An end-to-end audit of how a property photo travels from a broker's phone to a buyer's screen, and a strategy to build quality in from the first upload.
An end-to-end audit of how a property photo travels from a broker's phone to a buyer's screen, and a strategy to build quality in from the first upload.
Role
Product Design Intern
Timeline
6 weeks
Worked with
Engineering, Data Quality, Product, Design
Scope
Research, systems audit, strategy, PRD, UI
Every listing on Housing.com starts with photos that a broker or owner takes on their phone. Before a buyer sees them, they pass through four systems: upload, verification, optimisation and rendering, each looked after by a different team. This project follows that whole journey, not any single screen.
In real estate, the photo is the first site visit.
Before a buyer requests a call, they scroll. A listing gets a few seconds of attention, and the photos do most of the talking in that time. Many Housing.com listings were losing that moment.
/Too dark to read the room/Blurred after compression/Cropped out completely
PDPProperty detail pageSRPSearch results page
#1
Photos rank as the most useful feature in property search, ahead of price, location and description.
NAR, 2024
20x
Listings with fewer than 9 photos get 20× fewer enquiries. Image isn't presentation. It's conversion.
Zillow
11.9M
Listings on Housing.com: the multiplier on every ungoverned decision behind the pipeline.
Internal data, Feb 2026
02 · The answer
Build quality in, starting at the first upload.
Stop bad photos at the source, protect good ones on their way through and show every photo the way it deserves to be seen: Prevent, Preserve, Present.
Prevent
The guided upload
Instead of finding out after review, brokers now get guidance right beside the upload box and feedback the moment a photo lands. A dark photo is flagged on the spot, along with how to fix it.
Step 2 The checklist, the quality meter, the live listing-card preview.
Before
After
Preserve
The variant rules
A clear rule for which variant serves which slot, so compression stops softening photos that started out sharp.
Recommended for a later phase
Present
The rendering rulebook
I wrote the rules for how every listing photo is framed: three standard ratios, crop rules that protect what buyers look for, and a cover photo that stays put from search results to the property page.
1Three ratios. No ad-hoc exceptions.
2Never crop the roofline. Keep the floor and ceiling in frame.
3Three ratios. No ad-hoc exceptions.
03 · How I got there
It wasn't a photo problem. It was a pipeline problem.
Picture a broker in an empty flat, carefully clicking every room. Before those photos reach a buyer, four different systems check, compress, resize and crop them, and any one of them can quietly undo that effort. So I stopped judging photos and followed one all the way through.
UploadOwner / broker, Design
VerificationDQ team
OptimisationEngineering
RenderingProduct / Design
4 stages, 4 user groups. No single owner of the end-to-end experience.
Following one image through four stages
01 · Upload
5 upload entry points
Five ways in, five different sets of rules.
One flow locked orientation and room order. Another had no rules at all.
02 · Verification
~5 min per listing
A strong audit, running too late.
03 · Optimisation
~15 compression variants
Variants, but no rules.
A sharp original could reach buyers as a blurry thumbnail.
04 · Rendering
~39 web · ~32 mobile ratios
39 ratios for one photo?
Up to five on the home page alone. The same photo looked different everywhere.
The insight
Quality was checked at the end, not built in at the start.
Each stage worked on its own. But because nothing guided a photo at upload, every stage after it was cleaning up. The audit rejected, reviewers checked by hand, and rendering cropped whatever arrived. The fix wasn't a stricter filter at the end. It was moving quality to the beginning.
04 · What I expect
What I expect this to change.
Hypotheses, not measured results. Each one names the signal I'd watch.
Prevent
Guided upload flow
Picked up first
I expectFewer poor photos reach review.
Guidance beside the upload box and instant flags mean brokers fix a bad photo while still in the room.
Signal
Rejection rate at audit
Review time per listing
Present
Standardised rendering ratios
Picked up first
I expectEvery photo looks the same everywhere, so more buyers enquire.
Three ratios and crop rules that protect the roofline, floor and ceiling, with a cover photo that stays put.
Signal
Enquiries per listing
Preserve
Variant rules
Later phase
I expectSharp originals stay sharp for buyers.
One rule for which variant serves which slot, so compression stops softening good photos.
