700 downloads and stuck — five months later...

Thank you - "a meaningful change in the signal" is exactly what it feels like from the inside. Installs can be bought; emails can't.

Thanks - that's a fair map of the classic levers. Since you've walked the same plateau: which of the five moved your retention curve first, and what was the product?

On instrumenting early - agreed, and this month was exactly that: the app went from "some analytics" to end-to-end funnel tracking while the numbers are still small enough to be unflattering. I publish them anyway (28% week one, ~10% floor) precisely because waiting for numbers to look trustworthy is how you lose the window to steer by them. The email channel has its own instrument, humbler than a dashboard: a standing rule that every reply I send carries exactly one question. Three long emails so far - a sample too small to chart and too detailed to ignore.

On routing by acquisition source - I like it, and I owe you honesty about where it collides with my own constraints. My best traffic arrives by typed-in name search: no link, no referrer, no source bits to route on. And the standard fix - attribution SDKs - is off the table on purpose: no ad-tech in a privacy-first app, that's part of what people are buying. The one sliver where your idea survives both constraints is link-based campaigns on Android, where the plain install referrer carries a UTM and can pick an onboarding emphasis without any tracking beyond that. If the paid tests show a door worth doubling down on, that sliver is exactly where source-routed onboarding would start.

Happy to answer this one in the open, because it's measured now: the Threads-driven cohorts hold ~28% at week one and settle around 10% by week three. So yes - most of a spike is reach. But the tenth that stays is why the daily floor moved from a handful of installs a day in July to a few dozen now. The channel decays per-wave and compounds per-floor, both at once.

On being the bottleneck - agreed, and I've made peace with the shape of it: ten minutes a day is the one marketing cost I can always afford, and it can't be delegated without becoming ordinary marketing, which is exactly what makes it work. The paid creator placements I'm testing right now are precisely the experiment in whether any channel exists that doesn't need me in the room. First results within a couple of weeks.

And sure - if you want to go deeper, write to the app's support email from the store page; I answer everything there myself. Fair warning though: I keep the interesting parts of these conversations public - half this thread's value is that other builders can read along.

This thread has quietly reached consensus, and yours is the sharpest statement of it: one door per test, bundle stays inside, judge by the chain instead of the door's own applause.

Your last sentence is the cleanest job description in this whole thread: the door carries acquisition, the bundle carries retention, and neither should be graded on the other's work.

Where I'll be honest about the state of it: the rotation sits on my list as an idea still fighting for its budget. Every measurable spike I've had came from words in feeds, none from listing changes - so a month of storefront per door is a real cost against an unproven return, while ad creatives answer the same door question in days. That's the trade I'm weighing, not a plan I'm announcing. Have you actually run this rotation on something you ship? If you've seen what it moves - or doesn't - that data point would save me a month of finding out.

There's one load-bearing tool in here - taking it, with a constraint you didn't mention. Custom product pages: yes, and since mid-2025 Apple even lets you assign keywords so a CPP replaces the default page in search - but that part is live only for the U.S. storefront in English so far. My audience is Ukraine-heavy today, so the search play waits; the immediate use is the other half. I'm about to run door-specific ad creatives (calm place vs free calorie counter), and pointing each at a CPP whose screenshots match its door - instead of both landing on the generic page - is an obvious upgrade. That goes in the plan.

On the paywall block: the default is already an annual plan with a 7-day trial, and onboarding already ends with a personalized plan before any paywall. And I'd gently push back on "the blocker is almost never top-of-funnel" - my revenue was flat exactly as long as my visibility was flat, and moved the week visibility moved. Frameworks generalize; funnels don't.

One eyebrow, kindly meant: "increases conversion by 40%+" - whose funnel, measured how? Unsourced percentages are how founders end up mistrusting their own dashboards.

The share-card loop you describe is already built - the bonsai card exists precisely so a streak posted to Stories makes a friend ask what app that is. Whether it works is a number I'll have soon, not a vibe I'll report early.

When I traced the wave backwards, the trail ended in Threads - in those same ten-minute daily comments, not in anything that looked like marketing. The middle of the chain is stranger than you'd guess, and it deserves the full story. No link was involved at any step though: people typed the app's name into store search by hand. And your catch is the real one: it works precisely because it doesn't look like marketing, which also means it can't be scaled on purpose. I just get to keep showing up.

