I posted my iOS app on TikTok for six weeks. Marketing became the second product.

What I changed once production, attention, and subscription revenue stopped looking like one problem

I spent six weeks posting Pocket Orbit on TikTok while working a 9-to-5.

Pocket Orbit is the iPad note-taking app I was promoting at the time. I expected marketing it to mean finding a few useful formats, recording the product and learning which posts reached the right people.

Instead, every post created another small production chain. I needed an angle, a script, footage, an edit, a caption and an upload. If the same idea went to another platform, it needed another opening, crop and description. Then I still had to check whether the right file had actually published.

The posts received views, saves and questions about the app. Product activity moved during the same experiment. Premium subscriptions did not move the way I wanted.

Those facts looked as if they belonged to one story. They did not.

TikTok could tell me that somebody watched. Product analytics could tell me that somebody reached the app. RevenueCat could tell me whether a customer started a trial or subscribed. Without matching attribution and observation windows, none of them could tell me that one specific video caused the next event.

The useful result of the six weeks was not a viral case study. It was a clearer view of where my process kept breaking.

The launch created a second job

The first mistake happened before the first post.

I spent months getting the app ready and almost no time preparing the distribution process around it. Once Pocket Orbit was live, every channel felt urgent. TikTok needed native videos. X needed the reasoning written out. YouTube needed another title and description. Instagram wanted its own crop and framing.

Launch week was also the first time I learned how long one usable product video took.

At first, I treated every post as a separate creative project. I found a new idea, arranged a new setup, recorded new footage and made a new edit. That sounds manageable until it has to happen after work several times a week.

The first improvement was simple. I spent one weekend recording enough raw material for five of the next seven videos. I could still change the hook and edit without rebuilding the entire production setup every evening.

Batching did not make the ideas better by itself. It removed enough repeated setup for me to compare the ideas properly.

That distinction matters. A content system should not manufacture more posts than a founder can review. It should remove the repeated work around a real observation so there is enough attention left to judge the result.

The creative changed when I changed the recording

The format I expected to perform best was not the one that taught me the most.

One of the stronger results used a structure that was already familiar on TikTok. I kept the rhythm of the hook and the way the information unfolded, but rebuilt the message around Pocket Orbit with my own footage and claims.

That was not the same as copying another creator’s product, identity or result. I was borrowing a format the audience already knew how to consume.

Once I stopped treating originality as the only variable, another difference became obvious. The iPad recordings were working better than the early iPhone clips often enough that I changed the next batch.

The product was built for an iPad. On the larger screen, the note-taking use case was immediately legible. The interface had space. A hand using the Apple Pencil explained more than another sentence in the caption could.

I had grouped all of that work under one label: Pocket Orbit videos. The label hid the variables I should have been comparing.

Next time, I would keep the recording device, crop, background and first visible product screen state separate. A format is not only the script. The first frame can decide whether the product makes sense before the viewer reads a word.

Attention and revenue refused to become one story

Most early posts stayed in the low hundreds of views.

That number was frustrating because it looked precise while explaining almost nothing. It could point to a weak opening, a young account, poor audience fit, the wrong format or statistical variance, I could NOT know. One post could not tell me which.

The comparison became more useful when I stopped asking whether a number was good and started asking what it described.

Views described distribution inside TikTok. Average watch time and completion described attention. Saves, comments and profile visits described different kinds of intent. None of them described what happened after somebody reached the App Store.

During the same period, product analytics showed people arriving and signing up. RevenueCat showed customer activity. Seeing those lines move felt really encouraging, but it still did not connect one TikTok to one install or one subscription.

It also exposed a different problem.

Some people left during onboarding. The people who continued were not turning into premium subscribers at the rate I wanted. The flow asked for a trial decision before a new user had properly seen the product work.

I knew what Pocket Orbit did because I had spent months inside it. Somebody arriving from an short-form seconds long video had none of that context.

I started rebuilding the sequence so it could show real features in short loops before the paywall. That is still a hypothesis I am testing, not a conversion result.

The important change was separating the stages:

• TikTok attention
• App arrival
• Onboarding behaviour
• Paywall exposure
• Trial
• Subscription

Each dashboard could be accurate while the story connecting them was wrong.

The files became part of the experiment

The analytical problem was difficult, but the operational problem appeared every day.

One useful recording could become a TikTok edit, a YouTube Short, an X clip and an Instagram Reel. Reusing the raw product moment saved time. Uploading the same finished file everywhere would only have moved the mess downstream.

Each destination needed its own opening, crop, metadata and sometimes a different explanation. Once those versions lived in separate folders and upload documents, an edit to one did not reach the others.

The calendar showed when a post was meant to go out. It did not prove that the correct version had gone public. A successful upload did not always mean a viewer could see it. Planned, reviewed, scheduled, attempted and public had quietly become one status.

By the end, I could usually find the source footage. I could usually find a finished edit. I could usually find the planned date. I still could not answer the simplest question without opening several systems: what actually made it live?

That changed what I considered the reusable unit.

It was not the final caption or exported video. It was the real product moment underneath them. The source could remain stable while the presentation changed for each destination. Every version needed to stay connected to that source, its review state, its intended slot and its final publication result.

What I would change before the next launch

I would build the distribution process before launch week.

I would decide which product moments were worth recording, batch enough raw material to compare several posts, and keep each creative variable visible. I would reuse proven structures without borrowing somebody else’s identity or result.

I would treat the recording device and first frame as part of the argument. I would compare a series over the same observation window instead of asking one post to diagnose the account.

Most importantly, I would keep attention, product activity, trials and paid subscriptions separate unless the attribution actually connected them.

The next experiment is a controlled set of search-based carousels compared with the video formats I already used. I will keep the question, first-frame promise, device and observation window visible instead of changing everything at once.

I will check the same posts after one day, seven days and thirty days. If old carousels continue receiving search traffic, useful. If they disappear after two days, also useful.

I do not need the experiment to become a victory lap. I need it to make the next decision less random.

I posted my iOS app on TikTok for six weeks. Marketing became the second product. was originally published in Bootcamp on Medium, where people are continuing the conversation by highlighting and responding to this story.