.svg)
Funnel analysis is the process of tracking how users move through a defined sequence of steps, from first visit to sign up, from sign up to activation, from trial to paid, and identifying where and why they drop off. It is one of the most important analytical methods available to product and growth teams because it connects user behaviour directly to business outcomes. This guide covers what funnel analysis is, how to build and read a funnel, the most common mistakes teams make, and how to turn funnel insights into experiments that improve conversion.
Every product has a journey it wants users to take.
Visit the site. Sign up. Complete onboarding. Use the core feature. Upgrade to paid. Refer a friend.
Most users do not complete that journey. They drop off somewhere along the way, and without funnel analysis, you do not know where, how many, or why.
That is the gap funnel analysis closes. It makes the invisible visible: showing you exactly where users leave, which steps have the most friction, and which segments behave differently from the rest.
What Is Funnel Analysis?
Funnel analysis is a method for measuring how users progress through a defined sequence of steps and calculating the conversion rate between each one.
A funnel is defined by a series of events, they are actions users take in your product or on your website in a specific order. Amplitude, for example, lets you define a funnel by selecting the events that represent each step: page viewed, sign up started, sign up completed, first feature used, subscription started. Once defined, the funnel shows you how many users completed each step and what percentage dropped off before reaching the next one.
The output is a visual representation of your conversion pipeline, typically shown as a series of bars or steps, each one smaller than the last, with the drop-off rate between each step clearly visible. This gives you an immediate picture of where your biggest conversion losses are happening.
Why Funnel Analysis Matters
Funnel analysis is useful because conversion problems are rarely where you think they are.
Most teams have an intuition about where users drop off. They think it is the pricing page, or the onboarding flow, or the sign-up form. Funnel analysis tells you whether that intuition is correct, and more often than not, the biggest drop-off is somewhere unexpected.
Finding the real drop-off point changes everything about how you prioritise fixes. A team that spends three months optimising their pricing page when 60% of users are actually dropping off at the activation step is wasting its highest-leverage resource: engineering and design time.
Funnel analysis makes sure you are working on the right problem.
How to Build a Funnel in Amplitude
Building a funnel in Amplitude starts with defining the events that represent each step in the journey you want to measure.
The quality of your funnel depends entirely on the quality of your event tracking. If your events are not firing consistently, have inconsistent naming across platforms, or are missing key properties, your funnel will produce misleading results. Clean event tracking is the foundation everything else builds on.
Once your events are defined, navigate to the Funnel Analysis chart type in Amplitude. Select the events that represent each step in the order users should complete them. Set your conversion window, which is the timeframe within which users must complete all steps to be counted as converted. Amplitude's default is 30 days but the right window depends on your product and use case.
The funnel will immediately show you the conversion rate between each step and the overall conversion rate from first step to last. You can then apply filters to segment the funnel by user properties like device type, acquisition channel, plan tier, geography to understand whether different segments convert differently.
This segmentation step is where funnel analysis gets genuinely powerful. An overall conversion rate of 12% tells you something is wrong. A segmented funnel that shows mobile users converting at 6% while desktop users convert at 18% tells you exactly where to focus.
How to Read Funnel Analysis Results
Reading a funnel correctly requires looking beyond the overall conversion rate and asking four questions.
Where is the biggest absolute drop-off? The step with the largest number of users dropping off is your highest-priority problem, even if the percentage drop-off is not the largest. Fixing a step where 10,000 users drop off at 40% is more impactful than fixing a step where 500 users drop off at 60%.
Where is the biggest percentage drop-off? The step with the highest percentage drop-off often indicates the most significant friction point, something in the user experience or the product design that is creating a barrier.
Is the drop-off consistent across segments? A drop-off that affects all users equally points to a universal problem. A drop-off that is concentrated in one segment with new users, mobile users, users from a specific acquisition channel, points to a segment-specific problem that may require a targeted fix.
Is the drop-off new or persistent? Comparing funnels over time tells you whether a drop-off is getting worse (suggesting a recent product change broke something), stable (suggesting a chronic friction point), or improving (suggesting a fix is working). Amplitude dashboards surface the "what" but not always the "why" -- funnel data tells you where users leave, but you need qualitative data to understand the reason.
What to Do When You Find a Drop-Off
Finding a drop-off is the beginning of the process, not the end. The next step is diagnosing why users are leaving at that point.
The most effective diagnostic combination is funnel analysis paired with session replay. Funnel analysis identifies the step where users drop off. Session replay lets you watch exactly what those users did before they left, where they clicked, what they tried, what seemed to confuse or frustrate them.