Agreed on the split - store stays plain, doors live in the campaigns. One sharpening for your metric though: install-to-first-action stops working when onboarding manufactures the action. Mine ends with a plan built and the first task already checked - by design, nobody meets an empty screen - so every install produces a "first action" whether the door worked or not. The honest version is narrower: first action in the module the door promised. Calorie-door install logs food same day - door delivered. Checks the onboarding task and vanishes - curious download wearing a conversion's clothes.

Conveniently, another commenter already talked me into "first touched module" as a cohort dimension, and your framing slots right in: door → same-day action in the promised module → week-1 of that cohort. The doors finally compete on something other than my taste.

For what it's worth, my data leans your way: people who reach the food module average ~26 searches each. I'd quietly bet on the calorie door - which is exactly why it needs the test instead of my bet.

Honestly? Neither - and that's the answer I'd defend. Both options assume there was enough traffic for the data to mean something, and there wasn't: for those five months the app was getting a handful of downloads a day. The problem sat upstream of intent and retention both - almost nobody ever saw the app.

I did have analytics back then, enough to see retention wasn't great. But debugging retention on five users a day is rearranging furniture in a house nobody visits. The real bottleneck was visibility, and my paid attempts at it - small Telegram placements at $0.2-0.5 per install - bought downloads that never moved revenue.

The proof came this month, the first time real word of mouth happened and the funnel got actual volume: week-1 retention sits around 28% and settles near 10%. That's a real leak and I'm working on it - but it's finally a leak in a pipe with water in it. Fair warning on causality: the relaunch shipped alongside the visibility, so I can't cleanly separate "product got better" from "product got seen". But the orders of magnitude belong to visibility. Below some traffic threshold, "downloads without traction" isn't a retention signal or an intent signal - it's just silence with occasional noise.

This is close to what my data has been whispering. Food is my deepest engagement pocket - people who reach the calorie logger average ~26 searches each - but only a few percent of users ever reach it, because calorie hunters have never heard of me. We swim in different ponds. So my next ad test runs exactly your split: one creative sells the calm place, the other sells a single painkiller - free calorie counter, no ads, no account - and the store listing stays out of the experiment entirely. Ad creatives are cheap to A/B; renaming the front door mid-wave is not.

One honest wrinkle from my own numbers though: my two biggest acquisition days so far came through no pain point at all. Both were words in a feed - people read them and typed the app's name into store search by hand. Word of mouth walks through a door marked with nothing but the name. So my working model is a bit narrower than yours: the single-pain-point door is for cold traffic you pay for; warm traffic arrives already sold on the house.

And the retention half of your thesis matches my inbox exactly: three users have written to me at length so far, and none of them praised a module - all three described a week of their life with the app. Nobody falls in love with a calorie logger. They fall in love with the week.

This lands on the one asymmetry of a no-account app: I can't reach my users - except the ones who reached me first. Which, as you say, makes those emails the entire addressable signal, and they've already proven their weight twice over: calendar import exists because I asked one student what she was juggling the week she wrote (medical school, three jobs, tasks scattered across a phone calendar) - and her answer became the feature within ten days. Nobody in those exchanges has ever asked for "a module"; they describe weeks, not features.

What I haven't done is your systematic version - going back through every email with the same question: "what were you trying to fix the week you installed it?". The list is a dozen people, so it's an afternoon, not a research program - and the timing is right, because my pricing page currently prices the feature list (limits, modules, exports) and not a single sentence of what those people actually describe, which is some variant of "my life was in five apps and none of them cared". If ten conversations rewrite one paywall headline, that's the cheapest pricing work available to me.

The painkiller/vitamin split is the sharpest version of the door question yet - and you correctly spotted that the calorie counter was born from exactly that frustration (I refused to pay for one, so I built one). Two honest corrections from my side of the data though.

First, a correction that makes your point stronger, not weaker: the calorie counter itself IS fully free in SelfOS - logging, food search, barcode scanner, water entirely free with no caps. Premium only gates the extras: detailed macros, custom dishes, meal photos. So the honest version of your door copy exists: "count calories free - no ads, no account, your data stays on your phone." What I can't claim is "no subscription" (premium exists and I'd like it to keep existing :)), but the core painkiller - "stop paying just to log your food" - is a promise I already keep.

