Pipeline stages should be
verifiable by the buyer.

Open your CRM and read the stage names. If they are Qualified, Discovery, Proposal, Negotiation, you have a pipeline that describes what your sales team is doing. You want one that describes what the buyer has done. Those are different objects and only one of them forecasts.

The distinction sounds academic until the quarter closes at 61% of forecast and nobody can point to the week it went wrong.

For every stage, ask: could I prove this deal belongs here by showing a third party something the buyer produced? An email they sent. A calendar invite they accepted. A document they returned. A person they introduced.

If the only evidence is a rep's note, the stage fails. Notes are testimony. You want exhibits.

Five stages. Fewer than most teams run, because every stage you add is another place for a deal to sit looking healthy.

Notice that "sent a proposal" is not a stage. Sending a proposal is something you did. It says nothing about whether anyone will read it.

Once stages mean something, three reports that were previously decoration become useful. Stage conversion tells you where buyers actually stop rather than where reps stop updating. Time-in-stage becomes a real ageing signal instead of a measure of CRM hygiene. And stage-weighted forecast starts landing within a tolerable band, because the weights are now attached to observable events rather than to optimism.

In practice this is the moment a founder can walk into a board meeting and defend a number out loud. That is the whole point of the exercise.

Very early, with fewer than about thirty closed deals, you do not yet know which buyer actions predict a win. Do not import someone else's stage set. Track raw events for a quarter (every meaningful thing a buyer did in every deal you won and lost), then build stages from what actually correlated. A borrowed process applied to an unknown motion is just a more confident way of being wrong.