The web2app guide: how subscription apps grow by selling on the web first

A step-by-step guide to launching a web funnel, from the first ad to the first renewal, with data from real funnels.

Section 1

What web2app is

In the standard model, someone sees an ad, taps it, lands on the app’s store page and installs. If they subscribe, they pay inside the app, and Apple or Google keeps 15–30% of the payment.

Web2app changes the order:

• the ad sends people to a website first,
• there they answer a few questions about what they want and what’s getting in the way,
• they see a plan built from their answers and pay right in the browser,
• then they download the app and log in, and everything they paid for is already unlocked.

This sequence is called a funnel: a series of screens that takes someone from an ad to a purchase. A web2app funnel runs on the web and ends in your app.

Web2app is also written as web-to-app. In this guide, “web2app funnel”, “web funnel” and “funnel” mean the same thing.

A web2app funnel, step by step · Rise

Ad

→

Landing page

→

Quiz

→

Personal plan

→

Paywall

Subscription apps sell in four main ways. What sets them apart is where people pay.

The web part can look different. Teams use landing pages, advertorials (pages styled as an article or a review), video sales letters (one long pitch video) and quizzes. For subscription apps, quizzes are the most common, so that’s the format this guide covers: a few questions that end with a personal plan and an offer.

Web funnels are now common practice in every app category. They’re most widespread in:

• health and fitness,
• education,
• lifestyle,
• photo and video,
• finance and career.

Still, most US subscription apps don’t run web funnels yet and acquire users in other ways.

This guide is written for two kinds of readers:

1. founders and small teams that run their own user acquisition,
2. and growth teams at larger companies that are evaluating web2app.

By the end of the guide, you’ll know how to:

• explain how web2app works and why the biggest apps sell this way,
• plan an ad budget and run a first campaign on Meta, then add TikTok and Google,
• build a funnel from the first screen to the paywall,
• set up payments, app access and tracking so every purchase is counted,
• and read the results and decide what to test next.

Section 2

Web vs App Store

Selling on the web and selling through the App Store differ in six ways:

The payout delay matters once you buy ads. Ad platforms charge as the ads run, while App Store revenue from the subscribers those ads brought in arrives up to two months later. The team covers that gap from its own cash, which limits how fast it can scale ad spend.

The web brings more buyers and revenue

Web funnels also show better results (State of Web2App 2026):

• Conversion: with the same product and ad spend, a web funnel turns 3.0% of ad clicks into buyers, an in-app funnel 1.5%.
• LTV (lifetime value), the total a subscriber pays before they leave: in health and fitness apps, about $80–90 for web subscribers against $40–45 for app subscribers. That covers the higher cost of web traffic.
• Revenue mix: funnel scalers, apps that run many funnels at once, get about 90% of their subscription revenue from the web.

App Store and Google Play allow web2app

Apple’s App Review Guidelines allow people to buy a subscription on your website and then sign in to the app (section 3.1.3(b), Multiplatform Services). Google Play’s policies allow the same.

For years, what both stores restricted was app-to-web: a button inside the app that sends people to a web checkout. After court cases brought by Epic Games, US apps can now do this on iOS and Android. What the stores may charge on those sales is still being decided. None of this affects web2app, since people pay before they install.

The rules that do apply are the ones every subscription business follows: show the price and renewal terms before people pay, and make cancelling easy to find.

Apple also expects the same subscriptions to be on sale inside the app, so keep your in-app paywall for people who install directly.

Web2app comes with its own setup

Four things set the web2app flow apart:

1. More steps before the app. In the store model, people tap Install. In web2app, they first go through a quiz and pay on a web page. Some drop off along the way, so the funnel has to build trust.
2. You take over the store’s job. In the store model, the store handles payments, renewals, refunds, disputes, sales tax and cancellations. On the web they’re yours, along with the support questions that come with them.
3. It costs money before it pays back. You need a tool or a team to build funnels, plus an ad budget. Ads usually take a few months to pay back.
4. Fewer people reach the store for free. If everyone pays before they install, people who don’t pay never download the app. That means fewer installs and ratings for store search to work with, and less word of mouth.

Many teams run both: web funnels for paid traffic, and the regular store path with in-app purchases for everyone who finds the app on their own.

Section 3

Paid acquisition

Web funnels run mainly on paid traffic. Organic posts and creator videos can send people to a funnel too, but most funnel volume comes from paid user acquisition (UA): ads on Meta, TikTok and Google.

Organic reach depends on the platform’s algorithm: one video takes off, the next ten don’t. Paid acquisition gives you volume you can plan, scale up and switch off.

A web funnel also makes paid traffic more efficient. It sells on the first visit and reports each purchase to the ad platform, so the platform can find more people like the buyers.

Three metrics decide what a customer costs

Paid acquisition is profitable when a customer costs less to acquire than they pay you. Two metrics measure this:

1. CAC (customer acquisition cost): what it costs to get a paying customer, or your ad spend divided by the new paying customers it brought in,
2. and LTV (lifetime value): what that customer pays you over time.

CAC isn’t a number you set. It comes out of three others:

1. CPM: the cost of showing your ad 1,000 times. The ad auction sets it, and it depends on the country, the season and how many advertisers want the same people.
2. CTR (click-through rate): the share of people who see the ad and tap it. The creative drives this one.
3. CR (conversion rate): the share of people who tap and then buy. Your funnel drives this one.

Put together, CAC = CPM ÷ (1,000 × CTR × CR). Say your ads cost $20 per 1,000 views, 1% of viewers tap, and 3% of those buy. You pay $2 per click and about $67 per customer.

Because the three numbers multiply, small gains stack:

• Improve each one by 10% and CAC drops to about $50.
• Improve each by 20% and it drops to about $37, close to half.

For this reason, teams that scale work on the ads and the funnel together: a better creative, a better quiz and a better price all lower CAC.

