Check first-session return before polishing your AI app

A demo feels finished until the next open

You just shipped a weekend AI app that works, looks plausible, and does a useful thing. The hard part is whether the first session leaves a reason to come back. Most side projects stay side projects, and that is fine. The question is which one deserves another weekend.

A solo-built iOS app called Amana was released on the App Store on 6 August 2026 for the RevenueCat Shipaton.

In the analytics window from 6 August to 9 September, it had 40 unique users and no one opened it on a second day.

Eleven of the 40 users reached the paywall, and gross revenue was $9.41 from one transaction.

Eighteen of the 40 users took a photograph outside.

Across its YouTube videos, Amana drew 3,590 views, 27 likes, and 3 subscribers.

That is attention without return. Use it as a warning before you add another feature.

Repeat opens are the clearest business signal

In Amplitude's 2025 Product Benchmark Report analysis of more than 2,600 companies, a 7% day-seven return rate among the original cohort placed a product in the top quartile for activation and built a base for long-term retention.

In that analysis, more than 98% of new users had stopped being active by the two-week point in half of the products studied.

Productivity applications lost 71% of their Day 1 retention between Day 1 and Day 30, falling from 32.86% to 9.63%.

One app had onboarding completion above 90% on both iOS and Android, but most users had left by day two, so completed onboarding did not equal retained use.

A founder's product attracted 5,000 signups in under five days, but only a small fraction of those signups became paying customers.

In AI companies that passed $1 million in ARR, the top performers were the ones with early-stabilizing cohorts and net dollar retention above 100%.

Early return creates the base that later revenue can sit on. Audit your first session and second-day return data with that standard.

Marketing can hide a missing return reason

Amana's YouTube Shorts had 3,528 total views across 16 videos, and the channel's most recent 28 days included 2,939 views, 3.1 hours of watch time, and 2 subscribers.

After test devices were removed, Amana's weekly active users declined from 8 in the first week of August to 6 by the end of August.

The shape is the warning: no repeat base formed.

GitHub Copilot reached 67% of developers using the tool at least five days per week after installation, showing repeated weekly use tied to workflow integration.

That kind of repeat use comes from workflow, not novelty. Ask whether your app would still be opened after the initial excitement.

Test the return reason before you add features

Run this first-session return checklist with a fresh install and your second-day return data. Each item should be quick, and the answer should tell you whether to iterate, pivot, or stop.

• Name the return reason in a sentence: what will the user do next time?
• Put that reason in the initial session, not after a tutorial or paywall.
• Do a fresh install and ask what would make you open it again.
• Measure second-day opens, not just installs.
• Compare paywall reach to return: if most users see the paywall before the reason, move the paywall.
• Set a kill or iterate threshold before adding features.

The first two items separate a useful action from a reason to return. A photograph can be useful, but it may not create a habit.

The fresh install catches what you already know from memory. Second-day opens are the cheapest signal you have, because they do not require a new audience.

Paywall reach matters only after the return reason exists; otherwise you are asking strangers to pay for an initial impression.

Flat second-day opens after a fresh install mean the return reason is missing, so change it before adding features. If a paywall appears before that reason, move or remove it for a test. When the initial session creates a clear next action, give it a short test window and watch second-day opens. If the app only works as a trick, archive it and write down what you learned.