Backflip · Real estate lending

Cutting the right thing

Analyzer was expensive and slated for removal. The bill was real; the attribution was wrong. Nearly all of it came from an automated path to the same analysis. Cutting that path removed 90% of the cost and left the strongest retention driver in the product intact.

• Analyzer users retained 60% vs 9% at Month 1 and filed 85% more applications
• Kept the product, cut the one feature driving the bill: 90% cost reduction
• Renegotiated the vendor to usage-based pricing so the savings stayed real

Role

Senior Product Manager · Q1 2026

Scope

User research, cost attribution, executive persuasion, vendor renegotiation, scope reduction

Worked with

Design, engineering, sales, finance, legal

Timeline

About two months from the first cut-it conversation to the decision; the vendor renegotiation waited on the contract end date a few months later

The problem

Analyzer, the deal analysis tool real estate investors used to evaluate properties, was free, and it was how a lot of our borrowers found us in the first place. Sales already led with it. It also carried a large third-party data bill, and in a cost review that is all it read as: high spend, no direct revenue line attached to it.

The case for cutting it was entirely about the cost side of the ledger. Nobody had put a number on what it was worth.

What I did

I started with users rather than the spreadsheet. In borrower interviews, investors told me flatly they wouldn’t have used the platform without Analyzer. Then I checked whether the data agreed. It did, and the gap wasn’t marginal:

• Month 1 retention: 60% vs 9% — 6.7×
• 85% more loan applications per user

Some of that gap is selection: engaged investors are the ones who analyze deals. The interviews are what convinced me the causation ran the other way too, and the inbox-scanning cohort, which had the tool without the intent, retained no better than anyone else.

That reframed the decision. The question was no longer “is this feature worth its bill” but “are we willing to remove the strongest retention mechanism we have to save a line item.” Framed that way, the room agreed.

Winning that argument didn’t make the bill go away, though, and I didn’t want to defend the product again in the next cost review. So I went after the spend itself rather than treating the vendor bill as one undifferentiated number. Nearly all of it traced to a single feature, marketed and used as its own thing but really an automated front door to the same analysis. Investors connected their email inbox; we scanned it with AI to spot properties that looked like real deals, usually ones a wholesaler had sent them, and pulled those in on their own. It read and pulled data at high volume, and the investors using it were no more active and closed no more loans than anyone else.

Cutting it removed roughly 75% of data volume without removing the capability. Investors could still analyze any property they wanted; they just had to bring it to us rather than have us find it in their inbox. With volume down, I renegotiated the contract to usage-based pricing so the remaining spend tracked actual demand.

The spend was real. It just wasn’t coming from the thing we were about to cut.

Tradeoffs

Investors who had connected their inbox lost automatic deal discovery. A wholesaler email that used to surface a property on its own now needed them to notice it and run the analysis themselves. That is a real regression, and the usage data is what made it acceptable: those investors weren’t retaining or converting any better than investors who brought us deals directly, so the automation was buying data volume rather than outcomes. Keeping it would have preserved the entire cut argument along with it.

Usage-based pricing also trades a predictable bill for a variable one. That’s the right trade when volume has just fallen 75% and you want the savings to be real rather than averaged away in a flat rate.

What I’d do differently

Kill inbox scanning earlier, on its own merits. The usage data showing it bought volume rather than outcomes existed well before the cost review. It took a threat to the whole product to make anyone act on it, and I was one of the people who could have acted sooner.