Apartment List~3 min read

Renters favorited a property and never got in touch.

Two weeks, one PM, one engineer, four experiments, and three that failed.

  • Role: Product design and research
  • Timeline: 2 weeks
  • Team: 1 PM, 1 engineer, me
  • Impact: $50k → $500k placement revenue
  • Platform: Web
  • Platform: iOS, Android
0%

Renters contacting properties

after spotlights

$0k

Placement revenue

up from $50k, run-rate

Both were read at three weeks, ahead of the three months Apartment List normally took to report, and held through my departure in 2024. Both come from spotlights alone, not the program. Revenue is what property management companies pay for placement, not lease revenue attributed to better matching. The 80% may not share a denominator with the 50% this page opens on. Properties replying to renters did not move, which is the half of the exchange this work did not touch.

Overview

Half the renters who favorited a property never contacted it, and the experiment that fixed that was not the one the research pointed to.

50% of renters who favorited a property made no contact within 90 days. The intent was there and the next step never happened.

Worked with: 1 PM, 1 engineer, and Sales; 8 renters interviewed.

What I owned

I owned

  • Research with 8 renters, which showed trust working as a shortcut against choice fatigue
  • A competitive review of Airbnb and Amazon, and the placement principle it suggested
  • All four experiments and the order that determined what ran in two weeks
  • Leading and lagging measures, set before testing began
  • The lockup redesign, tested at roughly 8,000 sessions per arm

Decided with the PM and engineer

  • Test design and holdouts for the two experiments with volume risk
  • Pricing and packaging spotlights as sellable placement: Sales and the PM owned that, not me

Research

Renters used the management company as a shortcut against choice fatigue.

I interviewed 8 renters mid-search. The product knew which company owned each listing and never showed it. Product data added that 78% of searches had a cross-sell opportunity. A competitive review of Airbnb and Amazon showed both put credibility next to the decision.

Decision: test order

I ran experiments in descending expected value, which put the smallest result first.

We had two weeks and one engineer. The chatbot scored highest on impact and lowest on everything else, so it went last. I started with the lockup redesign because it had high confidence and was easy to build. Ordering by confidence puts the most trusted experiments first, which may teach the least. I made no mitigation for that.

Decision: placement

Paid placement worked where a free trust signal could not.

I rejected a free trust badge, because adding a logo to a company that responds badly means the platform vouched for it. I also rejected ranking editorially by response quality. We sold promoted placement, a slot above the organic feed. The cost is that paid prominence is not earned credibility, and renters who notice may trust the surface less. Holdouts measured it against a control.

The Deal Spotlight panel on a renter's matches page. Two promoted units from the same property sit in a bordered slot beside the organic smart-matches card, each showing a struck-through rent, a discounted price, and a specials-applied tag.
The one that worked. A paid slot beside the organic matches, and every headline figure on this page comes from it.

Redesign

Three other experiments produced a small lift, a withdrawn feature, and an unresolved problem.

The lockup put the company’s logo and contact details on the listing card. It was tested at roughly 8,000 sessions per arm over two weeks, 50/50 split, at 95% confidence. It won, and it was the smallest reason this worked. Bulk messaging let renters contact several properties at once. They did not use it, so we removed it.

The listing before the change. Address, unit details, a property website link, and a remote tour link, then a managed-by row naming the management company in grey text with no logo and no way to contact it.
Before. The company managing the building is named once, in the lightest grey on the page, with nothing to act on.
The redesigned lockup. A contact block heads the card with a phone number marked SMS supported, a 3D tour and video link, and a property website link, above the management company's logo and a row of three of its other properties under a see-more link.
After. A logo, a number a renter can call, and the company's other buildings underneath. It won its test, and it is the smallest reason this program worked.

Scope

Spotlights shipped. Bulk messaging was removed and the chatbot stayed live.

Renters engage with the chatbot and then do not contact the property. It passed every measure I set beforehand.

The leasing chatbot on a phone. A header reads welcome to Camden Station, I'm Lea, your virtual leasing assistant, over a swipeable photo gallery, a prompt asking whether the renter would like to see photos of anything else, and reply chips for community and amenities or floorplans.
It answers the question. That is the problem: a conversation good enough to end the question is a conversation that ends before anyone contacts the property.

Results

Renters contacting properties rose from half to 80%, and placement revenue from $50k to $500k.

The chatbot passes every measure I set, and renters still do not contact the property

I set measures about engagement, and a conversation that answers the question well enough to end it works against this funnel. Nobody established whether it helped renters at the funnel’s expense. It was still running when I left. I spent effort making the signal credible when it needed to be unavoidable.

Sam Cusano