Second, your closing question has a twist in the data: almost nobody ARRIVES for calories - the module reaches single-digit percent of users. But those who find it behave exactly like hunters: ~26 food searches per user, the deepest engagement in the app. My read is that this says more about my distribution than about demand: I fish in planner-and-self-care conversations, so I catch calm-place people, and the calorie hunters never hear about it. Which turns your frame into a testable ad hypothesis rather than a repositioning: keep the shelf as is, but aim a campaign at calorie-counter frustration and see if that door converts colder traffic better. That experiment just moved up my list - thank you :)

Your funnel-stage point lands, and I can answer your question with today's numbers rather than the March fog: the app is instrumented end-to-end now, and the cliff sits exactly where you point. Of first opens, ~63-65% finish onboarding, and ~28% are still there a week later. So yes - the biggest single loss is between the first open and the return, not in acquisition, and that's precisely the gap I'm working rather than widening the top.

Mobile spares me your installer-and-OS-warnings purgatory, but the "first screen with nothing in it" trap is universal - and it's the one thing I designed against hardest: onboarding ends with a personal plan already built, a bonsai already planted, and the first task already done. Nobody lands on an empty screen. Whether that's WHY the week-1 number looks the way it does, I can't prove - but the philosophy is the same as yours: the cheapest user to gain is the one you already have who almost stayed.

On your question - with the honest asterisk that "activate" is fuzzy here because modules ship enabled and users disable what they don't want (very few do): the first ACTION is tasks for a majority, though onboarding seeds that. The first health entry is water by a clear margin - it's the smallest possible promise, one tap for a glass. So the lived answer is: people arrive, do a task, log a glass of water, and the rest unfolds over days.

On Zarek - I'll be straight with you, since my post is literally the case study: automated posting into communities is the exact pattern that got me banned from Reddit despite writing every comment by hand and answering real questions. Moderators and automods don't punish bad copy, they punish the PATTERN - same product mentioned across communities on a schedule. A tool that industrializes that pattern will hit the same immune system harder, however good the copy is. The thing that made Threads work for me is precisely what can't be automated: I only ever reply to a person who already asked, and I'm accountable in every room I enter. That's the moat - and I'd gently suggest it's the constraint your product has to design around, not through.

Good push, and I can answer part of it with live numbers instead of intentions: onboarding completion currently runs at 63% (Android) and 65% (iOS) of first opens - measured per step, so I can see where the drop-offs sit. Week-1 return is ~28%. Module discovery is instrumented too (toggle events plus per-module first-use), and this week I added source-tagging and per-gate attribution to make those funnels cleaner.

The honest limitation is the one I confessed earlier in this thread: my "single-door launch" shipped confounded - rename, category move and four new modules landed within days of each other, plus a traffic wave right on top. So a clean before/after on positioning alone doesn't exist and can't be reconstructed; the wave cohorts are also small enough that a 5-point swing is noise. What I can do - and your framing sharpens it - is treat the current numbers as the baseline and read the NEXT deliberate change against them. That's the plan: baseline recorded now, one variable at a time from here.

The compounding argument is the strongest anti-rotation case in this thread, and it settles something I'd left half-open: earlier here I mused about seasonally rotating the lead pain. You're right that search, referrals and reviews all accrue to one message - so the door stays put, and any seasonal experiment will be an ad-campaign variable, not an identity change.

Your buyer-type instinct matches my data in a way that surprised me: calorie-counting is my stickiest module per user (the few who enter it search food ~26 times each), but those users are a single-digit percent of installs - a commodity audience passing through. The "one calm place" people are the ones who write the emails. So: door for the calm-place buyer, depth for whoever wanders into nutrition. Both real, different jobs.

Your closing question turns out to have a messier and more interesting answer than a clean either/or: the big wave of new users hit while the App Store still carried the OLD version - so they arrived at the planner-only app, and got the four health modules as an update a week later. A cohort that entered the narrow old door and had the house EXPAND around them. And here's the detail I keep coming back to: in all the emails since, nobody has ever praised a specific module. Every single one talks about the app as a whole - "calm", "everything in one place", "replaced my other apps". Which quietly supports your buyer-type theory: the people who write are "one calm place" people, whichever door they happened to walk through - with the honest caveat that my inbox is a dozen data points, not a cohort study.

That's the landing spot, yes - with one nuance from how it actually shipped: the "one door" isn't only the store pitch, it's the first-run experience. My onboarding commits fully to the health angle - a short intake quiz, then a personal health plan - and productivity (tasks, goals, habits) reveals itself only once you're inside using the app. The door and the hallway match, and that consistency matters more than the shelf copy alone.

