How to Clean Up a Sales Pipeline in One Week

Pipeline cleanup usually gets scheduled as a project and executed as a negotiation. RevOps produces a list of dead deals, reps defend each one, and half the list survives on the argument that the buyer is going to come back. Three months later the same deals appear on the same list.

The way through is to run the scrub in passes with defined rules, and to reserve rep judgment for the small set of deals the rules cannot resolve. Five passes, one week.

What does a dirty pipeline actually cost you?

Coverage is the obvious cost. ORM's customers typically carry more than 10 percent of pipeline untouched for twelve months, so a reported 3.8x is often a real 3.4x or worse.

The subtler cost is conversion math. Stale opportunities sit in the denominator of every stage conversion rate you calculate, which drags reported conversion down and makes healthy stages look broken. Cycle time inflates the same way, because a deal that died in month two but stayed open for a year enters your average as a fourteen-month cycle.

Deal values compound it. A pipeline averaging $80,000 per opportunity against closed-won deals averaging $40,000 will report roughly twice the revenue it can deliver, regardless of how clean the record count is.

Which deals come out first?

ORM applies a twelve-month rule for most customers and defines meaningful activity narrowly. A change in stage, close date, or amount counts. A logged call does not. A note does not. An email sequence does not. The record itself has to have moved.

Run this pass as a bulk action, not a conversation. The deals it catches have been dead long enough that no rep is actively working them, and asking for a defense on each one converts a two-hour job into a two-week one.

What do you do about close dates in the past?

A close date in the past is a record that has been abandoned rather than worked. Rolling those dates forward automatically is worse than leaving them, because it manufactures in-quarter pipeline out of dead opportunities and pushes the forecast further from reality.

Close-date behavior is worth watching after the scrub as well. The strongest signal that a deal is in trouble is a rep changing the close date. A deal that slips from one quarter into the next becomes less likely to close, even when it stays in commit. Track the count of close-date changes per deal as a standing field. Three or more is a deal slippage pattern rather than a scheduling detail.

How do you decide what is genuinely slow versus dead?

ORM groups opportunities with a machine learning model and predicts a close curve for each group. Those curves run from 1 to 80 weeks. Most groups carry their expectation before week 12, and very few extend past week 52. A 30-week-old enterprise deal in a group that peaks at week 26 is slow. A 30-week-old deal in a group that peaks at week 8 is finished, whatever the rep says about a budget cycle.

Without grouped curves, use segment medians. Take the median weeks-to-close for won deals in each segment, multiply by two, and treat anything beyond that as requiring an explicit defense.

How do you handle deals reps refuse to close out?

Most resistance is reasonable in intent. The rep does believe the account will buy eventually. The problem is that "eventually" and "this pipeline" are different things, and coverage math cannot tell them apart.

A nurture or recycled status solves it. The account stays visible, marketing can work it, and the opportunity leaves the coverage denominator. When the buyer re-engages, a new opportunity gets created with a real date. That also keeps your creation cohorts honest, because the re-engaged deal is counted as new pipeline in the period it actually became live.

Do not delete records. Closed-lost history is the input to win rate, cycle length, and every model built on them. Deleting a year of dead deals raises your reported win rate without changing a thing about the business.

How do you reprice inflated deals?

Compute the ratio of average closed-won value to average open value per segment over the last four quarters. If open pipeline averages $80,000 and closed-won averages $40,000, the ratio is 0.5. Deals sitting at more than 1.5 times the segment closed-won average go into pass 4 for a rep decision: reprice, or document the specific scope that justifies the number.

Report both the CRM value and the adjusted value from that point forward. The gap between the two columns is a standing measure of pipeline inflation, and it tends to close on its own once it is visible each week.

How do you keep it clean afterward?

Four standing rules cover most of it:

1. Flag any open deal with no stage, close date, or amount change in 60 days. 2. Flag any close date that passes without a resolution within five business days. 3. Flag any deal that exceeds its group's expected close window. 4. Flag any deal whose value exceeds 1.5 times the segment closed-won average.

Managers clear the queue during the weekly review. A pipeline that runs on these flags stays close to accurate permanently, which is the condition good forecasting practice assumes and rarely gets.

Does perfect data matter as much as teams think?

Every revenue team believes their data is uniquely bad and that it is the reason they cannot forecast. It usually is not. Garbage in does not have to mean garbage out. As long as the errors are consistent, a model can learn the bias and correct for it.

What breaks prediction is inconsistency. Reps who define stage 3 differently from each other, close dates that mean "when I hope" on one team and "when procurement signs" on another, amounts that include services on some deals and not others. Cleaning up a pipeline is mostly about making the errors uniform, and that is a smaller job than making them disappear.