Ads and product are measured by one shared metric: ROAS (return on ad spend). It’s the revenue from the customers an ad brought in, divided by what the ad cost. At 100% ROAS, you’ve earned back what you spent.

Ad spend usually pays back in 3–6 months

Paid acquisition requires spend up front. The first months go to learning what works, and profit comes after that. For a subscription app running its first web2app campaigns on Meta, use these numbers as a rough guide:

Payback here means the point where revenue from a group of customers covers what it cost to acquire them. On the way there, first tests on Meta with US traffic usually land at a CPA (cost per purchase) between $40 and $120.

As a rough reference for first tests:

• Under $40: a strong result for a new channel.
• $40–70: a typical target once the funnel and creatives work.
• $70–120: common while campaigns are learning; watch what those buyers pay in their first 30 and 60 days.
• Above $120: works only with high LTV, a large share of annual plans or a longer payback window.

Planning by number of purchases is more precise, since each milestone answers a different question. On Meta, at a CPA of $80–100:

Those figures cover only the spend behind the purchases. While campaigns warm up and creatives get tested, real spend tends to run 1.5–2 times higher.

Web funnels give ad campaigns better data than campaigns that send people straight to the store: the algorithm gets more signals, and every purchase is attributed to the ad that brought it. Campaigns can also optimize for the exact event you need, such as an upfront purchase.

Five rules for the first campaign

1. Optimize for purchases from day one. Ad platforms find more of whatever event you ask them for. Ask for installs or sign-ups, and the platform finds people who install or sign up, many of whom never pay. Ask for purchases, and it looks for people who buy.
2. Start with what already works: your best TikTok videos, the audience that converts in your app and your strongest market. The first campaign tests the new channel, so it should get the strongest inputs.
3. Keep to one market. Mixing the US with much cheaper countries in one campaign distorts every metric. Prove the funnel in one place, then expand.
4. Wait for 50 purchases before big decisions, and change one thing at a time. At 25 you can spot something broken. Before 50, results fluctuate too much to read. And if you change the creative, the audience and the offer at once, you won’t know which change did what.
5. Watch every step of the funnel. Track the ad click, the first screen, sign-up, paywall and purchase separately. CPA can look fine while one step in the middle is broken.

How to count CAC correctly

CAC decides how much you can spend on ads and when to scale. If it’s counted too low, paid ads look profitable when they aren’t, and the budget grows on numbers that don’t hold. Watch for three common counting mistakes:

1. Mixing cold and warm traffic. Cold traffic is people who’ve never heard of you. Warm traffic already knows you: they search for your brand, see a retargeting ad or open your email. Warm traffic converts cheaply, but its volume is limited, and your TikTok or cold ads created that awareness in the first place. Judge cold campaigns on their own CAC.
2. Using blended CAC. Blended CAC divides total ad spend by all new subscribers, organic ones included, which hides what paid ads really cost. Divide paid spend by the customers paid ads brought in.
3. Counting by calendar month. Subscribers who joined in January keep paying in February, March and later. A cohort view fixes this. It groups people by the month they joined and follows what each group pays over time, so you compare each month’s ad spend with the revenue it brought in.

Most teams add channels in this order:

1. Meta (Facebook and Instagram),
2. TikTok, once the funnel converts on Meta,
3. and Google Search.

Meta comes first because it needs less data to optimize. TikTok experience still applies: the videos that won organically become your first ads.

Start with Meta

Every ad platform needs a minimum number of purchases a week to optimize delivery. Until then, the campaign is in the learning phase, and results vary widely. Meta gets there with less data than TikTok, so it costs less to start:

• Meta: $300–400 a day for a first campaign.
• TikTok: about seven times your cost per purchase every day in each ad group. At a $120 CPA, that’s around $860 a day per group.

Meta also finds buyers through broad targeting. You don’t pick interests: the ad itself decides who it reaches.

First Meta campaign: the settings

In the first two weeks, walk through the funnel step by step and compare each step with the typical ranges in the Experiments section. Fix the first step that falls short before you judge anything after it, since later steps can’t be read until earlier ones work.

One pricing choice matters most at this stage: a paid intro offer, say $0.99 or $2.99 for the first week, in place of a free trial. A real charge, even a dollar, teaches Meta to look for people who actually pay. A free trial teaches it to find people who tap “Try for free”. The pricing section covers offers in detail.

Add TikTok once Meta works

TikTok isn’t a required channel. Its audience is younger and more sensitive to price than Meta’s, so most funnel steps convert a little lower there. Whether to add it depends on your product and who buys it: for an older audience, Meta and Google may be enough.

Launch TikTok with the funnel, offer and market that already win on Meta. If results are weak, the cause is the channel. For the first two weeks, while TikTok’s pixel (the tracking code on your funnel) learns, expect CPA to run 1.2–1.5 times your Meta CPA.

The biggest difference is creative. TikTok wants native, creator-style video, and a winning ad usually lasts 7–21 days. Once you scale, plan for 10–20 new videos a week. If your team already posts on TikTok, you already know how to make these videos.

Add Google Search for people who are already looking

On Meta and TikTok, the ad creates the need. On Google Search, people arrive with the need already typed into the search bar. That changes the funnel more than the campaign: the first screen has to repeat what they searched for, word for word if possible.

Set up Search in this order:

1. Start with Search campaigns on three kinds of keywords: your brand name, competitors’ names and the problem people search for.
2. Optimize for purchases with Maximize Conversions and no target.
3. Budget at least your expected CAC per day per campaign, ideally 1.5 times that.
4. After 30 purchases, add a Target CPA based on your real CAC. Change targets and budgets by no more than 20–30% every 3–5 days.
5. Once Search is stable, add Demand Gen (YouTube including Shorts, Discover and Gmail) and Performance Max. They take your best headlines to people who aren’t searching yet.

