TED Shorts, powered by Genuin

Business Context

TED is the global ideas platform with a large library and a very large social following, but the app had been deprioritised for several years and had not seen a significant update in its fifteen-year history since launching in 2010. Its download momentum was slipping.

Per the TechCrunch coverage, the app was downloaded roughly one million times in 2024, down about seventeen percent from 1.2 million in 2023, against more than 100 million lifetime downloads, and the app hosts 26 podcasts across topics including business, climate, and health. TED's Head of Product is quoted in that coverage on how deprioritisation affected monthly active users and annual growth.

The strategic bet was to convert external social demand into an owned in-app short-form experience, on TED's own terms, at a moment when short-form video was booming and a major social platform faced a potential ban.

The Product Problem

TED needed a branded, high-quality vertical video experience on its own surfaces that fit an educational content model and did not leak users or data away from TED.

• Brand and journey ownership vs. platform defaults: TED wanted shares to resolve on its own domain and did not want a generic shorts destination.
• Educational consumption vs. standard feed behaviour: TED wanted viewers to absorb the full gist of a talk, which pushed against default swipe behaviour.
• No redundant onboarding: TED's app already captured user interests, so the standard interest selection step was duplicative.
• Quality control vs. release velocity: TED is an extremely precise customer, sensitive enough that even a trivial change could trigger a full QA cycle on their side.
• Content-first discovery: TED is a content brand, but default search surfaced profiles and groups above videos.

The Integration Surface

The integration covered a branded, white-labelled experience on TED's app and web, with share URLs resolving on TED's own domain; the swipeable AI-personalised vertical feed placed in the middle of the app's navigation bar; brand identity configuration in the Brand Control Center covering colours, nickname, and categories; analytics via a Mixpanel integration and TED-defined custom events such as video completion and funnel reporting delivered through SDK callbacks, plus a custom reporting dashboard; and a set of TED-specific SDK customisations across web, Android, and iOS.

The public TechCrunch coverage independently confirms TED Shorts sits in the middle of the navigation bar, which matches the placement decision documented in our internal integration sessions, so the public record and the internal record line up.

The Six Customisations

The numbering is meaningful here: these were sequential asks over the life of the account.

1. Share URLs resolve on TED's domain. Passing the same video ID or slug so TED can consume it and keep the user's journey inside TED's ecosystem rather than leaking to a generic platform link.
2. The interest selection step removed. TED's app already captures interests on first open, so I had the popup removed and consumed TED's own interest data for feed personalisation instead.
3. Auto swipe tuned for educational content. TED asked for this so viewers absorb the full gist of a talk.
4. Custom login helper text. Consumers did not understand why a login or sign-up prompt appeared when they tried to act.
5. A custom engagement dashboard, plus a Mixpanel integration. The dashboard compares two embeds on comments and watch time; TED defines the custom events the SDK fires so TED measures in its own tools.
6. Cross-platform UI SDK detail screens. Coordinated across web, Android, and iOS so the experience stayed seamless, with platform-specific behaviours communicated to TED during QA.

Product Judgment Under Pressure

The auto swipe request is the one I want to be judged on. TED asked for a behaviour I did not think was right for a feed, and I said so. Auto advancing pulls control away from the viewer, and in most feed products that is a mistake.

But TED is not most feed products. Their content goal is comprehension rather than scroll volume, and a viewer who leaves halfway through a talk has not got the idea. I argued the case, lost it on the merits of their content model, and shipped it.

That is the pattern I would repeat: hold the opinion, put it on the record, then serve the customer. A product person who never disagrees is not adding judgment, and one who disagrees and then obstructs is not adding delivery. The record of the disagreement is what makes the decision reviewable later.

Turning One-offs Into Product

An account with this many customisations becomes a maintenance liability if every fix stays bespoke, so the question I asked on each one was whether the underlying problem was a TED problem or a problem every content brand has. When it was the latter, I pushed for the fix to move into the platform.

Four worked examples, one of them still a proposal:

• Login helper text: built for TED because users did not understand the prompt, now generalised to all consumers.
• Reusable rendering: rendering parameters built for TED's web embeds, now generalised to the web SDK.
• Custom event tracking: TED defined the events it wanted fired via callback, now a repeatable capability.
• Configurable search: search defaulted to profiles over videos for content brands. This one is proposed and has not shipped. Stating that honestly matters more than claiming it did.

Cross-Company Alignment

This was a customer product integration, so most of the value was in aligning teams and functions across two companies. The named individuals I worked with most closely are covered above in "I worked with"; this section covers the teams.

TED Side

Genuin Side

Challenges & Trade-offs

Key Learnings

• Fit the product to the customer's world. Removing the interest step and resolving shares on TED's domain are both about respecting the ecosystem the customer already built.
• Have a product opinion, and still serve the customer. Disagreeing on auto swipe and shipping it anyway are not in tension; the record of the disagreement is what makes the decision reviewable.
• Turn one-offs into platform value. The account stays maintainable when every fix is checked for whether it generalises.
• A demanding customer is a forcing function. TED's QA rigour improved the whole release process, not just their account.