In Amplitude, you can jump directly from a funnel drop-off into the session replays of users who left at that specific step. Amplitude Session Replay gives you the full behavioural context alongside the quantitative data, so you are not guessing at the cause of the drop-off, you are watching it.
Once you understand why users are dropping off, you have the evidence base to design an experiment. A well-grounded funnel insight becomes a strong hypothesis: "If we simplify the form at step three by removing the company size field, completion rate will increase because session replay shows 40% of users abandoning after reaching that field." That hypothesis is testable, directional, and grounded in evidence.
Turning Funnel Insights Into Experiments
Funnel analysis is one of the most reliable sources of high-quality experiment hypotheses. Drop-off points represent real friction, users who were motivated enough to start a journey but stopped before completing it. They are the users most likely to respond positively to a well-designed fix.
The 8-step A/B testing framework starts with identifying where to test, and funnel drop-off points are the most evidence-backed starting point available. Rather than testing based on intuition, you are testing based on measured user behaviour.
The most effective CRO and product teams run this cycle continuously: funnel analysis identifies the biggest drop-off, session replay diagnoses the cause, a hypothesis gets written, an experiment runs, and the result feeds back into the next round of analysis. Each cycle compounds on the last.
Common Funnel Analysis Mistakes
Using too many steps. A funnel with fifteen steps is not more insightful than one with five: it is harder to read and harder to act on. Start with the three to five steps that represent the core journey from acquisition to the primary conversion event. Add granularity once you have fixed the obvious problems.
Setting the wrong conversion window. A conversion window that is too short misses users who convert on a longer cycle. A window that is too long dilutes the signal by including users whose conversion was unrelated to the steps in the funnel. Match the window to your actual user behaviour, look at your conversion time distribution before setting it.
Not segmenting the funnel. An unsegmented funnel tells you that something is wrong. A segmented funnel tells you who it is wrong for and where. Always check funnel performance by acquisition channel, device type, and user type (new versus returning) before drawing conclusions from the overall data.
Treating funnel data as the final answer. Funnel analysis tells you where users drop off. It does not tell you why. Pairing it with session replay, surveys, and user feedback is what turns a drop-off into a diagnosed problem you can actually fix.
Want to turn your funnel insights into a compounding experiment programme?
If you are identifying drop-offs in your funnel but your experimentation programme is not building on those insights fast enough, the Experimentation Growth Engine automates hypothesis generation, prioritisation, and result evaluation -- so every funnel finding feeds directly into your next experiment.
👉See the Experimentation Growth Engine
Want to talk through your funnel setup?
FAQ
What is funnel analysis in product analytics?
Funnel analysis is the process of tracking how users move through a defined sequence of steps -- from first visit to sign up, activation, and conversion -- and measuring the drop-off rate between each step. It identifies where users leave a journey and how many, giving product and growth teams a clear picture of where conversion losses are happening.
How do you build a funnel in Amplitude?
In Amplitude, navigate to the Funnel Analysis chart type, select the events that represent each step in the order users should complete them, and set a conversion window that matches your typical user behaviour. Amplitude will show the conversion rate between each step and the overall conversion rate. Apply user property filters to segment the funnel by device, acquisition channel, or user type.
What is a good conversion rate for a funnel?
There is no universal benchmark -- conversion rates vary significantly by industry, product type, funnel length, and user segment. The more useful question is whether your conversion rate is improving over time and whether it differs significantly across user segments. A sign-up funnel converting at 8% for desktop users and 3% for mobile users points to a specific, fixable problem regardless of whether 8% is a "good" rate in your industry.
What is the difference between funnel analysis and cohort analysis?
Funnel analysis measures how users progress through a specific sequence of steps and where they drop off. Cohort analysis groups users by a shared characteristic -- typically the date they first signed up or used a feature -- and tracks how their behaviour changes over time, most commonly used for retention analysis. Funnel analysis answers "where are users dropping off?" Cohort analysis answers "how are different groups of users retaining over time?"
How do you diagnose a funnel drop-off?
Start with the funnel data to identify which step has the biggest drop-off and which user segments are most affected. Then use session replay to watch what users actually do before leaving at that step -- where they click, what they try, what seems to confuse them. Combine this with exit surveys or in-product feedback to get the user's own explanation. The combination of quantitative funnel data and qualitative behavioural data is what turns a drop-off into a diagnosable, fixable problem.
What conversion window should I use in Amplitude funnel analysis?
The right conversion window depends on your product's typical user behaviour. Look at how long users who do convert actually take to complete the journey -- if 80% of conversions happen within 7 days, a 7-day window captures most of your signal without distortion. If you set the window too short, you exclude legitimate conversions. Too long, and you include users whose completion was unrelated to the funnel steps you are measuring.
.png)