Translated to FunderVibe: lead with the no-equity angle not just in the tagline but in the first thing a new user actually does on the platform - and let the full model surface once they've taken the first successful step. Good luck with the build)

Honest answer to your closing question: store reviews are anonymous, so I can't trace that person's path - and I owe the thread a precision: my "one door" today lives in the name (Self-Care Tracker), the category shelf and the onboarding - the description itself still opens with the bonsai and honestly lists the whole house. So the narrow door is real but partial, and that review predates even the renaming - that person definitely came in through the wide framing. One review is an anecdote, not data - but it keeps me honest: the narrow pitch is a bet on acquisition efficiency, not a law of nature. The broad framing demonstrably converted people too.

What I'd still defend: the praise itself is a retention observation. Nobody installs an app because it promises to replace three others - they discover it did, then tell you. Narrow door or wide, "replaced 3 apps" is what the house earns after install.

And thank you for reading the "half-listen" as a decision rather than avoidance - you're the first commenter to name that distinction, and it was the loneliest part of the choice at the time

Thank you - this matches where the thread has landed, and your "cross-sell once they're on board" phrasing is a useful way to formalize it. For context, this is already implemented: the store listing leads with self-care tracking as the single entry point, and since the 2.0 relaunch the onboarding follows one path - a short intake quiz, then a personal health plan. The remaining modules are introduced only after installation, effectively serving as the cross-sell layer you describe.

The lesson that took me longest to internalize is that focus is a marketing decision rather than a product decision - the product can stay all-in-one as long as the message leads with a single need. Appreciate you confirming that from the practitioner side.

Went and pulled the actual numbers instead of answering from gut - with one honesty caveat: my onboarding auto-creates a first task, so raw task metrics flatter themselves. I've just shipped source-tagging to separate real actions from onboarding ones; clean task numbers in a couple of weeks.

But here's what's already clean, and it breaks the freeze-experiment logic anyway: the stickiest module is one almost nobody enters. Food tracking reaches only ~6-12% of users - yet those who use it average 26 food searches each, the deepest per-user engagement in the app (onboarding doesn't touch this module at all, so the number is honest). Meanwhile tasks are clearly the most-walked door even with the inflation discounted - people create and complete them daily far beyond what onboarding seeds.

So the answer to "which one would I keep" splits by job: tasks acquire, nutrition retains, the bonsai stitches them into one habit. Freezing 7 modules wouldn't reveal the product - it would break the division of labor that IS the product. I only know any of this because all 8 are on and instrumented.

This is the most concrete share-cards advice I've gotten, thank you. Honest current state: the cards exist but sharing is manual today - a button, not a moment. Your milestone-prompt framing is the missing piece, and it lands on infrastructure I already have (the app already detects milestones for celebrations). One milestone, one layout, prompt at the moment of pride - going on the list as designed, thanks!

"Designing around reality instead of against it" - I'm keeping that phrase:), it names something I did mostly by instinct. One honest note though: the share cards are still a bet, not a result. I genuinely hope they'll end up doing that distribution work for me in the stores, but the update has been live for such a short time that I haven't recorded a single case of it actually happening yet.

On the door point - agreed, and this thread has collectively talked me into it: one door on the shelf, the house revealed after install.

On privacy - you've actually touched the part of SelfOS I'm most deliberate about, and my answer is architectural rather than procedural: there is no data side to govern, by design. No accounts, no cloud, no backend - every gram of health data lives on the user's phone and dies with an uninstall. Export exists so the data is the USER'S in practice, not just in a policy. That's not compliance-readiness, it's compliance-avoidance in the nicest sense: the cheapest data breach is the database that doesn't exist.

You're right that the calculus changes the day I add accounts, sync or AI - and that's exactly why I'm in no rush there. For now, local-first is both my privacy story and my favorite feature to not maintain.

I kept treating IG as a second full-time job I was failing at, which is exactly why it kept not happening. Repackaging the advice content I already write daily into visual format is a completely different (and doable) task than "become a reels person". Especially since, as I confessed in the post, I talk to a camera like it's a tax inspector - but quote cards and carousels don't need my face.

The bonsai share cards as UGC - agreed, that's the bet. They shipped recently, so the honest status is: asset built, flywheel not yet spinning. The wave has to come from users posting trees they're proud of, which means the real work is upstream: more people reaching day 30 with a tree worth showing.

And you're right that "Replaced 3 different apps with this one" is the strongest line I have - a user answered the all-in-one objection better than I ever did. My plan for it is slightly different though: I want it as an ad slogan rather than a screenshot - the line that leads creatives and promo materials, where its job is to stop the scroll.