Add other channels once the funnel works

• AppLovin and other in-app ad networks. They show your ads inside other apps and can reach large volume. They also need a lot of creatives: some apps reach stable results at around 50 new ones a week. Plan for longer videos and playable end cards, the interactive screen that plays after the video.
• Pinterest, Snapchat and X. They work like Meta, with smaller, more niche audiences.
• Content ads such as Taboola and Outbrain, which place article-style ads on news sites. They work for advertorial funnels.
• Affiliate and CPA networks, where partners get paid per purchase. These work on the web because each purchase can be traced to the partner who sent it.
• Organic and creators. Organic TikTok and creator videos keep supplying tested hooks for ads and bring people who later search for your brand.

How to scale a working campaign

Once 3–5 creatives hold your target CPA, there are four ways to scale:

1. Raise the budget by 20–30% every 2–3 days on Meta, or 15–20% every 2–3 days on TikTok, while CPA holds. Bigger jumps send campaigns back into learning.
2. Gather the winners into one campaign, so the budget goes only to proven ads.
3. Add countries, first other English-speaking ones such as the UK, Canada and Australia, then new languages with a fully translated funnel and native-speaking creators.
4. Add creatives. On TikTok, every extra $1,000 a day of spend needs roughly 5–10 more new videos a week.

Make one of these moves per week. If CPA jumps more than 20% after a change, stop and let it settle for 2–3 days.

Plan around the calendar too. Ad prices rise sharply from late October to December, so avoid launching a new channel in late November.

Section 4

Web2app funnel structure and build options

From the outside, a web funnel looks like a quiz with a payment page at the end. Underneath, it’s a chain of six parts that have to work together, starting with the ad. Most of the work in running web2app goes into the parts people never see.

A funnel has six parts

1. The ad creative. The first screen of the funnel in practice: it makes the promise every screen after it repeats, and it decides who taps at all.
2. The funnel itself. The quiz or onboarding screens, the email step, the paywall with your plans and any offer after the purchase. It also has to work in every language and currency you sell in.
3. Payments. The checkout and payment processing, plus renewals, sales tax, refunds and disputes.
4. App access. The plan someone paid for on the web has to be available when they download the app and log in.
5. Tracking. Pixels and server events that report key events, such as sign-ups and purchases, back to Meta, TikTok and Google, plus UTM tags on every ad link.
6. Analytics. A product analytics tool such as Amplitude or Mixpanel shows what happens at each step. Revenue analytics shows what each group of customers pays over time.

Around these sit jobs that the store handles when people pay in the app. On the web, they’re yours:

• support for billing questions,
• refund requests,
• and receipts and emails about billing problems.

Two ways to build a funnel

Section 5

Creatives

The ad is the first screen of the funnel. It makes the promise that every screen after it has to keep, and it decides who taps at all.

Your best organic videos can become your first ads

With broad targeting, the ad itself decides who sees it: the hook, the person on screen and the promise attract a specific kind of viewer. So you need several creatives, each speaking to a different need. If you already have organic videos that perform well, use them to pull hooks, angles and messages for your first ad creatives.

Pick the 5–6 that performed best and re-cut them as ads:

• put the main message in the first 1–2 seconds,
• keep them vertical (9:16) and 15–35 seconds long,
• make sure someone speaks: a voiceover or a creator on camera,
• and end with a clear call to action.

For the first tests, plan at least six angles with three variations each:

• an angle is the core message: a pain, a promise or a situation,
• a variation changes the hook, the person or the format.

A new caption or filter on the same video doesn’t count as a new creative. It reaches the same people and wears out at the same time as the original.

Formats that tend to work for quiz funnels:

• a creator talking to camera
• POV or storytime
• green-screen reaction
• “I took this quiz” screen walkthrough

When you need more creatives than your team can shoot:

• Creators you already work with: ask them to record the same script in their own words. Testing new faces matters as much as testing new scripts.
• TikTok One: TikTok’s creator marketplace. It replaced Creator Marketplace and Creative Challenge in 2025. It’s paid, and the more you spend, the more creators you can reach.
• Spark Ads: a TikTok format that runs a creator’s post as an ad from their own account. It looks like their content and keeps its likes and comments.
• AI-generated video: TikTok’s own tools can make it, but it gets an AI label, which usually makes it perform worse than real creators.

Where to look for new angles:

• competitors’ ads in the Meta Ad Library and TikTok Creative Center’s Top Ads,
• the words customers use in your app reviews and support tickets,
• and the quiz answers that convert best.

Copywriting frameworks such as Life Force 8, a list of eight basic human desires, help turn those findings into hooks.

Test new creatives in a separate campaign

Creatives wear out, so new ones are needed every week. Test them in a separate campaign with $100–200 a day and the same targeting and optimization as your main one. Add about ten new creatives a week. This keeps untested ads out of the main campaign’s learning.

Read each creative from the top of the funnel down, because early signals arrive long before purchases do:

What to do with the results:

• Winners: a creative that hits your target CPA over 5–10 purchases moves to the main campaign.
• Creatives with no spend: leave them alone. The platform didn’t select them for delivery, which says nothing about their quality.
• Creative fatigue: when a proven winner’s CPA starts climbing after a few days or weeks of stable results, usually with falling CTR, the creative is wearing out. Replace it with new hooks on the same message. Don’t raise its budget.

Section 6

Onboarding

Onboarding is everything between the ad and the paywall:

• the quiz,
• the short screens in between,
• and the result screen with a personal plan, shown right before the paywall.

By the time the price appears, the person should see the plan as built for them. The price then reads as the cost of that plan.

Focus on the screens before the paywall

In an average web2app funnel, 13% of people who start reach the paywall and 3% buy (State of Web2App 2026). Most people leave before the price is even on screen, so the screens before it matter more than the paywall itself. Onboarding has three jobs:

1. Continue the story the ad started, so people feel they’re in the right place.
2. Build trust in a product they met half a minute ago.
3. Set expectations they’ll still agree with after paying. A promise the product can’t keep comes back as a refund.

