The challenge
An online real-estate listing portal selling 240,000 leads per month to 600 builders was bleeding developer subscriptions. Builders complained leads were stale, duplicated or low-intent. The portal had no way to score quality. Builders sat on leads for 9 hours on average, then blamed the portal when they did not convert.
How we deployed
- Trained an ML scoring model on 18 months of conversion data — output 0–100 quality score per lead.
- Built fuzzy-match duplicate detection across phone, email and project-of-interest.
- Enforced a 15-minute builder SLA — auto-reroute to next builder if first did not action.
- Issued weekly developer scorecards showing acceptance rate, time-to-action and conversion.
- Refunded low-quality leads automatically to keep complaint volume off support.
What changed
- Duplicate leads dropped 72% inside the first quarter.
- Builder time-to-first-action fell from 9 hours to under 15 minutes.
- Builder retention lifted 28% as scorecards proved lead quality objectively.
- Portal support tickets about lead quality dropped 81%.
- Premium subscription tier launched on the back of the new scoring engine.
"Our builders used to think we were sending them garbage. Now they see the scorecard, they see the SLA breach, they accept that the lead was good and they were slow."
— Product Head · Listing Portal