More invitations did not automatically mean better connections.
The useful outcome was not the tap on an invitation. It was a conversation that someone wanted to join and continue. We looked at activation, acceptance, conversation starts, reciprocal exchanges and premium conversion to understand where the experience stopped working.
The product question shifted from “How do we generate more interactions?” to “What makes a first interaction worth answering?”
Context made the first move easier to understand.
Interactions connected to a photo, an icebreaker or a profile detail gave the recipient a reason to engage. Generic invitations carried less information. I combined behavioural analysis with user research to turn that pattern into concrete changes.
A richer starting point
Use onboarding to help people build profiles with something to respond to.
A more specific first move
Attach a like to something on the profile and improve the icebreakers.
Make existing compatibility signals useful to people.
The recommendation engine already calculated compatibility, but that signal was not visible to users. We tested a recommendation experience organised around compatibility rather than chronological discovery. The experiment informed Super Match: fewer profiles, with stronger conversation potential.
From the question in the dashboard to the details in production.
I owned behavioural analysis, product discovery, research, experiment definition, interaction design, UX writing and delivery with engineering and QA. Working directly with the CEO kept product priorities close to both user behaviour and business constraints.
Measure the connection, not just the click.
Across the Mektoube initiatives reported in my CV, post-onboarding retention increased by 34% and invitation-to-conversation conversion by 12%. These are programme-level results across multiple changes, not the isolated effect of one screen.
Behavioural data tells me where to look. The design work is understanding what people need at that moment.
Evidence & scope
Figures reported in my CV. Measurement windows, baselines and attribution detail are not published here.