The usual order of screens

A quiz funnel usually runs in this order:

1. First screen that picks up the ad’s promise and starts the quiz.
2. Quiz about goals, problems and situation, with short screens of feedback, facts and reviews in between.
3. Email, asked in exchange for the result.
4. Loading screen while the plan “builds”.
5. Result: the personal plan and a prediction.
6. Paywall with plans and prices.
7. Upsell right after the purchase.
8. Download page that gets people into the app.

Tell one story from the ad to the paywall

A common storytelling framework breaks the path from the ad to the paywall into eight steps, and each screen of the funnel serves one of them:

1. Engage: an easy first step tied to the ad.
2. Build trust: facts, numbers and reviews.
3. Understand goals: what the person wants.
4. Understand pains and context: what’s in the way and what their days look like.
5. Give feedback: react to their answers.
6. Show the expected result: a prediction built from what they told you.
7. Give a sneak peek of the solution, without revealing all of it.
8. Offer a personal plan for the result they’ve just seen.

One promise from the ad to the paywall · Calm

Ad

“Work stress keeping you up?”

→

First screen

“less stress and better sleep”

→

Quiz

“improve your sleep and reduce stress”

→

Result

“Stress & burnout impact today”

→

Paywall

“reduce tension and feel more rested”

Calm’s funnel shows how this works. The ad asks whether work stress keeps you up at night. The first screen promises less stress and better sleep, the quiz asks questions about both, the result rates the person’s stress and burnout, and the paywall promises to reduce tension and help them feel more rested. One promise travels from the ad to the price, and every screen makes it more specific.

Repeat the ad’s promise on the first screen

Ad and first screen · Homemade Method

→

On the first screen, people decide whether the funnel matches the ad. The first screen needs:

• one headline that repeats the ad’s promise,
• one visual,
• one button,
• and a load time under about 2.5 seconds on a phone, inside the Instagram or TikTok browser.

A polished corporate page after a casual creator video breaks that match, and people leave.

Many funnels put the first question on the first screen itself, so engaging takes one tap. That question should:

• take one tap on a picture or a large button,
• continue the ad’s hook,
• and ask nothing personal.

Age, weight, income and other sensitive questions belong in the second half, after people have invested a few answers. If you lose more than half of people between the first and second question, the first question is usually the problem.

Write questions people recognize themselves in

Questions people recognize themselves in · Endel

Each question should make the person recognize their own situation. Together, the answers make the plan feel personal. Ask about:

• the goal,
• the main problem,
• the situation,
• and the motivation.

One format works well for this: a short statement people rate, such as “I need a little push to keep going when I’m tired. Agree?” People see themselves in it faster than in a direct question.

Ask only what you use. Every answer should change something later: the plan, a feedback screen or the paywall copy. Merge questions that ask about the same thing. Before the email step, use taps and sliders only, with no typing. The questions can also differ by audience: a funnel for women over 45 can ask, show and promise different things from one for men in their 30s.

Respond to answers between questions

Screens between questions · Fastic

Feedback

Reacts to the answers right away

Proof

Shows why the app works

Sneak peek

Shows the expected result before the price

Between questions, the funnel responds to the answers with three kinds of screens:

1. Feedback on answers. React to what someone said with a result or a tip, either right under the answer or on its own screen: “Your level is B1. That’s enough for everyday conversation; let’s get you to B2.” It makes the quiz feel useful.
2. Proof. The number of people using the app, ratings, research behind the method, press mentions and reviews from people who look like the person taking the quiz.
3. A sneak peek of the solution. A prediction chart with a date, a short explanation of how the method works, or a chance to try it: a selfie analysis in a style app, a palm photo in an astrology app, a short level test in a language app. Show enough to make the result real and keep the full result for after the purchase.

Ask for the email before the result

Email screens from real funnels

Babbel

A plain question, with the email terms below

Lasta

The email as the way to get the plan

Endel

The email as the way to see the results

Gyna

The plan, social proof and a line about privacy

Most people who reach your paywall won’t buy on the first visit. An email lets you follow up with them, and the emails that go out after someone leaves can bring back a meaningful share of revenue. Asking for it also adds a small commitment before the price, which tends to raise conversion.

Frame it as the way to get the result: “Enter your email to get your plan.” Then:

• put it right after a teaser of the result,
• add a line saying you won’t sell the address,
• and offer sign-in with Apple or Google so people don’t mistype it.

Healthy funnels keep 70–85% of people at this step. If you lose more, test moving the email one or two screens later.

Make the most of the loading screen

Loading screens from real funnels

Rise

Names the answers the plan is built on

Clever

Breaks the result into parts and adds a review

Reframe

Lists the parts of the plan and adds a trust badge

Gyna

Shows what is being analyzed

Flirtist

Shows the plan’s progress and adds a review

The “analyzing your answers” screen before the result is a short pause that shows the plan is built from the person’s answers. It also gives you a few more seconds to show something before the price. Use it to:

• show what’s being built from the person’s answers,
• tell the product’s story in a few short slides,
• or ask one or two last questions in pop-ups.

Keep it short for colder traffic: five seconds or less on TikTok. A longer loader makes sense only if every slide adds something new.

Build the result from the person’s answers

Result screens · Lasta

Goal and date

The goal from the quiz, with a date and weekly steps

Now vs later

Now vs after 4 weeks, marked as illustrative

Profile and plan

The person’s numbers and what the plan includes

The result screen turns the answers into a promise: the goal they chose, a date and how they’ll feel when they get there. Two elements make it personal: the line “Based on your answers” and a before-and-after of the person’s own numbers, such as where they are now and where they’ll be in three months. Keep the promise realistic and add “results vary” where it applies. Google requires it in some categories, and an inflated promise comes back as refunds.

