Özge Duman Karaboğa

Trust on Facebook Marketplace is never stated

Buyers rely on messaging to decide whether to go ahead with a purchase. I looked at how unresponsive sellers, automated messages and trust concerns push them to give up.

  • 5 interviews
  • 2 contextual inquiries
  • 3 weeks
  • October 2025
Marketplace listings with a buyer and seller exchanging messages
"This is not Amazon. You talk to real people."

What the research revealed

Every participant came with the same goal: find a good deal, close by, quickly. What they actually spent their time doing was working out whether the seller could be trusted.

  • Trust emerges from small signals. Profile age, past listings, how fast someone replies and how they write. A warm message builds confidence. A cold one ends the conversation.
  • Response speed was the biggest factor. Most expected a reply within the day. After that they moved on.
  • Buyers work around the platform. They changed search terms and managed multiple conversations with no way to tell them apart, and accepted the friction as part of the experience.

How I ran it

  • Generative interviews. Five semi-structured sessions of 30 to 60 minutes, with open questions that let buyers share real stories: missed pickups, broken promises, scam attempts.
  • Contextual inquiry. I watched two buyers use Marketplace in real time on their own devices, which showed behaviors they would never have described themselves.
  • Affinity mapping. Observations were clustered in Miro into 3 themes and 12 sub-themes. All participants signed an informed consent form.

What could be different

  • Response time indicator. A badge next to the seller’s name, similar to delivery apps showing wait times.
  • Verified activity badge. Account age and completed transactions, replacing the manual checking every participant described.
  • Proactive suggestions. When a seller doesn’t reply in time, surface similar active listings nearby instead of silence.

What went well

  • Narrowing the scope to buyers allowed deeper, more focused analysis.
  • A casual opening conversation reduced social desirability bias.
  • Contextual inquiry revealed behaviors participants would never have self-reported.

What I’d do differently

  • Early themes overlapped heavily and needed several rounds of refinement.
  • Separating findings from design opportunities was harder than expected.
  • Some interviews ran too long. I should have managed time more strictly.