Honest answer to the direct question first: bucketed together, for now. My cohorts exist but they're not segmented by entry point yet - partly a tooling gap (mobile source attribution is coarse), partly because until recently there wasn't enough traffic to slice without lying to myself. "First module touched" as a cohort dimension is very buildable though, and this comment just moved it up my list.

Your bonsai guess deserves a nuance: you're right that it's the differentiator, but it's the wrong kind of aha for a door. The bonsai's magic is deliberately slow - it grows from your streak, so its payoff arrives on day 4, not minute 2. It retains; it doesn't activate. The fast aha is the personal plan the onboarding builds for you in the first minute. So my working model is actually three layers: door (one pain), activation (your plan, instantly), retention (a small tree that quietly becomes yours). Naming that separation - door as marketing decision, not identity - is genuinely clarifying, thank you.

On seasonal rotation: the thread has me committed to holding one door until the current traffic wave settles and the baseline is readable - I've confessed elsewhere here that I lack the patience for clean experiments, so this is me borrowing discipline from strangers :) But January is the obvious test: health resolutions are the one season no health app should sleep through, and by then I should have the retention-by-door data to rotate on evidence instead of vibes.

If you mean inside the app - that's a question I actually wrestled with. Current answer: the onboarding commits to one door rather than presenting eight. It opens with a short quiz about you and builds a health plan - no module tour, no feature carousel. The other seven don't introduce themselves until you're in.

The "rotate by intent" version was on my whiteboard: ask people what they came for and shape the first screen around that. I shelved it out of fear - intent questions at step one add friction exactly where every extra tap costs installs, and a new user often doesn't KNOW their intent yet ("I want my life together" doesn't map to a module). My compromise is that the rotation happens one screen later: all modules are toggles, so people compose their own app in the first minutes - and plenty of them end up living in a module they didn't come for.

Long-term I'd love to A/B a soft version - not a questionnaire, just reading which module gets touched first and letting the home screen learn from it. If you meant the store message rather than the app: same philosophy, one door on the shelf, rotation only inside the conversation with an actual human.

Honest status of your chain: the middle and the end I can see - module usage and return cohorts are tracked, and this release added per-gate analytics so I know which feature people hit right before deciding about premium. The weak link is the source: mobile attribution for "arrived from a Threads comment" is coarse at best. I use tracked links where I can, but a store install laundered through an app-store page loses most of its origin story. So I can answer "which module retains" well, "which door retains" only approximately.

And your "stronger first impression" alternative is exactly what I can't rule out yet: trials picked up after the health-first onboarding shipped, but impression and retention are different organs. The current traffic wave will settle soon, and the 7-day curves underneath it will be the honest answer. I'd rather read them than guess - the one rule I try never to break is not sweetening my own numbers to myself.

Happy to dig in, since you're standing where I stood. The specifics: the bans were subreddit-level, not site-wide. The account was relatively new. Frequency was modest - a couple of comments a day, every one of them a genuine answer to someone's specific "recommend me something" request. The variable that killed it, I'm fairly sure, was naming my app by name in those answers with a young account. To a human I looked like a helpful person; to automod heuristics and tired mods I looked like every paid shill they've ever removed - new account, mentions same product, repeatedly. Same behavior, different immune system.

So your version is already different in the way that matters: you're not naming your thing unless directly asked. That plus an aged account with genuine karma in those communities is probably the sustainable form. My honest guess is that Reddit punishes the PATTERN, not the intent - and there's no way to show intent to an automod.

Why Threads tolerates it structurally: there are no mods and no automod standing between you and the reader - the algorithm decides reach, and the algorithm optimizes for engagement, not for community purity. A product mention that gets replies is a feature to Threads and a violation to Reddit. Neither is wrong, they're just different organisms.

If you do stay on Reddit long-term, I'd love to hear how it holds up - you're running the experiment I abandoned)

"The feeling being heavier than the numbers" - you compressed my whole March into one line. And yes, the comments-as-cofounder thing is not a metaphor for me: actual decisions in this update trace back to specific sentences from that thread. A stranger's comment from five months ago outvoted several of my own ideas, and the app is better for it.

