Facebook Marketplace Home Listing Composer

2016-2019 · Product, interaction, and experiment design

I made it simpler for sellers on Facebook to create high quality home listings, making it a trusted destination for home shoppers. I accomplished this by revamping the home listing composer to allow for comprehensive property information inputs. I streamlined data entry with quick toggles, and added advanced field search for more detailed cases.

Home listing composer detects structured data from free text descriptions

Context

For the initial launch of a dedicated home shopping experience on Facebook Marketplace, we prioritized simplicity by using the basic listing form with minimal required fields. This approach aimed to capture early user interest and intent, rather than offering advanced tools for professional sellers.

Initial, minimal home listing composer

Problem

Although we exceeded our targets for early adoption and home listing inventory, we observed a notable difference in the quality of direct home listings (via the Facebook app) compared to those from partners. This difference was evident in the number of photos and the detailed property information provided. We recognized that real estate transactions necessitate richer content, and our research indicated that home shoppers preferred structured details for better matching.

Low quality listings resulted in several key business metric drops:

Listing interactions went down (topline Marketplace metric)
Lead quality (seller side topline need)
Inventory partnership developed a negative brand perception (they don't want their high quality listings alongside lower quality).

Quality delta between direct and partner home listings

Solution

I collaborated with product managers, engineers, researchers, and analysts to enhance home listings on Facebook by adding more input fields without overwhelming sellers. This process involved experiments and user feedback loops, helping us find a solid direction.

To gather initial intent, I designed a new optional section of free text response categories beyond the required details to learn what aspects about a home sellers were most interested in adding.

Experiment: Anatomy of new category fields to gauge interest from sellers

The responses for each new field were analyzed, resulting in a comprehensive list of fields and logical groupings. Due to the substantial volume of new fields, it was necessary to reconsider how to present these form fields to avoid overwhelming sellers with too many choices.

Shipped: Redesigned home listing form with new sections & quick inputs

Experiment: Anatomy of new category fields to gauge interest from sellers

Early response

Despite sellers wanting to use the new structured data fields in the updated home listing form, adoption was low. Research revealed that people list homes on multiple websites and prefer minimal entry, using free text for the rest. Only a few professional sellers manually fill out the fields. This required us to think differently about how to raise direct listing quality without forcing a behavior.

Experiment: Prompt sellers to make details from free text entry both structured & searchable

Further experimentation

We tried detecting keywords in descriptions to auto-populate structured fields and ask sellers to opt-in, but it failed. It was seen as a common upsell tactic, and we asked at the wrong time—just before publishing the listing.

Winner: Prompting sellers in context

Winner

I hypothesized that timing was key. I redesigned the experiment to prompt users after they finished entering the text description. This proved effective, resulting in a high opt-in rate and no critical feedback.

Impact

50%+ increase in quality of direct listings in comparison to partner listings, defined by # of structured fields.