Joyn · Product Manager, Ad Inventory · 2020 to 2022
Charging Users for Showing Up
Leadership consensus at Joyn was clear: pre-rolls were essential for revenue. Three data points said otherwise. Here's how a contrarian read on the funnel led to a new ad model, cohort-level ad decisioning, and a direct disproof of the "we'll lose revenue" objection.
Watch-time growth post-launch
Negative impact on ad volume
Increase in retention
Context
Joyn was a free, ad-supported streaming platform where pre-rolls were the default revenue mechanism. Leadership consensus held that they were non-negotiable. The data disagreed.
Problem
As the PM owning ad inventory, I needed to decide whether to reinforce the pre-roll model or test an alternative. Three signals made the decision clear. 40% of first-time viewers abandoned their session on encountering a pre-roll, before experiencing a single second of content. 75% of heavy users went ad-free in most sessions because they had exhausted available inventory. And player crashes were re-triggering pre-rolls on session restart, punishing users for a failure that was ours. Pre-rolls were not just inefficient. They were destroying retention, under-monetising our most valuable users, and breaking trust at the worst possible moment.
Approach
01
Design the alternative
I designed a watch-time-based ad model: instead of a pre-roll, serve a maximum 15-second ad after 5 minutes of watched content. First impression of every session becomes content, not an ad. Crash-recovered sessions no longer re-trigger a pre-roll.
02
Introduce cohort-level ad decisioning
New to app users entered a low ad load experience, with frequency graduating upward as engagement deepened. For the first time, we were treating a first-time viewer and a power user differently, a core inefficiencies of the pre-roll model.
03
Secure executive alignment
I brought the three data points to C-level stakeholders across SevenOne Media, ProSiebenSat.1, and Joyn. The goal was to make the risk of changing feel smaller than the cost of staying still. The data did that work.
Outcomes
- →Watch-time grew 3% post-launch
- →Retention improved by 1.8%
- →Ad volume was unaffected — the anticipated revenue loss did not materialise
- →Users were no longer penalised for platform failures that had quietly eroded trust
Reflections
The hardest part was not the product design. It was earning the right to run the experiment. My job was to reframe the existing model as the gamble, not the change. The lesson: consensus is not the same as correct, and knowing the difference is the job.
Next case study
Building the world's largest live sports ad platform
Disney+ Hotstar