Four mistakes that lower conversion

1. Making the quiz shorter or longer for its own sake. What matters is whether every question and screen is relevant to this audience. The length follows from that.
2. Optimizing for reaching the paywall. More people on the paywall can mean fewer buyers, if the screens you cut were the ones that sold the plan. Judge every change by purchases.
3. Polishing the visuals and leaving the message as is. Words and meaning move conversion more than design.
4. Removing every bit of effort. Steps that ask something of people, like an email, a selfie or a photo of their palm, raise their commitment, and with it conversion.

Section 7

Pricing and the paywall

On the web, you set the prices, the plans, the trial and what happens after the purchase. Pricing control is one of the main reasons to sell outside the App Store. A single pricing change can shift a large share of a funnel’s revenue.

Research willingness to pay first

People pay when the value they expect is higher than the price. Value here is what the product does for them, plus the extras that come with it: the community, the feeling of a plan made for them. Before you test prices, find out what your audience would pay:

1. Run a short pricing survey. The Van Westendorp method asks four questions: at what price would the product feel too expensive, expensive but worth considering, a bargain or so cheap you’d doubt its quality. The overlap shows an acceptable range.
2. Rank features. A MaxDiff survey asks people to pick the most and least important feature from small sets, which shows what to put first on the paywall.
3. Talk to people who left. A short exit interview with people who cancelled tells you whether price was the reason or something else.

Continue the quiz on the paywall

Paywalls from real funnels

Fastic

1 week – 1 month – 3 months, with the 1-month plan selected

Lasta

28 days – 12 weeks – 24 weeks, with the shortest plan selected

Reframe

Monthly – quarterly – semiannual, with the quarterly plan selected

By the time people see prices, they’ve spent a few minutes in your quiz. The paywall should keep the angle, story and message that ran from the ad through the onboarding. A strong one usually has:

1. A headline built from their answers, such as “Your 8-week plan to lose 10 lb”.
2. The expected result again: the chart, the date, the before and after of their own numbers.
3. The plans, with one selected, and the price visible without scrolling.
4. The subscription terms next to the pay button: today’s price, the price after the intro period, how often it bills, that it renews automatically and how to cancel.
5. A short list of what’s included.
6. Reviews from people who look like the person reading.
7. A money-back guarantee, if you offer one.
8. Security and contact details: who charges the card, how payment is protected, how to reach you.
9. A short FAQ below the plans.

Urgency works too, as long as it’s real. A countdown has to count to a real deadline that stays the same on every visit, and a crossed-out price has to be one people do pay.

Show three plans with one selected by default

Three plans is the most common setup on web paywalls, because it gives people a choice without making them think too long. Two configurations work well for most subscription apps:

1. 1-week paid trial – 1 month – 3 months. An easy start for people who aren’t sure, and longer plans for those who are.
2. 1 month – 3 months – 6 or 12 months. Nudges people toward a longer commitment, which brings more money up front.

Small details change how people choose:

• which plan is selected by default
• the order of the plans
• the names: “1 month” and “4 weeks” read differently
• price per day or per week next to the full price
• whether the full price and the saving are visible

Start with a paid entry price

A paid entry addresses both.
It can be a paid trial, such as $1.99 for the first week.
Or it can be an intro offer: a discount on the first period, before the regular price starts.
New funnels often start at $0.99 or $2.99.

Paid trial and intro offer

Paid trial · Flo

A 30-day trial for €1, with an option to pay more

Intro offer · Rise

$19.99 for the first month, then $29.99 a month

Large intro discounts lift conversion, but the bigger the gap between the first price and the regular one, the more people leave or ask for a refund at the first full charge. Show the renewal price and date clearly. Free trials still work as an offer in a follow-up email to someone who left the funnel without buying.

People who leave the paywall without buying can get an offer too. A discount can bring some of them back. It also teaches people that waiting gets a lower price, so show it only to those who were unlikely to buy at full price.

Disclose all terms clearly before people pay

Before the pay button, people should see five things:

1. the price they pay today,
2. the price after the intro period,
3. how often they’ll be charged,
4. that the plan renews automatically,
5. and how to cancel.

In the US, the FTC enforces this under the Restore Online Shoppers’ Confidence Act (ROSCA). The law requires clear terms, the customer’s express consent and a simple way to stop recurring charges. Google Ads checks the same disclosures on your paywall.

Three more practices reduce refunds and disputes:

1. don’t say “free” when the next step is a paid subscription,
2. show the same plans and prices on the web and in the app,
3. and make cancelling easy to find.

Offer upsells to raise LTV

Upsells and upgrades from real funnels

One-time upsell · Simple

A one-time add-on, charged to the card already on file

Second subscription · Noom

Premium on top of the plan, with a trial, its own terms and a consent checkbox

Partner products · Noom

Physical products from partners, added to the program

Upgrade · Simple

A switch to a 3-month plan, with the new total, terms and consent

There are two kinds of offers, and they work differently at checkout:

1. An upsell is something extra on top of the subscription: a PDF guide, extra content in the app, a coaching session, a physical product. A one-time upsell can be bought in one tap, since the card is already saved. If the upsell is a second subscription, show its terms and a checkout again.
2. An upgrade replaces the plan someone just bought with a longer, pricier one, for example 1 month to 3 months. It changes the subscription, so it always needs its own terms and checkout. Frame it around the longer commitment and the better chance of reaching the goal. A discount alone rarely converts.

The people who take upgrades are the most motivated subscribers. Moving them to a longer plan can make the retention of the original plan look worse. Judge upgrades by the revenue of everyone who bought, across all plans.

Upsells convert best right after the purchase and can raise LTV by up to 30%. The global median upsell price is about $9.31 (State of Web2App 2026). People on longer plans accept pricier upsells: in the US, 3-month subscribers bring the highest average upsell value.