Building alone doesn't mean deciding alone - took me a while to learn the difference. Thanks for reading it the way it was written :)

The discipline you're describing is right, but my architecture makes the interview literally impossible - and I chose that on purpose. The app is local-first: no accounts, no backend, no user database. Everything lives on the person's phone. I couldn't email my paying users if I wanted to - I don't know who they are beyond anonymous store transactions. Privacy as a feature has a price, and this is it.

So I'm answering your question the only way I can: behaviorally. The latest release tags every paywall encounter with WHERE it happened - which limit, which feature, which screen the person hit before deciding to pay or bounce. In a month I won't have quotes, but I'll have a ranked list of named jobs that preceded every subscription. Written down, as you say - just by the analytics, not by the customer.

The ones who email me first are the exception, and their requests already steer the roadmap. But you're right that they're the loud minority - the data has to speak for the quiet ones.

Thanks! Community over raw downloads is fair - though my users live in wellness spaces, not developer platforms, so that's where I show up)

Thank you! Consistency really is the whole trick - most days it's just ten unglamorous minutes of showing up. Comments like yours make those minutes lighter :)

Thanks - and to be fair to the app, 700 was the starting point five months ago, not the current state; this post is the update on what happened since :)

Distribution so far, short version: manual comments on Threads is the engine (I answer people who ask for app recommendations in my niche - every growth spike came from that), Telegram ads as a background channel, and Apple Search Ads tested on popular and long-tail keywords - that one gave me nothing measurable. The longer version with the TikTok surprise is coming in a separate post soon.

Agreed on the diagnosis though - at my scale the bottleneck is eyes, not product.

You'll be glad to hear the door is already picked and it goes deeper than the shelf name. Since the 2.0 relaunch the whole first-touch experience commits to health: onboarding starts with a short quiz about you - parameters, activity, goal - and builds you a personal plan before you've seen anything else. The other seven modules don't even introduce themselves until you're inside. And judging by how trial starts picked up after this shipped, that single-door entry is what hooks people - they come for one clear promise, not for a brochure of eight.

So my August answer to my own March question turned out to be exactly your advice: one wedge at the entrance, breadth as the reason to stay. Where I was still wobbling was whether to rotate the door seasonally - between you and the commenter elsewhere in this thread who caught my confounded experiment, I'm convinced: hold the door, let it accumulate, read the data when the current wave settles. Thanks for sharing the lesson you paid for - "ours included" from a founder who lived it lands differently than theory)

This distinction deserves more words than it usually gets. Downloads measured my reach; the unprompted emails measured something else entirely - and they arrived only after 2.0 shipped, as if the app had to cross some invisible "worth talking to" threshold first.

What I can confirm from the inside: the "people care" signal is already writing my roadmap. Calendar import shipped ten days after 2.0 purely because the same request kept arriving in different words. That loop - someone asks, it ships, they see their words become a feature - creates a kind of user you cannot buy with any ad budget.

Where you've pushed my thinking is the last step: letting the caring users write the marketing too, not just the roadmap. My favorite review says "replaced 3 different apps with this one" - and I've been treating it as a compliment when it's actually copy. The person who cares tells you the message in their own words; my job is mostly to stop paraphrasing them.

The principle is real and I can confirm it from this month, not from theory. My app had a clean crash-free record until exactly the week real payments started - then two crash types appeared at once (one broken vendor ROM, one Android 13 framework race). If I hadn't been watching Crashlytics, I'd have learned about it from one-star reviews or never. So yes: silent breakage is not hypothetical, and "some will just leave" is exactly how it works.

On the checklist: error logs - watched daily, both platforms. Onboarding on a slow connection - this one I can answer with unusual confidence, because the app is local-first: no account, no network required, and I literally shipped a fix this week for the one place where a missing connection delayed startup. WebView console errors - covered by crash reporting plus my own error capture. But the spirit of your comment stands and I'd co-sign it for anyone reading: your users will not file bug reports, your logs have to do it for them.

Guilty as charged - it shipped confounded, and I knew it when I pressed the button. The honest reason is less strategic than it sounds: I simply wouldn't have had the patience to stretch this into three clean sequential experiments. And there's a smaller-print problem: clean experiments assume enough traffic to read them. Before the wave, my organic was a trickle - almost every install traced back to either a comment or a Telegram ad. A/B testing a shelf nobody walks past is its own kind of confounded.

You're right about the consequence though: I'm riding a traffic wave right now, and when it passes (sadly, they always do), the flat baseline underneath will be the readable result you're describing. And to be fair to my screenshots - they converted decently before the rename too, so the rotation idea is less "fix what's broken" and more "cheap monthly probe". Taking it.

