Funnel analysis: finding exactly where users drop off
Master funnel analysis to pinpoint conversion leaks, prioritize fixes by impact, and turn step-by-step drop-off data into real growth.
Most growth problems are not mysterious. A business is losing customers somewhere specific — a form that is too long, a checkout page that loads too slowly, an onboarding email that never lands. The hard part is not knowing that a problem exists; it is knowing precisely where in the journey the problem lives. That is exactly what funnel analysis is designed to answer.
What is a funnel and why does it matter
A funnel models a user journey as a sequence of ordered steps. Each step is a discrete action: a page view, a form submission, a feature activation, a payment. Users enter at the top and, ideally, flow all the way to the bottom. In practice, some fraction leave at every step — and the shape of those exits tells you everything about where to focus your energy.
A classic e-commerce funnel might look like this:
- Visited product page
- Added item to cart
- Reached checkout
- Entered payment details
- Completed purchase
For a SaaS product you might model: visited landing page → started free trial → completed onboarding → used core feature → upgraded to paid.
The funnel structure is simple. The insight it produces is not.
Step conversion vs overall conversion
Every step has its own conversion rate: the share of users who completed that step and also completed the next one. But the overall funnel conversion — the share who go all the way from step 1 to the final step — is not an average of those rates. It is their product.
Consider a five-step funnel with the following numbers:
Step Users Step conversion
──────────────────────────────────────────────────────
1. Visit product page 80,000 —
2. Add to cart 32,000 40.0 %
3. Reach checkout 22,400 70.0 %
4. Enter payment 18,144 81.0 %
5. Complete purchase 15,422 85.0 %
Overall conversion: 15,422 / 80,000 = 19.3 %
Cross-check: 0.40 × 0.70 × 0.81 × 0.85 ≈ 0.193 ✓The product rule has a powerful implication: a single weak step caps the entire funnel. If step 2 converts at 40 %, you cannot achieve an overall conversion above 40 % no matter how perfect every downstream step becomes. Fixing a late step when an early one is leaking is like patching a hole at the bottom of a bucket while there is a bigger hole in the middle.
Where to focus: absolute drop-off, not just the lowest percentage
A common mistake is fixating on the step with the lowest conversion rate. Rate alone ignores volume. Impact is always a function of both.
In the example above, step 2 has the lowest conversion rate (40 %), but let us calculate absolute drop-off — users lost at each step:
Step 1 → 2: 80,000 − 32,000 = 48,000 users lost
Step 2 → 3: 32,000 − 22,400 = 9,600 users lost
Step 3 → 4: 22,400 − 18,144 = 4,256 users lost
Step 4 → 5: 18,144 − 15,422 = 2,722 users lostThe answer is unambiguous: step 2 is where 48,000 potential customers disappear. Even a modest improvement there — say, lifting the add-to-cart rate from 40 % to 50 % — adds 8,000 users to the rest of the funnel and, at a downstream conversion of roughly 48 %, yields around 3,800 additional purchases. That is more than the total users lost at steps 3, 4, and 5 combined.
The first diagram below illustrates this visually.
Open vs closed funnels, and why time windows matter
Not all funnels work the same way.
A closed funnel requires users to complete steps in strict order within a defined window. A user who visits the checkout directly (skipping “add to cart”) would not count at step 2. This gives a cleaner picture of the intended journey but can undercount legitimate conversions.
An open funnel allows users to enter at any step. This is more forgiving but can inflate conversion rates if many users skip the top of the funnel entirely.
The time window is equally important. If your attribution window is 30 days, a user who visited on day 1 and purchased on day 35 does not count as a conversion. That is correct behavior — otherwise you are measuring different cohorts of users at different stages and the math becomes meaningless. A window that is too short (say, 1 hour for a considered purchase) will make your funnel look worse than it is; one that is too long will mix intent signals from very different sessions.
For most B2C funnels, 7–30 days is reasonable. For high-consideration B2B products, 90 days is common. For micro-conversions like “added to cart → checkout,” 24 hours is usually enough.
Segmenting funnels to find who drops and why
An aggregate funnel is a starting point, not a finish line. The overall 19.3 % conversion figure above hides enormous variation across different user segments. Segmenting your funnel is often where the real insight lives.
