How to write sales stage exit criteria, with examples

If your forecast is unreliable, the cause is almost never discipline. It is that your stage definitions describe seller activity, and seller activity is entirely under the seller's control. I have written about why that produces a number nobody can trust. This is the other half: how to actually write the replacements.

What an exit criterion is

An exit criterion is one fact that must be true before a deal moves to the next stage. Not a checklist, not a confidence percentage, not a judgment call. One fact.

It has to pass a single test.

That test does most of the work. "Discovery completed" fails it, because completed according to whom. "Champion identified" fails it, because a champion is an opinion until they do something. "Buyer confirmed a business consequence of not solving this, recorded in the notes" passes, because either the note exists or it does not.

The second rule follows from the first: the evidence has to come from the buyer. A rep can send a proposal on any Tuesday they choose. A rep cannot make a second stakeholder join a call. Put the gate on the side of the table you do not control and the pipeline gets a brake it has never had.

Five stages, before and after

Adapt the names to whatever your CRM already uses. The names are not the point.

Qualified

Usually means: the rep ran a discovery call.

Should require: the buyer has named a business consequence of not solving the problem, in their own words, and it is written down.

If you cannot state what happens to their business if they do nothing, you have not qualified anything. You have had a pleasant conversation. The test I use in pipeline reviews is to ask the rep to say the consequence out loud. If it comes back as a version of "things stay the same," there is no deal here. Not a slow deal. No deal.

Evaluating

Usually means: the rep gave a demo.

Should require: more than one person on the buying side has engaged, and the rep can name who else has to say yes.

Single-threaded deals fail at a dramatically higher rate than multi-threaded ones, and the failure is invisible until it is total. Your champion goes quiet in week six and there is nobody else who knows the deal exists. Naming the buying group is the stage. Everything else at this point is preamble.

Validated

Usually means: the rep answered the technical questions.

Should require: the buyer has confirmed the solution works for their environment, in writing or on a recorded call.

This is the stage most companies skip, and it is where deals die quietly between a good demo and a proposal that never gets a reply. Somebody technical has an unvoiced objection and nobody surfaced it. Requiring confirmation forces the conversation to happen while you can still do something about it.

Committed

Usually means: the rep sent a proposal.

Should require: the buyer has told you their process, their timeline, and who signs, and owns a next step with a date attached.

The word that matters is owns. If the rep owns every next step in the deal, the buyer is not participating. They are being pursued. Those are different situations and only one of them closes.

Contracting

Usually means: "we're in negotiation."

Should require: paper is with the buyer's legal or procurement, and you know the name of the person holding it.

"With legal" is where forecasts go to die. Legal is a person with a queue, not a department that receives things. If nobody on your side knows that person's name, the deal is somewhere in a building rather than in contracting.

The three cases that always break it

Every time I install this, the same three objections arrive within a week. They are all reasonable and all solvable.

Renewals and expansion

An existing customer has already named the business consequence, years ago. Forcing them back through Qualified is theatre.

Use a separate pipeline with its own criteria. For expansion, the useful gate is whether the buyer has confirmed a new problem or a new business unit, rather than a general willingness to keep paying. Renewal is a date. Expansion is a deal. Do not run them in the same pipeline and then wonder why your win rate looks strange.

Channel and partner-sourced deals

Your partner has the buyer relationship and you may never speak to the end customer directly. The criteria still apply, they just get verified through the partner.

The addition worth making: at Evaluating, require that you know whether the partner's rep has actually met the decision maker, or is relying on a relationship two levels down. Partner-sourced pipeline is systematically over-forecast because nobody applies the same scrutiny they would to a direct deal.

Long enterprise cycles

An eighteen-month deal will sit in Evaluating for two quarters and look stalled against criteria designed for a ninety-day cycle.

Do not soften the criteria. Add an aging rule instead: any deal in a stage beyond twice its historical median gets flagged for review, not moved. The criteria tell you where a deal is. Aging tells you whether it is moving. You need both, and conflating them is how companies end up softening definitions until they mean nothing.

What happens when you apply it

A large part of your pipeline moves backward. At one company it was slightly more than half.

That is not a disaster and it is not a performance problem. It is the first accurate picture anyone there had seen. The pipeline did not change size. The measurement changed.

This is the moment the whole exercise usually fails, so it is worth being deliberate about it. Say out loud, before you start, that the number is going to look worse and that this is the point. Then hold the line for a full quarter. If you treat the drop as a performance issue, you will have taught your team exactly one lesson, which is that honesty is punished, and the numbers will drift back to comfortable within a quarter.

How to make it stick

Willpower does not enforce this. The CRM does.

• Put the criterion in the field description so it is visible at the moment someone changes a stage, not buried in a document nobody opens.
• Add validation on the fields the criteria depend on. If Evaluating requires a named second contact, make the stage change fail without one. A rule that can be skipped will be skipped in the last week of a quarter.
• Report on criteria compliance for the first quarter, then stop. It is a habit-forming measure, not a permanent one.
• Run the pipeline review on three questions only. Who else has to say yes and have you met them. What happens to their business if they do nothing. What is the next step, what is the date, and who owns it.

Most reps will fail at least one of those three on most deals for the first few weeks. That failure rate is the actual state of your pipeline, and it is worth more than any weighted probability calculation you could run on top of the old data.

Where to start

Do not rewrite five stages this week. Take one, write its exit criterion as a fact the buyer produced, and apply it to every open deal in that stage. Let them fall where they fall.

You will learn more from that one stage than from a quarter of inspection, and it costs an afternoon.