Imagine you spent the last three months building your vibe code app from scratch. You launched, and the sign-ups started rolling in. Five hundred new users this week alone. It feels incredible, right? But here is the uncomfortable question nobody is asking you: where did last month's 500 users go?

If your app is leaking users as fast as it acquires them, you are not growing. You are just running on a treadmill that looks like progress.

This is one of the most common traps founders fall into, and the good news is it is completely fixable once you know what to look for. In this tutorial, you will learn exactly how to tell the difference between real growth and a leaky bucket. We will cover why sign-up numbers are misleading, what genuine retention actually looks like, and how to run a basic cohort analysis even if you have never heard that term before. By the end, you will know which metrics actually matter, how to read a retention curve, and how to turn that data into decisions that make your app stickier, stronger, and built to last.

The Leaky Bucket Problem Every Vibe-Coded App Founder Needs to See

Here is the math that most vibe-coded app founders never stop to run.

Your app is pulling in 500 new sign-ups every week. The dashboard looks great. The number keeps climbing. But your Day 30 retention sits at 3%, meaning only 15 of those 500 users are still around a month later. The other 485 are gone. Next week, you add another 500, lose another 485, and the cycle repeats. You are not building an audience; you are running a conveyor belt that empties itself every 30 days.

That is the leaky bucket problem, and it is more common in vibe-coded apps than most founders realize.

Why vibe-coded apps are especially vulnerable

Rapid ship cycles create a specific blind spot. When your entire build loop rewards speed, getting to launch and driving sign-ups feels like the job. There is no natural forcing function that surfaces churn early. The sign-up counter goes up, the Slack celebration fires, and the retention curve quietly collapses off-screen. By the time the numbers are unavoidable, months of acquisition spend have already flowed through a drain.

The gap between feeling growth and having it

Rising sign-up counts, a Product Hunt top-five finish, a mention in a newsletter: all of these feel like growth because they produce visible, shareable numbers. Real growth looks different. It is users returning to your app on Day 7, completing the core action your product exists to enable, and appearing again in the following week's cohort. One is a signal of attention; the other is evidence that your product is delivering enough value to earn a return visit.

Sign-up count is a lagging vanity metric. It tells you what your marketing did last week, not whether your product is working. The only number that predicts sustainable growth is retained users completing the core action, tracked across successive cohorts over time.

What this tutorial covers

By the time you finish reading, you will know how to instrument three distinct cohort types that reveal where your retention breaks down, how to read a retention curve and understand what its shape is telling you, and which specific actions to take based on what you find. No enterprise analytics stack required.

The leaky bucket problem exists because a specific class of metrics makes it invisible.

Vanity metrics are numbers that go up reliably whether or not your product is delivering value. Total sign-ups, total downloads, total page views: all three can climb while your actual user base sits at zero retained customers. They measure activity, not outcomes. As MasterClass defines them, vanity metrics "look good but lack actionable data" and "suggest growth without data to back it up." That is the precise trap.

The psychological layer makes it worse for vibe-coded app founders specifically. When you ship fast using low-code or vibe coding tools, you have already made a bet: that speed of build equals speed to product-market fit. Early sign-up numbers feel like confirmation of that bet. They are not. They are confirming that your landing page was compelling enough to collect an email address, which is a much lower bar. Founders in this position are often unconsciously looking for reasons the build was the right call, which makes them more likely to stop at sign-up counts and less likely to dig into what happened next.

Here is what that looks like in practice. A Product Hunt launch drives 1,200 sign-ups in 48 hours. The founder posts a screenshot, the comments celebrate, the dashboard shows 1,200 users. Ninety percent never open the app again after session one. The dashboard still reads 1,200. Nothing in that number tells you the product failed to deliver value to the overwhelming majority of people who tried it.

That 90% churn does not just disappear quietly. It creates what you can think of as acquisition debt: every unretained user adds noise to your analytics, drags down open rates and inflates churn on your email list, and corrupts the baseline you will later try to use for cohort comparisons. The more acquisition debt you accumulate, the harder it becomes to read whether future product changes are actually improving retention.

