1 in 3 Leads Was a Duplicate. Nobody Knew.

The Setup

A real estate coaching company running a high-volume book-a-call funnel. Meta Ads, YouTube Ads, and a referral engine pushing hundreds of leads into Close CRM every day. On paper, the pipeline looked healthy — sixty thousand leads across the tracked window.

The Problem

The marketing team was celebrating a strong lead month. The setter team was exhausted and couldn't understand why their close rate was dropping. The CEO was looking at both reports and couldn't tell who was right.

The ad agency's attribution report said lead volume was up 28%. The sales team's close-rate dashboard said conversion was down 19%. Both were technically accurate. Nobody could square the two.

When the CEO asked the setter lead what was happening on the calls, the answer was: "We're calling the same people over and over. They're annoyed. Half the time they've already talked to us this month."

What We Found

We pulled the full lead dataset from Close CRM and started looking for duplicates. Not exact duplicates — those the CRM already caught. We looked for near-matches: same email domain with different casing, same phone in a different format, same name at a different address, same IP within a narrow window.

Of the 60,000 leads in the reporting window, 20,000 were duplicates. Thirty-three percent of the pipeline was phantom.

It got worse when we cross-referenced with Stripe. Roughly 8% of the leads had actually converted on an earlier ad click that the CRM didn't know about — they were being worked as new leads when they were already customers.

The close rate wasn't dropping. The real close rate had been under-reported for months. The apparent drop was the volume of phantom leads growing faster than the real ones.

What We Built

A three-layer deduplication engine that runs on every new lead ingestion. Email is normalized (lowercase, strip-plus addressing, check common provider aliases). Phone is normalized across international formats. Name variations are matched using phonetic similarity. When a match is found, the system doesn't drop the lead — it merges the history, preserves the original touch, and flags it for the setter so they know this person's full context before they dial.

The historical clean-up ran once. The ingestion-layer dedup runs continuously. A weekly report shows the CEO how many duplicates were caught that week, and what the true acquisition cost looks like on the deduplicated lead base.

The After

Effective close rate — measured on unique leads — was 38% higher than the number the sales team had been reporting. The setters stopped calling the same people. Dialer efficiency went up immediately because the list was smaller and cleaner.

Most importantly, the CEO made a budget decision he couldn't have made before: he cut one of the three ad agencies. Their volume looked competitive on paper, but 51% of their leads were duplicates of leads from the other two agencies — they were claiming attribution for prospects already in the funnel. We showed him the data. The decision took ten minutes.

Total project time: six weeks. Total client time: three hours, spread across weekly check-ins.

Closing

The CEO told us six months later that this engagement paid for itself the day we ran the first dedup report. The budget reallocation from firing the third agency alone saved more than the annual retainer.