Section 8

Payments

Selling on the web means lower fees, faster payouts and control over prices and offers. It also means the work the store used to do becomes yours:

• taking payments,
• renewing subscriptions,
• refunds and disputes,
• and sales tax.

Choose who sells: a payment provider or a merchant of record

There are two ways to take payments on the web. A payment provider (PSP) processes payments for you, and you remain the seller. A merchant of record resells your product under its own name and takes on taxes, refunds and disputes:

Small teams often start with a merchant of record to skip the tax work. As volume grows and fees start to matter, they move to their own payment providers.

Sales tax on digital subscriptions depends on where the buyer lives: each US state has its own rules, and Europe and the UK charge VAT. A merchant of record handles it for you. With your own payment provider, you need a tax tool, such as Stripe Tax, and a process to file the returns.

Make Apple Pay the default

Checkouts from real funnels

Reframe

Renewal terms right above Apple Pay, a trust line below

Kure

Today’s price apart from the regular one, every method visible

Endel

Today’s price vs the regular one, Apple Pay first

Dr. Kegel

The discount and total, then a large Apple Pay button

Flo

Price today, a trial reminder and terms above Apple Pay

Lead the checkout with Apple Pay:

1. Select Apple Pay by default on iPhones, and make Apple’s official Apple Pay button the main one.
2. Keep the other methods visible, so people who don’t use Apple Pay can pick a card or PayPal without searching.
3. Show today’s price apart from the regular price, and put the renewal terms right above the button.
4. Add trust signals: secure payment, encryption and the name that will appear on the bank statement.
5. Test inside Instagram and TikTok. Their in-app browsers sometimes hide Apple Pay and Google Pay. If people only see a card form there, expect fewer purchases.

Where the checkout sits is worth testing too. The four common placements:

1. a separate screen after the paywall
2. a pop-up over the paywall
3. built into the paywall itself
4. an Apple Pay or Google Pay button right under the plans

Whichever you choose, build the checkout into your funnel. In FunnelFox’s experience, a provider’s default checkout, such as Stripe’s or Paddle’s standard payment page, converts up to 30% worse than a custom checkout built into the funnel. There are two reasons:

• it looks different from the screens before it,
• and it sends people to a separate page, which adds a step and a new web address some of them won’t come back from.

Expect some payments to fail

Some payments always fail. That’s normal, and it usually happens for one of these reasons:

• the bank asks for an extra security check that the person doesn’t complete,
• the bank declines the card,
• the card has expired or has no money left,
• Apple Pay or Google Pay doesn’t show up,
• or a technical error happens between systems.

A failed payment isn’t always lost revenue. Retrying it at the right moments brings back up to 17.5% of failed subscriptions, and the first three attempts recover almost half of what can be recovered (State of Web2App 2026). Back the retries with:

• an email asking for a new card,
• and a few days of grace before access stops.

Add a second provider before you scale

A payment provider can freeze your account in a day, for example after a spike in disputes or a risk review. With one provider, your cash flow stops at once: you can’t take new payments, and the money already in the account is frozen, so you can’t withdraw it or reinvest it in user acquisition.

A second provider removes that risk. Payment orchestration tools connect several providers in one place. They send each payment to the provider most likely to approve it, retry declined payments through another one and keep renewals running if a provider drops out.

You don’t need this on day one. Most teams start with one provider, then add a second one before they scale spend.

Make refunds easy to avoid disputes

A refund happens when a customer asks you for their money back and you return it.
A dispute, or chargeback, happens when the customer asks their bank. The bank pulls the money back from you, usually adds a fee and asks for evidence if you want to contest it.

Card networks track the share of your payments that end in disputes. Above their limits, you face monitoring programs, fines and, eventually, a closed merchant account.

Most disputes come from people who don’t recognize the charge, forgot about the renewal or couldn’t find how to cancel. That makes them preventable:

1. Use a clear statement descriptor, the name on the bank statement. It should match your app’s name, so people recognize it.
2. Send a reminder before renewals, especially before the first full-price charge after an intro offer.
3. Make cancelling and refunds easy to find. A refund you give costs you the payment. A dispute costs the payment, a fee and a mark on your record.
4. Use dispute alerts. Services such as Ethoca and Verifi notify you when a customer contacts their bank, so you can refund before the dispute is filed.

Build your own cancel flow

Cancel flows from real web subscriptions

Blinkist

Cancel in the account settings

→

A reason in one tap

→

50% off, with a cancel link below

→

Confirmation

nibble

Confirm or keep

→

Feedback

→

A discount, or skip

→

Cancel or keep the plan

→

Confirmation

On the web, you can set up your own cancel flow. Instead of losing a subscriber outright, you can offer an alternative:

• a pause,
• a cheaper plan or a downgrade,
• a discount on the current plan,
• a different plan length,
• or free extra time.

This helps you keep churn under control. Show the alternatives on one screen together with the full cancellation, and make every button equally visible. People should be able to cancel as easily as they can accept an offer.

If someone still chooses to cancel, ask why in one tap: too expensive, not using it enough, not seeing the value. The answers help you plan your product and pricing, and choose the right offers for win-back emails.

Section 9

Access in the app

After paying on the web, people download the app. If the app doesn’t know they paid, they see the paywall again. That leads to a support ticket, a second purchase in the app or a dispute with their bank. This step has to work for every buyer.

Guide buyers into the app from the download page

The page after the purchase decides whether people make it into the app. It should:

1. Confirm the purchase and what’s included.
2. Give the next steps in order: download, open, log in.
3. Show the right store button for the person’s phone, App Store or Google Play.
4. Remind them which email their account is on.
5. Offer help with a “Trouble logging in?” link to support.