The analytics industry has been forced to reckon with exactly this problem. The 2026 move toward AI-powered retention anomaly detection in mobile app analytics platforms exists for one reason: raw acquisition counts kept misleading founders into thinking things were fine. Even sophisticated teams kept optimizing for numbers that did not predict growth. This pattern is why digital marketing for SaaS companies is increasingly broken: the metrics most dashboards surface by default are the ones least connected to real business outcomes.

Sign-up count is not a growth metric. It is a starting point for a different question.

What Real Growth Actually Looks Like in a Mobile App

So what does the alternative actually look like?

Real growth has a specific operational definition: a growing share of each successive cohort is completing the core action at Day 7, Day 14, and Day 30. Not opening the app. Not logging in. Completing the one thing your app exists to do.

The industry benchmarks give this definition real weight. Across 31 app categories worldwide, only 25.3% of mobile users return on Day 1, and average retention drops to 5.7% by Day 30. Those numbers are the floor. If your app sits above them, you have evidence of product-market fit worth scaling. If you're below them, more acquisition spend will not help.

What "completing the core action" actually means

The core action is specific to what your app does:

Task manager: A user creates a task and marks it complete
Habit tracker: A user logs a streak entry
SaaS tool: A user reaches the moment the primary feature delivers its output (a report generated, a file processed, a workflow saved)

If you are not tracking that specific event, you are not measuring growth. You are measuring presence.

Active users vs. logged-in users

This distinction trips up a lot of mobile app analytics setups. A logged-in user opened the app. An active user completed the core action. These are not the same thing, and conflating them masks the real retention picture entirely.

A user can open your task manager, stare at an empty list, and close it. Your dashboard counts them as active. They are not. Building retention curves on login events makes a struggling product look healthier than it is, which delays the decision to fix what's actually broken. If you want to turn the first 100 into a feedback loop, you need clean signal from the start.

The shape that separates healthy apps from broken ones

Healthy retention curves follow a predictable pattern: a sharp drop in the first few days, then a flattening into a stable retained base. That plateau is the signal. It means a subset of users has found genuine, recurring value.

A curve that keeps declining toward zero, with no plateau, is not a marketing problem. It is a product-market fit problem. No acquisition channel fixes it. The only path forward is understanding why no one is forming a habit around the core action, and that diagnosis starts with cohort analysis.

Cohort Analysis Explained for Founders Who Have Never Run One

Knowing your retention curve shape is useful. Knowing why it looks the way it does requires a different tool entirely.

A cohort is simply a group of users who all signed up during the same calendar window, tracked forward together. Everyone who joined your app in the first week of March is one cohort. Everyone who joined the second week is another. You watch both groups over time and compare what they do.

That comparison is what cohort analysis gives you that a raw sign-up count never can: direction. A single retention rate tells you that 8% of your users came back at Day 30. It does not tell you whether that number is improving, declining, or holding flat as you ship new features. Cohort analysis shows the trend across successive batches of users, so you can see whether the product changes you made in week six actually moved the needle for the cohort that signed up in week seven.

What the Retention Table Looks Like

The basic format is a grid. Rows are cohorts: Week 1 users, Week 2 users, and so on. Columns are time intervals: Day 1, Day 7, Day 14, Day 30. Each cell holds one number, the percentage of that cohort still active at that interval.

Cohort	Day 1	Day 7	Day 14	Day 30
Week 1	28%	14%	9%	6%
Week 2	31%	16%	11%	8%
Week 3	33%	19%	13%	10%

Reading down the Day 30 column, you can immediately see whether retention is improving cohort over cohort. That directional read is the entire point. If you were looking at a blended average across all users, the improving trend in later cohorts would be buried inside one flat number. For founders who want to go deeper on connecting retention data to churn risk, Paid to Retention: Churn Prediction and Customer Health Tools covers the downstream signals worth tracking once your cohort baseline is in place.