Useful segmentation cuts include:
- Acquisition channel: paid search vs organic vs social vs email
- Device type: desktop vs mobile vs tablet
- User status: new vs returning
- Geography or language
- Pricing tier or plan (for SaaS)
The second diagram below shows why device type is so often revealing.
In this example, mobile converts at roughly half the rate of desktop on the very first step. If you only looked at the aggregate funnel, that signal would be diluted by the desktop majority. Segmentation reveals that the problem is almost certainly a mobile UX issue — perhaps a poorly placed call-to-action, a slow image load, or a form that is difficult to interact with on a small screen. That is a very specific, actionable finding.
Common mistakes that corrupt funnel analysis
Counting events instead of unique users. If a single user adds an item to cart three times, that should count as one user at step 2 — not three events. Event counts inflate the top of the funnel unevenly and make conversion rates meaningless.
Ignoring the time window. Without a fixed attribution window, users who converted weeks later get mixed into your “current” cohort. The funnel looks healthier than it is, and you cannot reliably compare one period to another. See the note in cohort retention analysis for a similar principle applied to retention metrics.
Survivorship bias in late-step analysis. By the time users reach step 4 in a five-step funnel, they are a highly self-selected group. They are already more motivated, more trusting, and less price-sensitive than the 80 % who dropped off earlier. Insights from that group may not generalize upstream.
Optimizing a late step when the leak is early. This is the most expensive mistake. If 60 % of users are lost at step 1, squeezing an extra 2 % from step 4 is rounding error. Always rank steps by absolute drop-off before deciding where to invest.
Not accounting for multi-session journeys. On mobile especially, users often research on one session and purchase on another — possibly on a different device. A strict single-session funnel misattributes these as drop-offs. Use a user ID (not a session ID or anonymous cookie) as your unit of analysis wherever possible.
Tie funnel findings to action
Funnel analysis answers where — it identifies the step that is leaking the most volume. It does not, by itself, tell you why users leave or what to do about it.
Once you have identified the priority step, the next moves are:
- Qualitative research: session recordings, heatmaps, and user interviews at that specific step. What are users seeing? Where does their cursor hesitate? What are they typing in the search bar right before they leave?
- Exit surveys: a short survey triggered when a user is about to abandon that step. “What stopped you from completing this?” yields signal that no quantitative tool can replicate.
- A/B testing: once you have a hypothesis (the form is too long; the price is unclear; the CTA button is below the fold), run a controlled experiment to validate the fix before rolling it out. The guide to A/B testing for business decisions covers how to size and analyze those experiments correctly.
This loop — funnel identifies where, qualitative research develops hypotheses, A/B testing validates fixes — is the foundation of systematic business analytics. It is repeatable, prioritizable, and directly connected to revenue.
For a broader view of how funnel metrics fit into a complete analytics stack, the glossary defines the core terms: conversion rate, drop-off, attribution window, cohort, and more.
Frequently asked questions
What is the difference between a funnel and a conversion rate?
A conversion rate is a single number: the share of users who completed a desired action out of those who had the opportunity. A funnel is a sequence of conversion rates measured step by step. The overall conversion rate of a funnel equals the product of all its step rates, so a funnel gives you far more diagnostic information than a single top-to-bottom rate.
How many steps should a funnel have?
There is no universal answer, but each step should represent a meaningful, discrete action that a meaningful share of users fails to complete. If step 2 → step 3 conversion is 99.5 %, those two steps do not need to be separate — they are not a decision point. Aim for steps that each represent a real behavioral threshold, typically 3–7 steps for most journeys.
Should I use a funnel for every metric?
Funnels are best suited to linear, goal-oriented journeys where sequence matters — onboarding, checkout, lead generation. They are less useful for exploratory or non-linear behavior, like how users browse a content site. For those cases, cohort retention and engagement metrics are more appropriate.
How do I handle users who skip steps?
In a closed funnel, users who skip a required step are excluded from the denominator at that step. In an open funnel, they enter wherever their first recorded action is. The right choice depends on your product: if skipping step 2 and landing on step 3 directly is a valid path (e.g., direct checkout without an explicit “cart”), use an open funnel. If the steps represent a gated sequence where skipping is impossible by design, use a closed funnel and investigate any skip anomalies as data-quality issues.