Funnels need refuelling, loops feed themselves
Awareness, install, activation, revenue. The funnel describes one user's path and it is useful for that. Its problem is that it says nothing about where the next user comes from. Every week starts at zero and needs a new input, usually money. A loop takes the output, a user, and turns it into the next input. Installs lift rank, rank brings installs. Reviews write listings, listings convert. Revenue funds ads, ads bring revenue. The framework is Reforge's, from 2018, and it is still the model growth teams use because it answers the question the funnel does not.
What is new is that the stores themselves now close loops that used to be manual. Apple's ranking weighs retention and rating recency, so keeping users is directly a discovery signal. Google Play's Ask Play reads your description and your website, so a clear listing is also an answer engine strategy. And an agent can run the loop every day, which is what loops always needed and rarely got.
The organic loop, and the brief that starts it
Search on the stores and the web is one part. Short form content, comments and reviews are the rest. The organic loop runs at zero cost and is where most apps get their first thousand users. Its parts: keywords per country, a listing that carries them and reads well to App Store Tags and Ask Play, a hook a day drafted from reviews, every comment and review answered, and web pages that rank for the same terms and get cited by ChatGPT, Perplexity and Gemini when someone asks which app to use.
The loop closes when a review becomes a listing fix and a hook, and the hook brings a user who leaves the next review. Nothing in it needs money. Everything in it needs consistency, which is the part an agent is good at.
Run the organic loop for [app] with no budget. Every day: post one hook drafted from our reviews, answer every review and comment, and log what people asked. Every 14 days: ship one listing change from the keyword plan and measure it. Every Monday: report impressions, downloads, trials and day 7 retention per country, and tell me which review or comment became a listing change or a post this week. Do not spend anything.
The paid loop, and when the agent is allowed to open it
Revenue funds ads, ads bring users, their revenue funds more ads. The loop only closes if a user pays back the cost of acquiring them inside a window you can finance, usually 60 days. That is why the paid loop starts from the organic winners: your best keywords become Apple Ads on your own terms, your best hooks become TikTok Spark Ads, and the bids follow measured trial to paid rather than installs.
The 2026 detail: custom product pages, now up to 70 per app, can be linked to keywords and appear in organic search as well as behind ads. So the same page that converts an ad click also converts an organic one, and the paid loop feeds the organic loop instead of running beside it. Google's custom store listings show a median 10 percent conversion lift in campaigns for the same reason.
Do not spend until day 30 retention holds above 20 percent and we know our trial to paid rate. Then price a paying user per channel from our own data. Start with Apple Ads on our own top keywords at $20 a day, with a custom product page per keyword group. Raise a campaign by a third each week only while payback stays inside 60 days, hold it when payback drifts, pause it with a reason when it leaves the window. Never exceed the monthly cap I set, and ask before proposing a higher one.
The retention loop, the one most teams skip
Users who stay are what make the other two loops pay. A cheap install that leaves on day three is a loss. The retention loop watches day 1, 7 and 30, finds the biggest leak, and fixes it: the onboarding step, the push permission prompt, the win back offer for lapsed trials. Every fix raises the value of every install the other loops bring, and since the stores now rank on retention, every fix also lifts discovery.
Google Play made two of these fixes cheaper in 2026: delayed charging retries failed payments inside the app and cuts involuntary churn by up to 18 percent in Google's own numbers, and the account recovery window doubled to 60 days. Both are settings, not features, and an agent can tell you whether they are on.
Read day 1, 7 and 30 retention per cohort from App Store Connect and Play Console. Find the single largest drop between two days and tell me what happens in the app at that moment. Draft the fix: an onboarding change, a push prompt moved after the first win, or a win back message for lapsed trials on day seven. Check that delayed charging and the 60 day recovery window are enabled on Google Play. Show me the August cohort against July every Monday.
How the three loops feed each other when one agent runs them
This is the part a tool cannot do and an agency does once a quarter. A repeated complaint in reviews becomes a screenshot caption, which is indexed, which lifts a keyword, which makes the Apple Ads campaign on that keyword cheaper, which funds a TikTok test of the hook that came from the same complaint, whose comments become the next reviews. The signal crosses four surfaces in one week. Nobody has to remember to carry it. That is the difference between running loops and having three separate projects.
Treat every signal as input for every loop. When a review complaint repeats, propose the listing fix, the content hook and the retention change together. When a hook wins on watch time, propose it as a Spark Ad and put its words into the screenshot captions. When a keyword's organic rank rises, lower the bid on it and move the budget to the next term. Report the chain, not the parts.
What this means on Monday
Stop asking how many installs did we buy. Ask which loop moved and why. If organic impressions rose, which listing change did it. If paid payback improved, which retention fix caused it. The report that answers those questions is the plan for the week, and it is the only meeting the loops need.