A pipeline stage should be defined by something the buyer did, not something you did. Get that right and the probabilities are no longer a judgement call, because you can count them out of your own history.
Most pipelines are inaccurate for one reason: the stages describe seller effort, so they advance whether or not the buyer is moving.
A default set of six
| Stage | Entered when the buyer | Typical time in stage |
|---|---|---|
| New | Responded, or arrived through a form or referral | Days |
| Qualified | Confirmed the problem, the budget and who decides | Days to a week |
| Discovery complete | Shared enough detail for a specific proposal | One to two weeks |
| Proposal sent | Received a priced proposal they asked for | One to three weeks |
| Negotiation | Raised terms, redlines or procurement | One to four weeks |
| Closed | Signed, or told you no | n/a |
Each definition in the middle column is an event somebody else caused. That is the test. "Demo booked" is a buyer action. "Demo prepared" is not. "Proposal sent" is weak on its own and strong when it reads "received a proposal they asked for", because the asking is the signal.
Where the probabilities come from
Not from the CRM defaults. Those numbers are a guess about a generic business, and they are almost always too generous in the middle stages, which is why forecasts built on them read high all year and correct in December.
The calculation is ordinary. Take every deal from the last twelve months that ever reached a given stage, including the ones that died there. Count how many eventually closed. That fraction is your probability for the stage.
Two things usually surprise teams the first time they do it. The proposal stage converts worse than anyone expects, often badly enough that sending more proposals is the wrong response. And the qualified stage converts far better than the default suggests, which means qualification is doing more work than it is given credit for.
Redo the numbers every six months, or after any real change to pricing, market or team. A probability set that has not moved in two years is a decoration.
Why stage definitions drift
They drift because nobody owns them. A rep moves a deal to negotiation because the conversation felt like negotiating. Another waits for a redline. Both are being reasonable, and within a quarter the stage means two different things, which makes the forecast an average of two definitions.
The fix is a sentence per stage, written down, visible in the CRM at the moment of the change rather than in an onboarding document somebody read once. And one person who owns the definitions and is allowed to say no when a fourteenth stage is proposed.
Time in stage is the measure that earns its place
Conversion rate tells you what happened. Time in stage tells you what is happening, early enough to act.
A deal sitting in proposal for three times the normal duration is not at 50 per cent whatever the stage says. Ageing is the most reliable signal of a deal that has quietly died, and it is the one number most pipelines do not surface, because it requires knowing what normal is for each stage, which requires the history above.
Two rules make it usable. Every stage gets an expected duration. Anything past twice that duration is reviewed and either moved, dated or closed. A pipeline that is only ever added to will forecast badly no matter how good the stage definitions are.
Empiraa Signal is a CRM and more. It is also the B2B prospect database, the enrichment, the email sequences, the booking link and the proposal tool, in one system with unlimited users on every plan, so everyone who touches a deal can keep its stage honest without it costing another seat.
Related reading: how to build a sales pipeline from scratch, and lead qualification frameworks compared.
Sales pipeline has a definition on this site.