You Do Not Need Enterprise Tooling to Start

Mixpanel, Adjust, and AppsFlyer all include cohort analysis as core functionality, and they do it well. But the rest of this tutorial is built around a self-serve path that works before you have any of those platforms set up, using the event data you already have or can start logging today.

The Three Cohort Types Every Vibe Coding App Should Instrument

Now that you know what cohort analysis is, the next step is understanding which flavors to actually build. There are three, and each answers a different question about your app's health.

Time-Based Cohorts: Is Retention Getting Better?

Group users by the week or month they signed up, then track what percentage return at Day 1, Day 7, and Day 30. This is your baseline. It answers one question directly: are successive cohorts retaining better as you ship improvements?

A concrete example makes this real. Say you built a budgeting app with a low-code tool and shipped a UI redesign in week eight. When you pull time-based cohorts, users acquired before the redesign retain at 7% by Day 30. Users acquired after it retain at 18%. That gap is the signal that justifies doubling down on the new design, not a gut feeling or a tweet about how much cleaner the interface looks.

Behavior-Based Cohorts: Where Does Activation Break?

Instead of grouping by signup date, group by what users did (or skipped) in their first session. Users who completed onboarding versus those who bailed. Users who hit the core action in session one versus those who bounced before it.

This cohort type locates the activation gap precisely. If users who completed onboarding retain at 22% at Day 30 and users who skipped it retain at 4%, you have identified a specific, fixable problem. The next section covers what to do with that diagnosis.

Acquisition-Based Cohorts: Is Your Best Channel Also Your Best Retention Channel?

Group users by the source that brought them: organic search, paid social, referral, or direct. Then compare Day 30 retention across those groups. Often, your highest-volume channel is not your highest-quality channel.

This matters especially for vibe coding apps that rely on Product Hunt launches, Reddit posts, or influencer shoutouts for early traffic. Those channels generate genuine spikes, but the users they produce tend to be curiosity-driven rather than need-driven. They sign up, poke around, and leave. If you do not segment them out, their churn drags down your overall retention number and makes your product look worse than it actually is for users who arrived with real intent.

Understanding how acquisition channels map to downstream user behavior is what separates a clean retention picture from a distorted one. Segment your channels before you draw any conclusions from the aggregate numbers.

How to Run Basic Cohort Analysis Without an Enterprise Analytics Stack

Knowing which cohort types to track is step one. Actually running the analysis is where most founders stall. Here is a four-step path that works without an enterprise analytics budget.

Step 1: Instrument Three Events Before Anything Else

You need exactly three events logged in your database: a signup event (with timestamp and acquisition source), a core action event (the thing your app exists to do), and a session start event. No tool, paid or free, can produce cohort analysis without these three. The most common mistake in vibe-coded apps is discovering, weeks later, that the no-code backend logged page views by default but never recorded the core action. Page views measure visits; they do not measure value delivery. If you built on a no-code platform, check the webhook or function layer and add explicit event logging there before you do anything else.

Step 2: Run the Foundational SQL Query

Once events are in a database like Postgres or Supabase, this query gives you Day 7 and Day 30 retention by cohort week:

WITH cohorts AS (
 SELECT
 user_id,
 DATE_TRUNC('week', signup_date) AS cohort_week,
 acquisition_source
 FROM users
),
activity AS (
 SELECT DISTINCT
 user_id,
 DATE_TRUNC('day', event_timestamp) AS active_date
 FROM events
 WHERE event_name = 'core_action'
)
SELECT
 c.cohort_week,
 COUNT(DISTINCT c.user_id) AS cohort_size,
 ROUND(100.0 * COUNT(DISTINCT CASE
 WHEN a.active_date BETWEEN c.cohort_week + INTERVAL '6 days'
 AND c.cohort_week + INTERVAL '8 days' THEN a.user_id END)
 / COUNT(DISTINCT c.user_id), 1) AS day_7_retention,
 ROUND(100.0 * COUNT(DISTINCT CASE
 WHEN a.active_date BETWEEN c.cohort_week + INTERVAL '29 days'
 AND c.cohort_week + INTERVAL '31 days' THEN a.user_id END)
 / COUNT(DISTINCT c.user_id), 1) AS day_30_retention
FROM cohorts c
LEFT JOIN activity a ON c.user_id = a.user_id
GROUP BY c.cohort_week
ORDER BY c.cohort_week;

