Stage exit criteria decide when a deal is allowed to move forward in your pipeline. They sound like admin, but they determine what every stage means, and therefore whether your stage probabilities, weighted pipeline and forecast can be trusted. The key test: is each criterion something the buyer verified, or something the seller did?
Worked example: seller activity vs buyer evidence
| Stage | Seller activity (weak exit criterion) | Buyer-verified exit criterion |
|---|---|---|
| Discovery | Held a discovery call | Buyer confirmed a specific problem, who owns it, and a reason to act within a stated time |
| Evaluation | Delivered a demo | Buyer agreed written success criteria; the economic buyer is named |
| Proposal | Sent a proposal | Buyer received pricing and confirmed the scope matches their success criteria |
| Negotiation | Sent the contract; got a verbal yes | Buyer confirmed the selection in writing with an agreed close plan; their legal or procurement team is engaged (redlines or a vendor form received) |
Delivering a demo is necessary, but it is entirely in the seller's control. A demo can be given to the wrong person, with no agreed next step, and the deal still moves to Evaluation.
Why buyer-verified criteria matter
Consider a team whose Evaluation stage was defined as "demo delivered". Every deal that got a demo counted as Evaluation, whether or not the buyer had agreed what a successful evaluation would look like. The team then redefined the stage: a deal reaches Evaluation only when the buyer has agreed written success criteria. Four quarterly cohorts under each definition, all old enough to have outcomes:
| Evaluation defined as | Deals that reached the stage | Won | Stage-to-close win rate |
|---|---|---|---|
| "Demo delivered" (seller activity) | 300 | 45 | 45 ÷ 300 = 15% |
| "Buyer agreed written success criteria" (buyer evidence) | 120 | 36 | 36 ÷ 120 = 30% |
Under the old definition, the stage held 2.5 times as many deals, and each one carried half the probability, because real and hopeful deals were averaged together.
Activity-based criteria cause three problems:
- Stage movement measures effort, not buying progress. The seller can move every deal forward without the buyer doing anything.
- They're easy to game at quarter end. Send a batch of proposals, and the pipeline looks further along.
- Each stage mixes real and hopeful deals, so its probability becomes an average of both and predicts neither well.
After a team rewrites a stage with buyer-verified criteria, fewer deals reach it, and those that do should win more often. Mark the date of the change, don't mix old and new history, and recalculate the stage probability from the new definition once three or four quarterly cohorts are old enough to have outcomes.
What counts as buyer evidence
Four types of evidence work well:
- Words in writing: an email or message from the buyer confirming something, such as a reply to a meeting summary that asks "Did we get this right?"
- Actions: the buyer introduced the economic buyer, involved IT, or shared data.
- Artifacts: their requirements document, vendor form or contract redlines.
- Dated commitments: a meeting on their calendar, or an agreed next step with a date.
"They loved the demo" and "they said it sounds good" are not evidence. The test for every stage move is one question: how do you know? What did the buyer say or do, and where is it recorded?
Writing good exit criteria
- Keep each stage to two or three criteria. More than that and nobody checks them.
- Make each one observable, and give it a home in the CRM: a checkbox with a date and a link to the email or document.
- Remember that one stage's exit criteria are the next stage's entry criteria. Write them once.
- If your buyers can't confirm something in writing (some public-sector rules prevent it, for example), choose a different kind of evidence, such as a dated scope review meeting. Don't drop the requirement for evidence altogether.
Common mistakes
- Using seller activities as criteria. "Demo delivered" says nothing about the buyer.
- Writing criteria nobody records. If the evidence isn't on the opportunity, the criterion can't be checked or audited.
- Keeping old probabilities after a redesign. The new stage is a different population of deals; recalibrate from its own history.
- Treating forecast category as a stage. The stage records what has happened; the forecast category is the rep's judgment about this period. Keep both. Honest stages are also what make forecast accuracy improvable.
The free RevOps Foundations course introduces exit criteria as part of the Process layer in Lesson 4, "The ROI Stack™".