How we took Flamme's D30 retention from 0.8% to 50%+ with organic UGC

The numbers:

• 50M+ organic views
• 200K+ users
• D30 retention from 0.8% at the start, to 4.8%, then to over 50%
• $0 on paid acquisition

Here's how we ran it, and why Lauriane and I now run the same system for other consumer apps as Agniverse Media.

Views were never the goal

Most UGC agencies sell views. Views don't pay salaries.

A video can do a million views and bring in people who delete the app two days later. You see the spike on TikTok and nothing in your cohort chart. D30 retention tells you whether UGC brought the right people, so that's the number we reported on, every week.

What actually moved D30

UGC brought people in. What kept them was rebuilding activation and habit loops inside the app.

• The habit moved into onboarding. Flamme's core loop is a daily question for the couple. We put it inside onboarding, so the first thing a new user did was the thing we wanted them doing every day.
• The product was built for two. Relationships are multiplayer, but most relationship products are built like single-player apps. Flamme was incomplete without the other person, so the app had to bring the partner in.
• The feature did the inviting. Women usually joined first. Their partners came in because the feature pulled them in.

TikTok was our product lab

We didn't treat TikTok as marketing. We treated it as a product testing surface. Comments gave us raw user language, the exact words couples used for their own problems, and those words went into the next round of hooks.

The comments were also the test for whether a video worked. If people are not asking what is the app, what is the product, where can I get this, then it is not content that converts.

Every brief starts from the moment the app clicks

We built each brief around the point where a new user first gets value from the app, and the habit that brings them back. Not around the trending sound of the week.

When the video sets the right expectation, the person who downloads already knows what the app does for them. They don't churn on day two because the app turned out to be something else.

A system, not one lucky video

One viral video followed by six weeks of silence isn't growth. We run a fixed loop. Test hook angles across formats, publish, read the numbers, update the brief, repeat.

Here's how we picked hooks on Flamme. Every angle got 3 to 5 hooks, and every hook was tested with 5 to 10 videos. TikTok first shows a post to 200 to 300 people, so one post on its own can just be unlucky. Then the views decided:

• Under 1,000 views, we killed the hook.
• Between 1,000 and 5,000, we mutated it.
• Above 5,000 to 10,000, we scaled it.

One day was one testing cycle. At one point we were putting out 150 pieces every day to test a single hook.

We call it the VSC framework (viral, scalable, convertible). Every brief goes through it, so results don't depend on one creator having a good day.

Operators, not account managers

Lauriane Nayal and I ran Flamme's UGC ourselves. We still run every Agniverse Media account ourselves. No account manager three months into the job, no handoff after the sales call.

Start this week

1. Find your activation moment in your data: the action on day one that users who are still there at D30 had in common.
2. Move your core habit into onboarding, so a new user does it before they leave the first session.
3. Rewrite your creator briefs so every video shows that moment, in the creator's own words.
4. Treat each day as one testing cycle. For every hook, decide kill, mutate or scale from its views, and write the decision down.
5. Put D30 by source next to views in your weekly report. If they move apart, believe D30.

What it has done for other apps

Client names stay private. The numbers don't.

• Health app: 9M organic views and D30 retention up 9 points
• Travel app: 22M organic views
• Edtech app, pre-launch: 3M views and 20K pre-registrations before the app went live