Unseen Job Postings
A surface for postings users had never seen — end-to-end conversion doubled in its first cycle.
Exposure → View
Company
Remember
Timing
H1 2023 (initial cycle)
Role
Senior Product Designer
Team
PM, Design, Data, Engineering
Contribution
Owned design end-to-end
Improved stage
Exposure
→
View
→
Apply
→
Hire
Overview
The goal was more applications, the stage that feeds hires. Pushing on it exposed a gap: we didn't actually know how our users found postings and applied. So instead of guessing, we ran user interviews.
Problem
We segmented users on two axes (new vs. existing, active vs. passive) and focused on active existing users: the group the product could move fastest, with the quickest feedback loop. Interviews showed a pattern. Frequent visitors entered the main feed and scrolled to hunt for postings added since their last visit. The feed was making them do memory work. Exposure had stopped meaning discovery.
Hypothesis
If repeated exposure of already-seen postings was the bottleneck, a dedicated surface for unseen postings should lift exposure-to-view first, and the gain should carry down the funnel instead of evaporating at the next stage.
My Role
User interviews · segmentation
Surface
Entry point
Interaction
Stage-by-stage funnel metrics
Stakeholder alignment
Design Decision
I designed an emphasized banner at the bottom of the main screen on entry, showing how many unseen postings remained: a dedicated space where users see only those. It sits visually separated from the feed, and tucks away when the user shifts to other behavior.
Considered and rejected: a newest-postings sort. For active users this check was the first job of every visit, and a sorted list still gave no marker of where they had left off. It needed a dedicated device, not a sort option.
On entry, the banner counts the postings you haven't seen; scroll into other content and it tucks away.
"Unseen" meant postings the account had never opened, measured by clicks. Postings a user had scrolled past unclicked could still qualify; recency-weighted relevance scoring kept long-circulating postings from resurfacing, so the space stayed fresh rather than recycled.
Result
2x
end-to-end conversion (0.06% → 0.12%)
first cycle, H1 2023: every stage improved
Funnel Conversion
Before
After
Exposure → View
6% → 10%
View → Apply
36% → 41%
Apply → Hire
2.7% → 3.0%
Measured
interviews with 10 users (recruiting at least five per segment), then an A/B test at 50:50 over two weeks with active job seekers, tracking whether exposed users checked the banner first, returned to explore, and used it repeatedly. No ranking-logic changes shipped in the window.
The chain was measured through to hires (the point where Remember earns), and every stage moved together. So a doubling this early compounds through every stage below it.
Later Validation
H1 2023
First cycle: funnel doubled end-to-end
Q4 2024
Relaunch: 2,345 direct applications
2025
~2.5x application conversion of a comparable module
Different periods and conditions, not directly comparable to the 2023 cycle. Same mechanism, two years later.
Trade-offs & Guardrails
A stakeholder worried the banner would cut time-in-app. I made two arguments, and we agreed to watch downstream metrics as guardrails; they held. Posting quality was settled by the funnel itself: view-to-apply rose from 36% to 41% instead of falling.
1
Users run multiple job platforms in parallel, so dwell time spent hunting serves no one's goal.
2
Remember is an MAU product people return to when job-seeking, not a DAU community where longer sessions mean health.
What This Shows
It looks like a very simple feature, but it made the user's goal easier to reach. The placement and the structure came from looking past the product itself, at how people actually behave when they are trying to change jobs. For working professionals too busy to comb through listings, it became a feature they genuinely needed.
