact — an accepted invite rather than a sent one, a document with buyer edits rather than a shared link, a named procurement contact who has actually replied. Raise the artifact bar the moment you see gaming, and say plainly why you're raising it.

Quarter-end compression. Everything above degrades in the last ten days. Pressure spikes, and the culture you've built gets stress-tested exactly when it's weakest. The practical defense is to pre-commit the last-two-weeks protocol *before* the quarter's pressure arrives: what gets escalated, who calls the buyer, what discounting authority exists, and — critically — that the forecast number is frozen at a defined point and not re-litigated hourly. Teams that renegotiate the number daily in the final week teach their reps that the call was never real.

The model-as-cudgel failure. Mentioned above but worth restating as a risk: probabilistic scoring becomes toxic the instant a manager uses it to adjudicate rather than to inquire. It also fails quietly in the other direction — models trained on historical data reproduce historical bias, so if your team sandbagged for three years, the model has learned to sandbag too. Retraining after a process change takes several quarters of clean data. Don't trust the model's calibration during the transition period.

Segment and geography confounds. Rolling all segments into a single accuracy number hides real problems. Enterprise deals slip for structural reasons — procurement, security review, budget cycles — that SMB deals don't have, and a rep carrying two enterprise deals has irreducibly lumpy forecasting math regardless of skill. Same for a rep in a market with a long holiday shutdown. Segment your accuracy reporting or you will coach the wrong people. Similarly, a rep with a 6-deal quarter cannot be evaluated on forecast accuracy the way a rep with 60 deals can; the small-N problem is real and the honest move is to say so out loud rather than pretending the metric means the same thing for both.

The renewal and expansion blind spot. Most forecast call design assumes new logo. In a business where 60–80% of revenue is renewal and expansion, the same call has to handle a fundamentally different risk profile: renewals are high-probability-until-suddenly-not, and the sandbagging pattern inverts — CSMs tend to over-call renewals because churn signals are uncomfortable to surface. Run renewals through a distinct evidence standard (product usage trend, executive sponsor still employed, support ticket sentiment, contract auto-renew mechanics) and consider a separate call entirely. Applying new-logo forecast discipline to a renewal book is one of the most common structural mismatches in RevOps.

A practical rollout plan

Do not roll this out as an announcement. Announcements about culture change are correctly read by sales teams as noise. Roll it out as a sequence of small, concrete changes where each one is independently defensible.

Weeks 1–3: instrument silently. Before changing anything, capture the baseline. Store every rep's weekly call, every manager's roll-up, and the eventual actual, for at least one full quarter if you can afford the wait, or use whatever history exists in your CRM's snapshot tables. You need to know the current bias and variance per rep, per manager, per segment. Do not share this data yet, and do not use it retroactively — using baseline data punitively is the fastest way to poison the entire initiative.

Weeks 4–5: rewrite the category definitions. Convert every stage and forecast-category definition from probability language to evidence language. Do this *with* two or three respected reps in the room, not in a RevOps vacuum — the definitions have to be achievable in the actual sales motion, and reps will tell you immediately which criteria are unrealistic. Publish the definitions somewhere permanent and referenced, not in a deck.

Week 6: change the meeting, announce the scoring. Same week, both changes. Announce symmetric accuracy scoring and its separation from quota. Restructure the call: roll-up only, deal surgery moves to a separate session, ranges instead of point numbers. Say explicitly, in the first meeting, that a voluntary downgrade is a *good* contribution — and then when the first rep does it, respond in a way that visibly rewards it. That single moment does more than the entire policy document.

Weeks 7–14: hold the line through one full quarter. The transition period is where it breaks. Pressure will build to revert at quarter end. The two things that must not bend: deal surgery stays out of the roll-up call, and no rep is criticized in the group setting for a number. Managers will need coaching here more than reps will — the instinct to publicly probe is strong and usually well-intentioned.

Quarter 2: publish the accuracy chart. Once you have a clean quarter under the new process, publish forecast accuracy — symmetric, by rep, visible to the team. This is the step that feels risky and is actually the payoff. When the scoring is genuinely symmetric and genuinely separate from comp, publishing it creates constructive peer pressure rather than fear. If you don't have real confidence that the separation from comp is credible, delay this step; publishing too early re-creates the exact dynamic you removed.

Quarter 3 onward: tighten and automate. Now bring the model in as a reconciliation partner, tighten the tolerance band, and start automating evidence capture so reps aren't hand-entering artifacts — calendar integration for meeting confirmation, document tracking for mutual action plans, email signal for buyer engagement. Every artifact you can capture automatically is one less thing a rep can fake and one less thing they have to type.

One resourcing note: this is a RevOps program, not a sales-management initiative, and it fails when it's owned by whoever runs the meeting. Someone has to own the data pipeline that stores weekly snapshots, computes symmetric error, segments it correctly, and produces the chart. That's a real, ongoing responsibility — typically a fraction of one analyst's time, but it has to be assigned. The most common way this whole effort dies is that the baseline instrumentation stops running three weeks in and nobody notices until the quarter is over.