Replace core_action with your actual event name. The output is a table of cohort weeks, each showing the percentage of users still completing the core action at Day 7 and Day 30.

Step 3: Visualize Without Spending Money

Connect Metabase (open-source, free) to your Postgres or Supabase database and paste that query in; a cohort table renders in under an hour. For smaller datasets, a Google Sheets pivot table works fine. If you want pre-built cohort views without writing any SQL, FunnelKeeper's funnel dashboards surface the same retention picture across signup, activation, and return visits in one place.

Step 4: Review on a Fixed Cadence

Cohort data is only useful if you look at it consistently. A 15-minute weekly review comparing your two most recent cohorts against the prior four-week average is enough to catch a retention shift before it compounds. Set a recurring calendar block and treat it like a standing meeting.

How to Read a Retention Curve and Know What It Is Telling You

Once your cohort table is built, you need to know what the shape of the data is actually telling you. Three curve patterns show up repeatedly, and each one points to a different problem or opportunity.

Declining to zero means every cohort eventually churns completely. No stable base forms. This is a product-market fit problem, not a marketing problem, and scaling acquisition before fixing it will only accelerate your losses.

Flattening curve means retention drops sharply in the first few days, then levels off into a stable retained base. This is the healthy signal. Your job here is to lift that floor, not to panic about the early drop.

Smile curve means retention dips, then recovers later. This usually points to something seasonal or a specific feature that pulls users back. When you see it, investigate what triggered the recovery before assuming it will repeat.

Where You Stand Against the Industry

Mobile app retention data shows the average Day 30 retention across all app categories sits at 5.7%. A vibe-coded app hitting 10-15% at Day 30 is genuinely outperforming the industry median and is in a solid position to scale acquisition. Below 3%, the leaky bucket problem is acute; adding more users at that point accelerates losses, it does not compound growth.

Cohort-Over-Cohort Improvement Matters More Than Any Benchmark

If your Week 12 cohort retains better at Day 30 than your Week 8 cohort, your product is improving. That directional trend is the most important signal in your entire analytics setup, especially in the first six months. Moving from 3% to 8% over three months is evidence of a product that is learning. Sitting at 12% with no improvement across cohorts is actually a weaker position than it looks.

Channel Quality Shows Up in Retention Curves

Split your retention curve by acquisition source. If organic search users retain at 14% Day 30 but paid social users retain at 2%, the paid social spend is generating acquisition noise that makes your overall retention look worse than your product actually is. Knowing which channels produce retained users versus one-time visitors is one of the highest-leverage decisions in your funnel. This is exactly where conversion optimization tools organized by funnel stage become useful for matching the right growth levers to the right problem.

The Most Commonly Missed Red Flag

Watch for strong Day 1 retention followed by a sharp drop at Day 7. This pattern is easy to misread as a feature problem, but it almost always points to onboarding. The first experience is compelling enough to bring users back the next day. Something in the following days fails to deliver ongoing value or a clear reason to return. That is an onboarding sequencing problem, and it is fixable without rebuilding your core product.

What to Do After the Cohort Diagnosis: Turning Retention Data Into Action

Once you know what your retention curve is telling you, the next step is acting on it. The diagnosis points to one of four problems, and each has a different fix.

If your curve declines to zero, stop spending on acquisition. More users into a product that retains nobody does not build a business; it builds a bigger bill. Instead, pull a behavior-based cohort of the small slice of users who did return and talk to them directly. Find out what job they were hiring your app to do. Those conversations are worth more than any dashboard at this stage.

