Conversion funnel: where users leave and what the designer does with it

The conversion funnel is a loss map. On the left are all the users who started. On the right are those who completed it. Between them are all those who left at every step.

Most teams look at the funnel for the final conversion: “We have 4% conversion from landing to buying.” This is an important figure, but useless for action. Where exactly is the other 96 percent lost? First step? The last one? Equal on each?

The answer determines what to do. And it's the designer's responsibility to find the break point and eliminate it.

## How the funnel works: basic mechanics

A funnel is a sequence of states the user goes through. Between each of the two states there is a conversion – the percentage of users who have moved on.

** Simple registration funnel:**

Landing (100%)
35% (conversion to registration)
Registration form (35%)
72% (completion rate form)
Confirmation of email (25%)
68%
First entrance (17%)
↓ 80%
Completion of onboarding (14%)
55%
First key action (8%)

The final conversion rate is 8%. But where is the most lost? On the transition from landing (losing 65%). On the registration form - 28%. After email confirmation, 32%.

The biggest absolute losses are on the landing. But this does not always mean that you need to work there in the first place: sometimes “screening out” on the landing site – normal – leave untargeted users. But onboarding losses (out of 25% that reached registration, only 14% complete onboarding) are a real problem.

## Three types of funnels

### Acquisition funnel (acquisition funnel)

From first familiarity with the product to registration or first purchase.

Typical steps:

• Advertising display/organic search
• Going to the site
• Viewing the landing
• Click on CTA
• Filling out the form
• Completion of registration/purchase

Who works with this funnel: marketing (traffic and advertising budget) + design (landing, form, CTA).

### Onboarding funnel (activation funnel)

From registration to the first “aha moment” – the moment when the user understands the value of the product.

Typical steps:

• Register
• First entry into the product
• Interface familiarization
• First key action
• Getting a valuable result

This is the direct area of responsibility of the designer.

### Conversion funnel (monetization funnel)

From active user to payer.

Typical steps:

• Active user
• Reaching the free plan limit
• View pricing page
• Start checkout
• Completion of payment

## How to read the funnel: what is important

### Look at absolute numbers, not just percentages

“Conversion from step 2 to step 3 – 85%” – good or bad? It depends on how many people are on step 2. 85% of 10,000 — you lose 1,500. 85% of 100, you lose 15. Priority depends on scale.

### Compare it to benchmarks

Converting from the registration form to the completion of the registration 60% is a bad thing? For a form with 10 fields, it is normal. For a form with 2 fields – very bad.

Comparative benchmarks:

• Landing → registration: 2–5% (regulation for SaaS)
• Registration form → completion: 70-85%
• Onboard completion: 50–70%
• Activation rate (achieving aha moment): 30-60%
• Free-to-paid conversion: 2–8%

### Break it down into segments

The aggregate funnel hides the patterns. Build a funnel separately for:

• Mobile vs. desktop
• Different channels of attraction
• New vs. returning
• Different countries or languages

It often turns out that the funnel is good for desktop and catastrophic for mobile. Good for organic traffic and bad for paid traffic.

### Look at the dynamics in time

Conversion rate = 4% is a snapshot. “Conversion rate has dropped from 6% to 4% over the past month.” Always look at the trend.

## Diagnostics: Why users leave

When a break point is found, you need to understand why. Data says "where," but not "why.".

### Quantitative methods

Heatmaps. Where they click, where they look (eye-tracking if any), how far they scroll. Click next to the button, but not on it? Aren't they scrolling to important content? - The CTA needs to be raised.

Session recordings. Record real user sessions (Hotjar, FullStory, Microsoft Clarity) Look at the sessions of users who dropped on a particular step. What were they doing before they left?

Rage clicks / Error clicks. Repeated clicks on one place are a sign that the user expects a reaction and does not. Or the element looks clickable, but is not.

Form analytics. Which fields are filled, which are left blank, where users are the longest. If the “phone” field causes 40% of refusals to fill out the form, maybe it is optional?