Send the same instructions by email, since many people install later or on another device. And watch how many buyers never open the app. People who paid but never used the product are the ones most likely to ask for a refund or go to their bank.

Link the purchase to the person in the app

1. Who the person is. The app needs an identifier that matches the purchase. It can be a user ID passed from the funnel through a link, the email from the funnel or a sign-in with Apple or Google. A full account isn’t required: every extra step in the funnel loses some people.
2. What they paid for. The app needs to know which plan they bought and until when. The funnel sends each purchase to the place your app checks for access: a subscription platform such as Adapty, RevenueCat, Qonversion or Superwall, or your own backend. Renewals, cancellations and refunds have to update it too, so access ends when the payments do.

There are four ways to do this:

Most teams combine two: the automatic link as the main path and a login link or code by email as the fallback for anyone it doesn’t reach.

To make the hand-off from web to app work:

• Pass the context, not just the login. The link can carry the plan, the goal from the quiz and the language, so the app opens on the right screen and skips the onboarding the person already finished on the web.
• Keep personal data out of the link. Pass a short-lived, single-use token, and have the app exchange it for a session.
• Plan a fallback. If the link doesn’t trigger, restore the person’s data from your backend on the first app launch.
• Test on a fresh install, with no previous app data, before you launch.

Keep web and in-app subscriptions in one place

If your app keeps its in-app paywall, a person who already paid on the web can still see it and pay again in the app.

To prevent that, keep all subscriptions, from the web and from the store, in one place: your subscription platform or your own backend. The app checks that one place, so a web buyer sees their plan instead of the paywall.

Section 10

Tracking and analytics

Tracking does two jobs:

1. it tells ad platforms who bought, so their algorithms find more buyers,
2. and it shows you what happens at every step and what each group of customers pays over time, so you know what to fix and where to spend.

Send purchases back to ad platforms

Ad platforms find new buyers by learning from the ones you already have. So every purchase in your funnel has to be reported back to the platform that brought the buyer.

This usually starts with a pixel: a small piece of code on your funnel pages that sends events, such as a purchase, to Meta, TikTok or Google. Browsers and ad blockers stop some of these reports, so the best practice is to send the same events from your server too, through Meta’s Conversions API, TikTok’s Events API or Google’s enhanced conversions.

The main events to send:

Funnel platforms can send these events for you. If you build the funnel yourself, your developers set them up. Either way, check these details, or purchases get lost or double-counted in your reports:

• Count each purchase once. When both the pixel and your server report the same purchase, give it the same ID in both. The platform then sees one purchase, not two.
• Don’t lose the click ID. When someone taps an ad, the platform adds a code to the link: fbclid on Meta, ttclid on TikTok, gclid on Google. That code is how the platform knows the purchase came from its ad. It has to stay with the person on every screen of the funnel. If it gets lost, the purchase won’t show up in the platform’s reports.
• Report every purchase, whatever the source. A buyer who came from TikTok or an email still helps Meta learn what a buyer looks like.
• Label every ad link. Add UTM tags: short labels at the end of the link that tell your analytics which network, campaign and ad a visitor came from. Add the IDs too, since names change and IDs don’t.
• Trust your own numbers first. Platforms also count purchases from people who only saw an ad and didn’t click, so their reports usually show more sales than your payment data. Compare the two.

Track where people drop off

Ad platforms see a few events. Your own analytics should see every step. Tools like Amplitude, Mixpanel, Google Analytics or PostHog work for this, as does an in-house dashboard if you have a data team. Send them these events:

Save user properties

Events tell you what happened. User properties are details attached to each person. They tell you who did it and let you compare one group with another. Save these for everyone who enters the funnel:

1. One user ID that’s the same on the web and in the app, so you can follow a person from the ad to their renewals.
2. Email.
3. Where they came from: UTM parameters plus the campaign, ad set and ad IDs.
4. Their answer to each quiz screen, so you can see which answers lead to purchases.
5. Demographics you collect: country, age, gender.
6. Technical details: device, operating system, browser, language.
7. A/B test variants they saw.

Check spend vs ROAS every day

This report compares what you spent on ads with the revenue those ads brought in, by channel, campaign and creative. It shows which ads pay off and which don’t, so you know where to add budget and what to turn off. If you check only one report every day, make it this one.

ROAS is usually read at fixed points. D7 ROAS is the revenue a group of customers brought in their first 7 days, divided by what it cost to acquire them. D14 and D30 work the same way.

Next to it, keep the in-between numbers that warn you early, such as cost per quiz start and cost per paywall view. They move days before purchases do.

Read this report on your own payment data. Platforms also credit themselves with purchases from people who only saw an ad, so their totals run higher than what reached your bank account.

Keep two dashboards: trends and daily work

Revenue analytics built into funnel platforms show ARPU (average revenue per user), LTV and subscription health next to the funnel steps. That covers much of the trends view without a separate tool.

Track revenue by cohort

Group customers by the month they joined and follow what each group pays over time. Compare these cohorts by plan, channel and country to see which customers are worth paying more for.

Plan length changes the picture a lot. After 12 months, about 3% of people on weekly plans are still paying, against about 10% on monthly plans (State of Web2App 2026). Annual plans bring most of their revenue in the first month, while monthly and weekly plans build it up over time.

Plan ahead with a financial model

A financial model predicts what a decision will bring before you make it. If you spend this much at this CPA, it tells you:

• how many customers you get,
• what their cohorts will pay month by month,
• and when you earn the spend back.

The model can be a spreadsheet with four blocks:

1. spend per month,
2. expected CPA,
3. the revenue curve per customer from your cohorts,
4. and the resulting payback month.

Next to it, keep a plan-versus-actual report. It compares each month with the model and projects where you’ll land by the end of the period. This report is also the main document for an internal business case.

