Pipeline architecture
Definition
The deal stages in your CRM, each with a meaning: what a deal in it has done, and what moves it out. Your forecast is read off this map.
How it works
Each stage needs an exit criterion that is an observable buyer action, something anyone can verify from the record: the buyer booked the meeting with the signing group, or returned the redlined contract. A stage that advances on the seller’s confidence advances on mood, and when the stages stop matching how the company sells, sales stops updating them; every forecast and report built on top becomes fiction. Every pipeline I have inherited had more stages than the sale had steps, added one manager at a time and never removed. Your stage count is the number of commitments a buyer makes between first conversation and signature; a two-call sale and a nine-month committee sale need different maps.
What to watch for
Ask a rep to define each stage from memory; a stage they cannot define is a stage nobody is updating. Look for stages holding almost no deals and for neighbouring stages whose definitions overlap. Then pull the age of open deals per stage and set it beside how long your won deals spent in the same stage; a stage holding deals far older than any deal that went on to close is where deals go to sit, and everything in it inflates your pipeline coverage. Check whether your stage probabilities are still the vendor defaults nobody chose.
The question you ask
“What does a buyer have to do for a deal to leave this stage?”