Deal Stage Regression

Deal stage regression is an opportunity moving backward to an earlier pipeline stage after it had already advanced. A deal in negotiation returns to evaluation because a new security reviewer joined and reopened the technical assessment. The record now shows something most pipelines never capture: a deal that lost ground it had already covered.

What a regression actually tells you

Forward movement is ambiguous, because a rep can produce it with a picklist edit. Backward movement is almost never volunteered, so when it appears, something concrete happened. The common causes are a new stakeholder entering late, a scope change that resets the commercial terms, a champion leaving, or a budget owner reopening a decision that had been treated as made.

Each of those materially changes the close date. That is the reason regression is worth tracking as its own event rather than as an ordinary stage change.

Why teams suppress it

Regressions look like failure on a rep's board, so reps avoid recording them. The deal stays parked in the late stage instead, holding its original amount and its original close date, and it keeps the weight that stage implies right up until it disappears from the forecast.

That is the worse outcome. A regressed deal that is honestly recorded is a manageable problem with weeks of warning. A deal frozen in negotiation for two months is a surprise at quarter-end. Managers get the behavior they reward, so the regression rate should be read as a hygiene metric, not as a rep scorecard line.

Measure it at the boundary

Compute regression rate per stage boundary rather than per pipeline. A concentration of regressions out of one stage points at that stage's stage exit criteria being too easy to satisfy, since deals are entering it before they have earned the position and getting pushed back once someone checks.

Two other cuts are worth running. By rep, to separate a coaching problem from a process problem. By week of quarter, because regressions clustered right after quarter close usually mean the previous quarter's late-stage bucket was inflated.

Pair regression with the close-date signal

ORM identifies a rep changing the close date as the strongest available signal of deal slippage, and the earliest signal as the absence of any signal at all: no activity, no field changes, no notes. Regression sits between those two. It is louder than silence, and when it lands before a close-date edit it is the first hard event you can act on.

Build the alert as a pair. A stage regression with no accompanying close-date change is the deal to inspect first, because the record is asserting that a deal with new work ahead of it will still land on the original date. Treating that contradiction as a review trigger removes a recurring source of error from forecast accuracy work, and it does it early enough in the quarter to change the outcome.