A question back: what's a realistic ceiling for pure store organic, no marketing at all, for a niche health app? Ballpark installs/month. I've never seen honest numbers on this and I suspect they're humbling.

If you could offload one part of that loop tomorrow — marketing, customer research, prioritization, sales or operations — which one would you hand off first?

Honest answer: one channel, done very manually. Threads. I basically hunt for my users by hand - I search for comments where people ask "recommend me a productivity app" or "what do you use for wellness tracking" and I answer like a human, not like an ad. Most of those comments do nothing. Once in a while one takes off, and that single comment brings more installs in two days than a quiet month. Every growth spike this app ever had traces back to one of those.

Around that engine I did try paid: I run Telegram ads fairly often - they're more of a background hum for me, only a few channels ever performed really well. And I tested Apple Search Ads on both popular and long-tail keywords - that one gave me nothing. Not "meh ROI", literally nothing I could attribute. So my honest ranking: manual Threads comments > Telegram ads > App Store search ads. Next on my list is Instagram - I want a visual shopwindow that works while I sleep.

Your situation sounds like the same lesson from the other side: 661 technically perfect pages are the product, not the distribution. If I were you I'd go where people literally ask the questions your calculators answer (Reddit, niche forums, Threads) and answer them by hand for a month. Slower than SEO promises, but it doesn't wait for Google's trust.

Agreed in principle, and it's roughly what I'm doing. I'd just split the "door" in two, because they're different decisions.

The acquisition door is the one you mean, and health won it: the app moved to Health & Fitness, and onboarding now builds a personal plan from your body parameters - water, calories, sleep, steps. Health leads, and the productivity half arrives as the surprise.

The door inside the app I deliberately left wide. After onboarding everyone sees all modules, because the cost is asymmetric: removing a module you don't want is one tap, while a module you never switched on is one you'll probably never discover. So people trim the house instead of furnishing it.

On measurement itself: I do see which modules get used, module toggles and per-module events are all tracked. What I don't have yet is an honest link between that and retention, because the traffic wave is two weeks old and the monthly cohort hasn't closed. Usage I can read today; "which module actually brings people back" needs another month before I'd quote it. And I'll have a look at amami.dev - thanks for the pointer!

Thank you! Taking your questions in order.

Conversion from download to active user: honestly, too early for a number I'd trust. The app has only had two weeks of real traffic, so the weekly and monthly windows haven't closed yet, and the denominator right now mixes people who installed before 2.0 with a cold wave that arrived last week. I'd rather wait for the cohort to mature than quote something flattering and meaningless. That's the subject of my next post.

Feedback: yes, and it works better than I expected. There's an in-app rating prompt, and it has brought 27 reviews on Google Play and 24 on the App Store. Separately, people started emailing me directly after 2.0, and those emails have effectively become my product team - one of them turned into a shipped feature in ten days.

Directories: I'm on AlternativeTo, though I still haven't updated the listing for the new version, so that's on me. Product Hunt I keep circling and not committing to. If you've launched a consumer app there, I'd genuinely like to know whether it was worth the day it costs.

And you're right about doubling down: the channel that brought the first 700 is Threads, and that's the one I keep hammering. Everything else has been a distraction so far.

I know that dashboard-refreshing far too well!

The most useful thing I did wasn't a feature: it was giving the app an emotional core. A bonsai tree that grows with your streak, so the streak isn't a number in a corner but a living thing you're responsible for. Water, a task, a logged meal all feed the same tree. That's the one change that moved retention rather than installs.

Honest about the size of it though: it retains somewhat better than no tree at all, not dramatically. And I redesigned the bonsai visuals in the latest version, so I'm still waiting on the retention numbers for that one. I'll be able to say something real in about a month.

Yes, it paid off, and the clearest sign isn't a number: the product finally feels whole. Before 2.0 people were landing on half an app and I knew it. Now it's the thing I described from day one instead of a promise. (I also stopped paying anyone for a calorie counter, which was a genuinely motivating part of the plan 😄)

The timing is funny too. Right now the app has caught a wave of new people, and I'm spending my days fixing things that only break at volume - the kind of problem you simply cannot find with 700 downloads and no traffic. Tiring, but it feels completely different from before: this is working with a live product real people depend on, not polishing something in silence. That part is genuinely wonderful.

Thank you! And yes, the emails are the thing nobody warns you about. Downloads are a number; a stranger taking twenty minutes to write you a bug report with screenshots is something else entirely.

