When the pipeline is fiction, the fix is unglamorous

A founder I work with told me this week the friction in his pipeline was his own data.

Small B2B SaaS team, founder-led. His CRM has hundreds of contacts. He looked at it and realised most of those contacts hadn't been touched in 12 months, half of them were people he no longer recognised, and the lead-quality scoring sitting on top was effectively fabricated. The forecast, the next-best-action lists, the pipeline coverage report were all computed against contact records that no human had verified in over a year.

So this week, he stopped outreach. Paused the motion. Put delivery first while he and his team go through the data to figure out what's real.

I've watched this exact sequence three times this quarter. The founder doesn't notice the rot for months because the system keeps producing reports. Charts still draw. Forecasts still calculate. Then something forces a closer look, a board ask, a missed quarter, a renewal review, and the data underneath turns out to be a story.

The pattern is structural, not careless. Here's what causes it.

Why CRM data rots quietly

At 5 to 50 person founder-led companies, nobody owns CRM hygiene. The founder is selling. The first sales hire is selling. The implementation partner left after go-live. There's no RevOps function and no SDR layer cleaning records as they go. So the contacts pile up. Tags get applied inconsistently. People change roles or companies and the records lag.

The lead-scoring model keeps running, but it's running on data nobody is checking.

The thing that breaks the silence is usually a specific question the founder asks himself. Of these 47 deals in my Q2 forecast, how many have moved in the last 60 days? Or: of these 312 marketing-qualified contacts, how many would I personally call this week?

Then the audit starts.

What a real pipeline audit looks like

Three things, in this order.

First, define what "real" means before you sweep anything. A contact is real if a human at this company verified it in the last 90 days. A deal is real if it has moved at least one stage in the last 30 days, with a named next-step and a date. Anything that doesn't meet those two tests isn't in the pipeline. It's in a separate bucket called "unverified."

Second, sweep the records by hand. Founders try to do this in a single afternoon and burn out. The version that works is one focused hour a day for two weeks, with one person owning the call on each record. The team I'm watching this week scoped 14 working days for the cleanup.

Third, restart the motion against the verified set only. Outreach to anyone in the unverified bucket is paused until that record has been re-touched. The pipeline shrinks. The forecast contracts. Both of those things are the point. A small forecast you can trust is more valuable than a big forecast you can't.

The trap to avoid

Every founder I've watched do this is tempted to skip step one. They want to sweep the records first and figure out what counts as real along the way. That doesn't work. Without the bar set up front, the audit drifts into "is this contact still relevant" and stalls in week one.

Define the bar before you grade the work.

One thing worth doing this week

Pick a single sales report you currently rely on. Pipeline coverage, forecast, lead funnel, whichever one you actually look at. Take ten contacts or deals at random from that report. For each, write down two things. The date the record was last verified by a human. The date the deal last moved a stage. If half of the ten fail one of those tests, your report is a story, not data.

What good looks like

A founder who can point to a forecast and say: every deal in here has moved this month, every contact has been touched in the last quarter, and the names I see on this list are names I would personally call. The forecast might be smaller than last quarter's. It will also be the first one in a while that doesn't surprise you.