### Quality methods

Usability test on a specific flow. Give the user a task to go through the problem step and watch. 5 users will show 80% of the problems.

**Interview with departed users. ** Those who have not completed registration or onboarding are a valuable source of knowledge. The letter “We noticed you didn’t complete the setup – can we ask why?” gives surprisingly honest answers.

**In-app polls. ** At the moment of leaving the problem step, a short question: “What prevented you from continuing?” Even 10-15 answers give patterns.

## Typical causes of ruptures and how to fix them

### Gap 1: Landing → Registration (conversion rate < 2%)

** Possible causes:**

• The headline does not resonate with the audience
• There is no clear next step
• No, trust signals
• Slow page loading
• Disparity between advertising and landing (message mismatch)

What a designer does:

• Testing different formulations of the value proposition
• Makes the CTA more visible and specific
• Adds social proof above fold
• Optimize image size and download speed
• Creates landing pages for specific advertising messages

### Gap 2: Registration form (completion rate < 65%)

** Possible causes:**

• Too many fields
• Misunderstood error reports
• No realtime validation
• Complicated password requirements without explanation
• No progress in multi-step form
• No trust (no SSL, no privacy policy next to the form)

What a designer does:

• Remove optional fields (only email + password at the start)
• Adds inline validation with understandable messages
• Shows progress bar in multi-step form
• Adds “show password” to the password field
• Link to the privacy policy next to the submit button

### Gap 3: Onboarding (completion rate < 50%)

** Possible causes:**

• It’s a long way to get to the first valuable result
• Tutorial from slides that no one reads
• The user does not know what to do next
• The product seems complicated before it becomes useful
• Mandatory steps that could have been postponed

What a designer does:

• Shortens the path to aha moment
• Replace the tutorial with contextual hints
• Makes the first action as simple as possible (templates, examples)
• Removes optional steps from mandatory flow
• Adds progress bar so that the user can see that the finish is close

### Break 4: Aha moment → Regular use

** Possible causes:**

• The user received the first result, but did not understand how to continue
• No return mechanism (notification, email, streak)
• The value of the first result is insufficient for a second visit
• The bar is too high for the second result

What a designer does:

• Makes the next step obvious after the first result
• Launches Welcome Email Sequence
• Adds the mechanics of preservation and continuation (draft state)
• Reduces the complexity of second use

### Gap 5: Free-to-paid conversion (<2%)

** Possible causes:**

• The user does not see the difference between free and paid
• Paywall arrives before the user realizes its value
• The form of payment causes distrust
• No answers to objections at the time of decision
• The user did not find the desired feature in the paid plan

What a designer does:

• Makes paywall contextual – appears when the user wants more
• Simplifies checkout (minimum steps, save the card)
• Shows the specific value of a paid plan (not a list of features, but “what you can do”)
• Adds a trial period without entering a card
• Creates a clear comparison of plans

## How to build a funnel in analytics

In Amplitude: Use Funnel Analysis. Select events in the right order, set a time window (for example, “the user must complete all the steps in 7 days”), get a breakdown on each transition.

** In Mixpanel:** The Funnel report works similarly. Particularly useful breakdowns on the properties of the user.

In Google Analytics 4: Explore → Funnel Exploration. Allows you to build open funnels (the user can skip steps) and closed (strict order).

No special instruments: You can create a funnel through SQL requests for event data. You need the help of an analyst, but the data will be more accurate than in ready-made tools, if the analytics are custom.

## Prioritization: Where to Start Funnel Optimization

When you find several break points, you need to choose where to start. Use the influence × force matrix:

**High Impact + Low Effort: Do it right away. Remove the extra field from the form, clarify the CTA, add progress bar.

**High impact + high effort: plan as a separate project. Onboarding redesign, value proposition change, new pricing page structure.

**Low impact + low effort: * If you have time. Small cosmetic improvements.

**Low impact + high effort: don't do it.

To assess the impact, consider this: if conversions increase by 10%, how many additional users will reach the end of the funnel?