Small correction though, since it changes the maths: there's no marketing team here, it's just me. Your LinkedIn/TikTok finding is interesting because mine points the opposite way, and I think it's because I don't use these platforms for reach at all. Threads is my one channel, and it works not because the algorithm spreads my posts but because I can search for the conversation already happening - "which water tracker do you use?", "what app for tasks and long-term planning?" - and answer that person directly. One install at a time, and it produced every growth spike this app has had.

Instagram is what I want to grow next. My search-and-answer trick doesn't exist there, so organically it hasn't given me much yet, but I read that as needing a different mode: a shopwindow plus paid promotion instead of conversations. The app is visual enough to earn that - the bonsai tree that grows with your streak, the themes, the share cards. No Google Ads experience here, but Instagram ads are exactly what I'm considering, so: did you try paid promotion there, or only organic? That's the gap in my data before I spend money.

Distribution being the harder half - painfully yes. The product is the part you fully control; distribution is the part where you're a guest in someone else's system, and every one of those systems has its own rules and its own immune response.

When I moved to Health & Fitness I rebuilt onboarding around the health side: it asks for your body parameters and builds a personal plan from them - water, sleep, calories, steps - and then ends by having you create your first task, which is the bridge back to the productivity half. So people don't walk in through one door and discover the rest later; they arrive with several modules already switched on. Whatever cross-module adoption I see, I handed it to them.

What I can say is that it's working better than I feared. Water and food get tracked heavily, tasks do get created, and "everything at once" doesn't seem to overwhelm people - my worry was that eight modules would read as clutter, and so far it reads as choice. Which module eventually wins is still open; I need more months before I'd trust an answer.

Thank you!

Before committing to 2.0 I went looking at the products that tried to be everything at once. What I found it was a handful of things that had been sitting in pre-release for a very long time. All-in-one doesn't usually die on the spreadsheet; it dies on sheer volume of work. Eight modules is eight products' worth of edge cases. My read is that it survives mostly when the person building it is also the person who needs it every morning - that's a far more stubborn reason to keep going than a roadmap.
My reference point was Notion, but I wanted the opposite trade-off. Notion will be anything you want - after you assemble it yourself, block by block. I wanted the functionality to arrive already built, and the customization to be a toggle rather than a construction project. That's why every module is optional: you're not designing your system, you're switching off the parts you don't need.

On the poll - agreed, and a quiet version of it is already running. Since 2.0 users have been emailing me (bug reports with screenshots, feature requests), and that inbox has become my product team. On top of that, the modular design is a continuous survey in itself: I can see which modules people actually switch on and off, which tends to be more honest than what they'd tick in a form. What I still owe myself is tying that to retention instead of just counting it.

Thank you! The three doors metaphor is a keeper. It matches what I see in reviews: the "replaced 3 apps" person and the "one calm place" person are clearly not the same buyer.

Funny enough, I almost built a version of "different first impressions": letting people pick their modules during onboarding. I backed off for one reason - I was afraid nobody would ever open a module they didn't pick on day one, while removing one takes a single tap anytime. So for now everyone sees the whole house first and trims it down, not the other way around.

And yes - retention is exactly where my focus already lives. Your framing is a good reminder that it's the right way:)

This is very close to what actually works for me - I just run it in public threads instead of DMs. My Threads playbook isn't "post more": I search for specific people asking specific questions - "where do you track your water?", "which calorie tracker do you use?", "what app for tasks + long-term planning?" - and answer each one personally, suggesting SelfOS where it genuinely fits. One person, one conversation, one install at a time. That's what produced every growth spike this app has had, and it's exactly your "learn which words make people pay" phase: by now I know which phrasings convert, because I've typed them hundreds of times at real humans, not into a keyword tool.

I ran the identical playbook on Reddit and got banned within two weeks 😄 Same behavior, different immune system.

But where you've got me is "watch them use it": with a no-account, on-device app I literally can't see anyone's usage, and live 1:1 conversations with existing users have been rare - mostly whoever emails me first.

Thank you! And good news - shopping lists can already be shared, there's a share option that sends the list through the system share sheet, so you can drop it straight into whatever messenger your partner actually uses.

Real-time sync between two phones is the honest gap though - that one genuinely needs a backend, and it's on the long-term radar. But whenever it comes, the rule stays the same: accounts optional, the app fully useful without one. That part of the philosophy isn't negotiable.