If Day 1 retention is low despite strong sign-up numbers, the problem is activation, not the product itself. Users are arriving but not reaching the moment where your app delivers its core value. The fix is collapsing the time between sign-up and that value moment as aggressively as possible. Cut onboarding steps, remove anything that delays the core action, and get users to a win in the first session. Then A/B test the revised flow and read the results cohort-by-cohort. A week-over-week improvement in Day 1 retention is the signal you are moving in the right direction.

If Day 7 is your drop-off point, users understood the product but lost the habit before it formed. A re-engagement trigger fired at Day 5 or Day 6 can interrupt that drift. The trigger needs to be specific to the core action, not a generic check-in. "You have three tasks due this week" works; "We miss you" does not. That specificity is what makes the message feel useful rather than needy.

If acquisition cohorts diverge sharply by channel, the answer is almost never to scale the highest-volume channel. It is to scale the highest-retention channel, even if it produces fewer raw sign-ups. Quality cohorts compound over time; low-retention cohorts keep refilling the leaky bucket and distort every metric you track. Retention and renewal is an ongoing tracking requirement, not a one-time read.

If running these cohort views and attribution comparisons manually sounds like a lot to maintain, FunnelKeeper is built for exactly this. It surfaces sign-up, activation, and retention data in one place, with attribution signals included, so founders working on SaaS or vibe-coded apps can see the full funnel without writing SQL or stitching together separate dashboards.

The Only Growth Metric That Actually Matters

Once you've diagnosed your retention curve and identified where users are dropping off, one truth becomes hard to ignore: sign-up count was never the number you should have been watching.

The only metric that predicts sustainable growth is retained users completing the core action. Not total sign-ups. Not total downloads. The percentage of each cohort that comes back and does the thing your app exists to do. Cohort analysis is the tool that makes that number visible, and without it, you are flying on a dashboard that shows progress while the product stalls.

Here is what to do this week:

Instrument three events. Log a signup event with a timestamp and acquisition source, a core action event, and a session start. These three data points are the minimum foundation for any cohort analysis. Without them, no tool or query will give you meaningful results.
Run the time-based cohort query. Pull your existing signup data, group users by the week they joined, and calculate what percentage of each cohort was still active at Day 7 and Day 30. The industry average sits at 5.7% retention by Day 30. That is your first benchmark for comparison.
Read the curve shape. Is it flattening into a stable base, or declining toward zero? That shape tells you whether you have a retention foundation to build on or a product-market fit problem to solve first.

One important mindset shift: beating 5.7% by next week is not the goal. Moving from 3% to 8% Day 30 retention over three months is a real signal that the product is improving. Directional progress across successive cohorts matters far more than hitting an absolute number on the first attempt.

This is where vibe-coded apps have a genuine structural advantage. Because you ship fast, you can also run retention experiments fast. Each new feature, onboarding change, or re-engagement trigger produces a new cohort. Cohort analysis connects each shipping decision directly to a retention outcome. That feedback loop, run consistently, is what turns a fast-shipping habit into compounding growth rather than a leaky bucket that resets every week.

Conclusion

Most vibe-coded apps die not from lack of features, but from lack of visibility. Sign-up numbers lie. Retention curves tell the truth.

The core takeaways are simple: sign-up metrics mask churn; cohort analysis reveals whether real growth is happening; retention curve shape diagnoses your product-market fit; and directional improvement across cohorts matters more than any single benchmark.

Your immediate action is straightforward. Instrument three events today: signup with source, core action, and session start. Run your first weekly cohort query this week. Read the curve. Then ship one targeted change and measure whether the next cohort improves.

You already have the speed advantage. Now pair it with the right feedback loop. Fast shipping plus cohort-informed decisions is the combination that builds compounding growth. Start measuring, keep iterating, and turn that leaky bucket into a foundation worth scaling.

Vibe-Coded App Analytics: Is Your App Growing or Just Filling a Leaky Bucket?