## What to monitor regularly

The funnel is not a one-time analysis, but regular observation.

Weekly:

• Abrupt changes in conversion rate at any step (> ± 10% weekly change - signal)

Monthly:

• Trends in key transitions
• Cohort comparison (this month vs. last)
• Breakdown by segment

After each release:

• Comparison of the funnel before and after changes

## Funnel analysis checklist

• The funnel is built with absolute numbers, not just with percentages
• Analysis by segment (mobile/desktop, channels, new/returned)
• Found the point with the greatest losses
• Qualitative analysis of this point (heatmap, session recording, test with users)
• A hypothesis about the cause of the rupture is formulated
• Design solution with expected effect
• Planned A/B test or clear before/after measurement

## Non-standard breakpoints: where users lose in an unobvious way

The classic breakpoints - the registration form, checkout - are usually under control. But there are less obvious places where the funnel bursts quietly.

### Email confirmation as a bottleneck

After registration, many products require email confirmation. It seems like a simple step. This is where 20-40% of users can be lost.

Why are they leaving

• The letter gets into spam
• The user switched to another tab and forgot
• The email did not arrive immediately, the user lost patience
• I don’t know what to do if the letter doesn’t come

What do you do

• The "Check Email" page should give specific instructions: "Letter from noreply@example.com, subject "Confirm email""
• The “Send Again” button is visible without any extra clicks
• “Change email” if the user is wrong
• For Critical Products: Magic link removes confirmation step altogether

### 404 pages in the funnel

The user clicked on the link from the email, received a 404 and left. This is not a hypothetical scenario: ad links expire, UTM options break routing, pages move without redirects.

How to control:**

• Monitoring 404 errors in analytics (especially referrer – from advertising, email)
• a 404 page should help you find what you’re looking for, not just say “not found.”
• Redirects when changing URLs are required

### Slow loading as an invisible drop-off

The user clicked - sees the spinner - waits - leaves. In analytics, this looks like a high bounce rate, but the real reason is performance.

According to Google, with a delay of 1 → 3 seconds, the probability of failure increases by 32%. At 1→5 seconds - by 90%.

** How do I find out:**

• Core Web Vitals in Google Search Console
• Lighthouse score of specific funnel pages
• Real User Monitoring through New Relic, Datadog, or free web-vitals. js

Quick wins:

• Image optimization (WebP, correct dimensions)
• Lazy loading for images below fold
• Transfer of non-critical scripts to async

## Funnel for returning users: a separate story

New users and returnees are different audiences with different tasks. The crater needs to be built for each individual.

**A user who returns after 7 days is likely to remember the product. He needs to get into context quickly and move on from where he left off.

**A user who returned after 90 days has almost forgotten. It needs mini-onboarding: "That's what's changed, that's where you've been doing before.".

**The user who opened the reactivation email is cooled but interested again. He needs concrete value right away, without going through onboarding again.

### Design for returning

Last session state. Show what the user did last time: “You worked on Project X.” Continue

Changelog / What's new. For a user who has returned after a long absence - a short list of what has changed during this time. Not the full release notes, but "3 things that are now easier.".

**Restoring the unfinished. ** If the user has started the Flow and has not finished, show it to him when returning. You have an unfinished form from April 7. Continue

## Mobile funnels: specifics and differences from the desktop

Mobile users behave differently, and their funnels are different.

** Interruptions.** Mobile users are frequently interrupted: call, notification, need to get off the bus. The funnel should support interruptions - maintain a state, allow you to return.

Keyboard flow. Filling out forms on mobile is painful. The wrong type of keyboard (number instead of email, regular instead of phone number), autocorrection, which breaks the password, the inability to see what you type - all this creates friction and increases the error rate.

Vertical scrolling. ** On mobile users are more likely to scroll than on desktop. Long landing is normal for mobile. But every screen has to keep an eye on it.