Clean your data before you trust it

Three things distort funnel numbers more than any others:

1. Clicks that never reach the page. If many ad clicks don’t turn into page views, the page loads too slowly, usually inside the in-app browser, or the ad promised something the page doesn’t show.
2. Bots. Ad platforms send their own crawlers to check your pages, and other bots visit too. Filter them out of your analytics by user agent, the name a browser or bot reports about itself. Meta’s crawler, for example, calls itself facebookexternalhit. Unfiltered bots inflate visits and make conversion look worse.
3. Double counting. A purchase reported by both the pixel and the server has to carry the same event ID, or it’s counted twice and ROAS looks better than it is.

Section 11

Experiments

Once the funnel runs and the reports work, growth becomes a loop:

1. find the step where people leave,
2. form an idea of why,
3. test a change,
4. and keep what earns more.

The order of tests matters as much as the ideas.

Check that the funnel works before you test

If a step is broken, the test result shows the broken step, and you learn nothing about your idea. Apple Pay might not show up in the TikTok browser, a button might not work on some phones, or purchases might not reach your analytics. Before your first experiment, and after every big change, check that:

1. The funnel works in in-app browsers. Open it from Instagram, Facebook and TikTok on both an iPhone and an Android phone, and check that Apple Pay and Google Pay show up.
2. People move the way you expect. Session recordings from a tool such as Hotjar show where they hesitate, tap the wrong thing or leave.
3. A real purchase works end to end. Buy with a real card, open the app, check that the plan you paid for is available, then refund yourself.
4. People come back after payment. If the payment happens on another page or domain, make sure they land back in the funnel.
5. Every event fires once, with value and currency on purchases.

Test the steps before the paywall first

To find the weak step, compare your funnel with these typical ranges for cold traffic in Tier 1 countries:

Fix the first step that falls short before you judge the ones after it. Later steps can’t be read until earlier ones work. If people reach the paywall but few of them buy, check:

• the price and the offer,
• the trust signals on the paywall,
• and which payment methods show up.

Test these blocks separately, roughly in this order:

1. the first screen and the first question
2. the story: the order of questions, feedback screens and proof
3. the result and the personal plan
4. where the email step sits
5. the paywall: layout, plans, the default plan and trust signals
6. pricing: prices, paid trial or intro offer, plan lengths
7. the checkout: where it sits and which payment method comes first

Test whole messages. A change that carries one new idea across several screens, such as a new promise that runs from the first screen to the paywall, tells you more than a new word or button color.

Find test ideas in your own data

1. Segments. Compare conversion by quiz answer, country, age and device. If one group buys far more than the rest, tune the creative, the funnel and the paywall toward it, or give it its own funnel.
2. Interviews. Talk to recent buyers and to people who cancelled. Jobs-to-be-done interviews ask what was happening in someone’s life when they went looking for a solution, and they give you the words to use in ads and on screens.
3. Usability tests. Watch five people go through the funnel after any big change. Confusing screens show up fast.
4. Questions inside the funnel. One research question in the quiz, such as what people have tried before, collects answers from everyone who passes through.
5. Competitors. Their live funnels show what they’re testing right now.

Change one thing per test

1. Split traffic randomly between versions over the same period, so ad changes and seasonality hit both equally.
2. Change one thing per test and keep the ads the same. Your best 3–5 creatives make a good control.
3. Wait for 25–50 purchases per version before deciding. Below that, differences are mostly noise.
4. Run two or three tests at a time at most, so each gets enough traffic.
5. Judge by purchases and revenue per visitor. How many people reach the paywall says little on its own.
6. Write down every test: the idea, the result and the decision, so the team doesn’t retest old ideas.

Most funnel platforms split traffic between versions for you.

Section 12

Launch checklist

Every step from this guide, in the order you’ll do it. Tick them off as you go.

Plan

1. Set a budget: about $300–400 a day for a first Meta campaign, and expect ads to pay back in 3–6 months.
2. Choose how to build the funnel: with an AI coding agent or on a funnel platform.
3. Choose how to take payments: your own payment provider or a merchant of record.
4. Make sure you can export customers and subscriptions if you switch tools.

Creatives

1. Pull hooks and angles from your best organic videos.
2. Prepare at least six angles with three variations each.

Funnel and paywall

1. Repeat the ad’s promise on the first screen, and make the first question a single tap.
2. Ask for the email before the result.
3. End with a result screen built from the person’s answers.
4. Show three plans with one selected by default, and start with a paid entry price.
5. Show the price, billing period, renewal and how to cancel next to the pay button.
6. Add an upsell after the purchase.
7. Set up languages, currencies and regional prices for every market you sell in.

Payments

1. Make Apple Pay the default and keep other payment methods visible.
2. Set up retries for failed payments.
3. Set a clear statement descriptor and send renewal reminders to prevent disputes.
4. Handle sales tax through a merchant of record or a tax tool.
5. Build a cancel flow with alternatives and the full cancellation on one screen.

App access

1. Build a download page with clear steps into the app.
2. Make sure the app recognizes buyers, with a login by email as a fallback.
3. Keep the in-app paywall for people who install the app directly.
4. Keep web and in-app subscriptions in one place, and check that status updates after every purchase, renewal and refund.

Tracking

1. Send purchases to ad platforms from both the pixel and the server, with one ID per purchase.
2. Keep click IDs through every screen, and add UTM tags to every ad link.
3. Send funnel events and user properties to your analytics.
4. Make a real test purchase in the Instagram and TikTok browsers on iPhone and Android.

Launch

1. Start one Meta campaign optimized for purchases: broad targeting, one market, three ad sets by angle.
2. Change nothing before 50 purchases, then change one thing at a time.
3. Test new creatives in a separate campaign.

After launch

1. Check spend vs ROAS every day.
2. Fix the weakest funnel step first, then test the steps before the paywall, with 25–50 purchases per version.
3. Add a second payment provider before you scale.
4. Add TikTok and Google once Meta CPA is stable.