The short version
A common CRM mistake: stages named after what the rep did (“Call 1 done,” “Demo sent”) instead of what the buyer has committed to. Activity-based stages make forecasting a guessing game, because “I sent a demo” tells you nothing about intent.
Behaviour-based stages force a rep to justify advancement with evidence, not effort: Qualified (confirmed pain and a budget signal), Evaluating (multiple stakeholders engaged), Committed (verbal agreement on next steps and timeline), Contracted. Rebuilding stage definitions this way, with explicit exit criteria written into the CRM, is usually the single highest-leverage half-day of CRM hygiene work a sales team can do.
Writing exit criteria that cannot be faked
A good exit criterion is something a manager could verify without asking the rep. “Buyer confirmed budget range” is soft. “Budget range stated in writing or on a recorded call, with the call link attached” is checkable. In HubSpot or Salesforce, make that evidence a required field or a validation rule on the stage change, so the deal cannot move without it.
Keep each criterion to one line and phrase it as a buyer action: “The buyer has named a second stakeholder and agreed to include them.” Avoid verbs that describe the rep, such as sent, presented or followed up. If a sentence still makes sense with the buyer deleted from it, it is an activity stage and should be rewritten.
Handling stalled and regressing deals
Behaviour-based stages make it legitimate to move a deal backwards. A deal in Evaluating whose champion goes silent for 21 days should drop to Qualified, or into a Stalled status, not sit in place flattering the forecast. Add a rule that any deal with no buyer-originated activity for a set number of days, commonly 14 to 30, is flagged for review.
Treat backward movement as information, not failure. Track it as its own report: which stage do deals regress from most often, and what was the stated reason? A cluster of regressions out of Committed usually points at a missing stakeholder or an unresolved legal or security step that your stage definition should have demanded earlier.
Rolling it out without a revolt
Run the migration in one sitting with the team in the room. Print a mapping table of old stage to new stage, and let reps challenge individual deals live. Expect an uncomfortable first hour, because several deals will visibly drop a stage. That honesty is the point, and it is far cheaper to absorb in a workshop than at quarter end.
Freeze the old reports the day you migrate and rebuild the three that matter: pipeline by stage, stage-to-stage conversion, and average days in stage. Do not try to compare new conversion rates with historical ones; the definitions changed. Start a fresh baseline and annotate the dashboard with the migration date so nobody misreads the dip.
Worked example (illustrative)
A hypothetical team of six reps carries 80 open deals under activity-based stages. After a half-day rebuild, each deal is re-staged against buyer-evidence criteria.
What changed: The pipeline looks smaller on day one, but the forecast error shrinks because every late-stage deal now carries evidence. The drop came from removing deals that had only ever had rep activity.
Do it: the steps
Tick them off as you go. Your progress stays on this device.
Traps to avoid
- ×
Keeping both old and new stage names “for continuity”. Two vocabularies guarantee neither gets used properly.
- ×
Letting managers override stage by hand in the pipeline review. Once one deal is moved without evidence, every rep learns the criteria are optional.
- ×
Writing criteria that depend on opinion, such as “strong interest” or “good fit”. Replace them with something dated, named or attached.
What to track
- Stage-to-stage conversion rate
- deals that advance divided by deals that entered the stage in the period; expect it to settle after about two quarters
- Average days in stage
- mean time from stage entry to exit, tracked per stage; spikes show where deals stall
- Forecast accuracy
- absolute gap between forecast at quarter start and closed-won, as a percentage; trending down is the goal
Benchmarks are typical planning ranges, not guarantees. Calibrate to your own data.
Questions people ask
How many stages should a pipeline have?
Most teams do well with four to six. If reps cannot say from memory what separates adjacent stages, you have too many. Merge stages until each one has a distinct buyer commitment as its entry condition. Add stages only when a measurable difference in win rate or deal velocity justifies the extra data entry.
Should win probability be set per stage?
Only after you have at least two quarters of honest stage data. Until then, use your own measured stage-to-close rates rather than vendor defaults, and review them quarterly. Default percentages such as 10, 25 and 50 are placeholders, not evidence, and they distort the weighted forecast when stages are loosely defined.