Touch targets. Buttons and links should be large enough to be pressed with a finger (minimum 44×44px). Small elements create tap errors – the user clicks the wrong thing.

### Separate analytics for mobile

Always look at the funnel separately for mobile and desktop. Typical picture:

• Desktop: conversion rate 5%, mobile: 2%
• The problem is that 70% of mobile traffic
• Total conversion: ~2.9%, although desktop conversion is excellent

Don’t mess with the data – the breakdown by device often reveals the main problem.

## How to build a funnel culture in a team

The funnel is not a one-off project. It's a constant practice. And it only works when the whole team looks at it regularly.

Weekly funnel review. Once a week - 15 minutes on the main numbers. Not a detailed analysis, but a look at key metrics: what has changed, whether there are anomalies.

**Agree that every week someone from the team (designer, product, analyst in turn) makes an analysis of one particular transition in the funnel: where is the loss, why, what can be done.

Data before design. Before redoing any funnel step, look at the data. Not because of bureaucracy, but because intuition is often wrong. It often seems that the "obvious" problem is one screen, and the data shows that the real problem is three steps earlier.

Document the results. After each change in the funnel, record what happened, what happened, what happened. This is the basic training of the team and the basis for future decisions.

## AI and Conversion Funnel: How to Find Breaks and Understand the Causes

AI helps interpret funnel data, generate hypotheses about the causes of ruptures, and prioritize points for improvement.

### Prompt: analyze the funnel and find priorities

Here are the details of our funnel [onboarding/registration/purchase]:

Step 1 [Title]: [N] users
Step 2 [Title]: [N] Users
Step 3 [Title]: [N] Users
Step 4 [Title]: [N] Users
Step 5 [Title]: [N] Users

The context:
- Type of product: [Description]
Main source of traffic: [description]
Devices: [% mobile/desktop]

Analyze the funnel:
1. Calculate conversions at each transition
2. Identify the most critical gap (highest losses × their position in the funnel)
3. Propose 3-5 hypotheses about the causes of each major gap
4. How to test each hypothesis (method and what you need)
5. Prioritize Improvement Points: What to Fix First

### Prompt: interpreting session data

If there is data from session recordings (Hotjar, Clarity) - AI helps to find patterns:

I looked at 20 user session entries for [step funnel where we lose].

Here's what I observed:
[Observation 1, for example: users scrolle down before pressing the button]
- [observation 2]
- [observation 3]
...

Interpret these observations:
1. What are users looking for or not understanding?
2. What UX problems explain each behavior?
3. Suggest specific design changes for each problem
4. How do you measure what changes have helped?

### Prompt: compare funnels by segment

We have funnel data for two segments:

Segment A (mobile):
[step data]

Segment B (desktop):
[step data]

Find:
1. What steps are the biggest difference?
2. What usually causes such a difference between mobile and desktop?
3. What do I need to check to confirm the cause?
4. Suggest 3 Specific Design Changes to Improve Mobile Funnel

### Prompt: Write a funnel optimization plan

Current situation:
Main gap: [step X → step Y], conversion [%]
Best Benchmark Scenario: [%]
- Data on the cause: [what found through heatmaps/sessions/interviews]

Create an optimization plan:
1. Hypothesis (in the format if → that)
2. Design solution (specific)
3. Verification method (A/B test / before-after / test with users)
4. Metric of success (primary and guardrail)
5. Expected financial impact (help calculate if you give data on LTV)

Make it as a 1-page brief to agree with the product.

### Prompt: find non-standard break points

Losses often don’t happen where they are obvious. AI helps you think more broadly:

Here's our funnel and data:
[data]

I already know about the major gaps. Help me find the non-standard ones:

1. What points between steps are often missed when analyzing funnels in [product type]?
2. Where can users get lost outside the main funnel (email confirmation, 404, slow download)?
3. Which segments are worth checking out separately - maybe they have a different picture?
4. What external factors can affect our funnel that are not visible in the product data?