Related questions

What's the difference between sandbagging and prudent conservatism?

Conservatism is a consistent, disclosed discount applied for known risk. Sandbagging is an undisclosed, variable hedge that exists to manage the caller's downside. The test: if the rep would call it differently in a setting with no personal consequence, it's sandbagging.

Should forecast accuracy ever affect compensation?

Generally no for individual contributors — it re-creates the incentive to game. Some organizations tie a small component of frontline-manager comp to roll-up accuracy, which is more defensible since managers aggregate, but even there it tends to push the hedging one level up rather than eliminate it.

How do you forecast when a single deal is 40% of the quarter?

You don't forecast it, you scenario it. Report the number both with and without the deal, name the specific gating events, and give the board a range rather than a point. Pretending a binary outcome is a probability-weighted number misleads everyone downstream, especially finance.

Does AI-based deal scoring replace the forecast call?

No. Scoring replaces the *argument about probability*, which frees the call to do what humans do better: surface information not yet in the system. A buyer's CFO left last week — no model knows that until someone says it out loud.

How should the forecast call change for a PLG or self-serve motion?

Radically. With thousands of small transactions, individual deal inspection is meaningless; you forecast from cohort conversion curves and leading product-usage signals instead. The sales-assisted expansion layer on top still needs a traditional call, but the base is a statistical exercise owned by RevOps, not a meeting.

FAQ

Why do reps sandbag even when leadership says there's no penalty for missing?

Because reps calibrate to observed behavior, not stated policy. If a manager has ever visibly reacted badly to a miss — a sharp question in a group call, a mention in a QBR, a note of "reliability concerns" — that lands harder than any number of assurances. Trust in the new system is rebuilt through several consecutive quarters of consistent response to misses, not through announcements. The first miss under the new process is the real policy statement.

What single change gives the biggest reduction in sandbagging?

Symmetric accuracy scoring — treating over-delivery against the call as exactly the same magnitude of error as under-delivery. Most organizations quietly celebrate beating the number, which is precisely the reward that sustains sandbagging. Removing that asymmetry, and being visibly consistent about it, moves behavior faster than any tooling change or meeting redesign.

How do you keep the call from becoming an interrogation?

Structurally, not behaviorally. Move deal-level diagnosis into separate, smaller sessions so the roll-up call only reviews deltas and evidence gaps. Cap per-deal airtime at a few minutes. When a deal needs real work, the correct response in the roll-up is "let's take that offline Thursday," said every time without exception. Discipline about the room is more reliable than discipline about tone.

Should the forecast number be a single figure or a range?

A range from the rep, a single figure from the system. Humans are poor at producing calibrated point estimates and good at bounding outcomes. Collecting low/likely/upside and computing the roll-up centrally removes the rep's need to embed personal risk tolerance in one number, which is where most sandbagging originates.

How long before the change actually shows up in the numbers?

Directional bias typically flattens within one to two quarters, since it's a behavior response to a changed incentive. Variance — the actual precision of the forecast — improves over three to six quarters, because it depends on slower changes: better qualification, cleaner stage criteria, real mutual action plans, and enough clean post-change data to recalibrate any predictive model.

Does this approach work for renewals and customer success forecasts?

The principles transfer but the evidence standard must be rebuilt. Renewal risk is signaled by product usage decline, sponsor turnover, and support sentiment rather than by meeting artifacts. Also note the bias inverts: CS teams tend to over-call renewals because raising churn risk feels like admitting failure. The blameless framing matters just as much, applied to a different failure mode.

Sources

• https://hbr.org/2010/12/stop-losing-sales-to-customer-indecision — Harvard Business Review on sales process and buyer decision dynamics
• https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights — McKinsey Growth, Marketing & Sales insights on commercial operations
• https://sre.google/sre-book/postmortem-culture/ — Google SRE Book, "Postmortem Culture: Learning from Failure" (the blameless post-mortem model)
• https://www.gartner.com/en/sales — Gartner sales practice research on forecasting and revenue operations
• https://hbr.org/2019/07/how-to-make-your-sales-forecasts-more-accurate — Harvard Business Review on sales forecast accuracy
• https://www.salesforce.com/sales/analytics/sales-forecasting/ — Salesforce overview of sales forecasting methods and categories
• https://en.wikipedia.org/wiki/Goodhart%27s_law — Goodhart's law, the mechanism behind evidence theater
• https://rework.withgoogle.com/print/guides/5721312655835136/ — Google re:Work guide on psychological safety in teams
• https://otexts.com/fpp3/ — Hyndman & Athanasopoulos, *Forecasting: Principles and Practice* (forecast error, bias vs. variance)

Related on PULSE

• How do you set pipeline coverage ratios by segment instead of using a blanket 3x?
• What belongs in a deal review versus a forecast roll-up call?
• How do you measure forecast accuracy when one deal dominates the quarter?
• What evidence should define each sales stage instead of probability percentages?
• How do RevOps teams instrument weekly forecast snapshots in the CRM?
• How should renewal and expansion forecasting differ from new-logo forecasting?