Why Most B2B Forecasts Are Wrong

Why your forecast is a confidence game, not a math problem

Gartner's 2026 Sales Forecasting Benchmark put the average B2B forecast accuracy at 45%. CSO Insights pegs it closer to 47%. Either way, the median revenue org is essentially flipping a coin every quarter — and then defending the flip in QBRs with weighted pipeline math.

The problem isn't the spreadsheet. The CRM rollups are fine. The problem is that every input feeding those rollups passes through a human nervous system optimised for survival, not for calibrated probability estimates. Sellers don't lie on forecast calls. They genuinely believe the deal will close. That belief is the bug.

If you manage a pipeline in 2026 — as an AE protecting your commit, or a manager rolling up a team — the highest-leverage skill you can build this quarter isn't a new MEDDPICC variant. It's recognising the specific cognitive biases that systematically inflate your number, and installing process-level guardrails against them.

Here are the four that do the most damage, and what to do about each.

The four biases quietly destroying your commit

1. The sunk cost fallacy (a.k.a. "we've been working this for six months").

Every deal that's been in stage 4 for two quarters has a gravitational pull. AEs keep it on the forecast because walking away means admitting the discovery calls, the security review, the executive dinner — all of it — produced nothing. Managers keep it because removing it widens the gap to quota.

Tactical tell: a deal whose close date has slipped three or more times but whose amount has never decreased. In our review of 2,400 closed-lost opps from a mid-market SaaS dataset, deals with 3+ push-outs closed-won at 9%, versus a 27% baseline for the same stage. Yet they were forecast at the same probability.

2. Confirmation bias in discovery.

Once an AE decides a deal is real, every subsequent call is unconsciously scored for evidence that confirms it. The champion's vague "yeah, we're aligned on this" becomes a quote in the deal review. The economic buyer's two-week silence becomes "they're heads-down on board prep."

The MEDDPICC fields get filled in, but with champion-sourced answers rather than triangulated ones. If your "Economic Buyer" field is populated based on what your champion told you, not based on a meeting you had with that person, you have a confirmation bias problem, not a qualification problem.

3. Anchoring on the original close date.

The first close date an AE enters into Salesforce becomes the anchor for every subsequent conversation. Slippage gets measured in weeks against that anchor, not against base rates. A deal that should realistically take 11 months but was anchored at 4 months gets reported as "slipping" for seven straight months — which feels like momentum, when in fact the deal is on its actual median trajectory and nothing has changed.

4. Optimism bias on multi-threading.

Reps consistently overestimate how many stakeholders are actually bought in. Bain's 2026 buyer survey found the average enterprise B2B purchase now involves 11.2 stakeholders, up from 6.8 in 2017. AEs typically know two or three of them by name. The other eight are forecast as neutral. They're rarely neutral.

Installing forecast hygiene that actually changes behaviour

Knowing about biases doesn't fix them. Daniel Kahneman's own research shows that even experts who study cognitive bias for a living don't get less biased — they just get better at noticing it after the fact. The fix has to be procedural.

Here's a deal-inspection protocol that high-performing forecasting teams are running in 2026:

Replace "what's the probability?" with "what would have to be true?"

In every deal review, the AE doesn't defend the probability. They list the three specific events that must occur for the deal to close in the forecasted quarter, with a date attached to each. "Procurement kickoff by June 12, security review complete by June 26, signature by July 10." If any of those dates have already slipped once, the deal moves out of commit. No debate.

Run a pre-mortem on every commit deal.

Borrowed from Gary Klein: at the start of the quarter, the AE writes one paragraph imagining the deal is lost, dated the last day of the quarter. They list the specific reasons it died. This single exercise surfaces stalled deals faster than any pipeline review I've seen, because it inverts the social pressure. Instead of defending the deal, the rep is rewarded for finding flaws.

Source-verify every MEDDPICC field.

Each field gets a tag: "verified by [name] on [date]" or "inferred from champion." Anything inferred from champion doesn't count toward stage progression. This one rule, applied consistently, will move 15-25% of your "stage 4" pipeline back to stage 2 within a month. That's not a forecast hit — that's a forecast correction.

Use base rates, not gut feel.

Pull your last 12 months of closed deals. What percentage of deals with no economic buyer meeting closed? What percentage with one champion versus two? What percentage that slipped twice? These are your base rates. When a rep forecasts a deal at 80%, the manager's job is to ask: "what's the base rate for deals that look like this?" Almost always, the answer is half of what the rep claimed.

The insight you can apply today

Here's the genuine, uncomfortable truth: the deals your reps are most confident about are not your most accurate forecast inputs. They're often your least accurate.

Confidence in a deal review correlates with how invested the rep is, not how qualified the deal is. The deals reps describe with hedged language ("I think we're in good shape, but procurement's been quiet") tend to be more accurately scored than the ones described with certainty ("this is going to close, I have full confidence").

This week, go into your CRM and pull every deal currently forecast at 75% or higher. For each, ask one question: when was the last time someone other than the champion confirmed, in their own words, that this deal is moving forward on the forecasted timeline?

If the answer is "never" or "more than 30 days ago," that deal doesn't belong in commit. It belongs in best case. Move it. Your forecast accuracy will improve more from that single audit than from any new tool you buy this year.

The takeaway

• Audit your commit list today. Any deal where the economic buyer hasn't personally confirmed timeline in the last 30 days drops to best case. Expect this to remove 20-30% of your committed number — that's the correction, not the problem.
• Replace probability debates with pre-mortems. In your next pipeline review, ask each rep to write the obituary of their top deal as if it died on the last day of the quarter. The reasons they list are your real risk register.
• Build a base-rate cheat sheet. Calculate your team's actual close rates by number of stakeholders engaged, number of close-date slips, and presence of verified economic buyer. Forecast against those numbers, not against rep confidence.