Funnel analysis is an analytics method that measures how many users progress through an ordered sequence of steps toward a goal, such as signing up, completing a purchase or activating a feature, and where they drop off along the way. In analytics, it turns a stream of individual events into a small set of step-to-step conversion rates that show which stage of a process loses the most people.
A funnel is defined by its steps, their order, the counting unit and a conversion window. Changing any of these choices changes the numbers, which is why a well-specified funnel definition matters as much as the chart that displays it.
How funnel analysis works
Funnel analysis works by matching each user’s events against an ordered list of steps and counting how many users reach each step within the allowed time. The first step defines the entry population; every later step counts only users who completed all earlier steps in the required order.
The main outputs are the step conversion rate, which divides the users at one step by the users at the previous step, and the overall conversion rate, which divides the users at the final step by the users at the first. The complement of each step rate is the drop-off rate, the share of users lost at that transition.
The table below shows a simple checkout funnel and how the rates are derived.
| Step | Users reaching step | Step conversion | Overall conversion |
|---|---|---|---|
| Product page viewed | 20,000 | Not applicable | 100% |
| Added to cart | 3,000 | 15.0% | 15.0% |
| Checkout started | 1,800 | 60.0% | 9.0% |
| Payment submitted | 1,260 | 70.0% | 6.3% |
| Order confirmed | 1,197 | 95.0% | 6.0% |
In this example, the largest absolute loss occurs between viewing and adding to cart, while the largest relative loss after intent is shown occurs between cart and checkout. Both views are useful, because a small percentage improvement at a high-volume step often adds more completions than a large improvement late in the funnel.
Key parameters of a funnel definition
Every funnel depends on a few definitional parameters that analytics tools expose as settings. Two funnels with identical step names can produce very different numbers if these parameters differ, so they should be documented alongside the result.
| Parameter | Common options | Effect on results |
|---|---|---|
| Counting unit | User, session, account, order | Defines what a conversion means |
| Conversion window | 30 minutes to 90 days | Longer windows raise conversion rates |
| Step order | Strict, relaxed, any order | Strict order excludes skipped or repeated steps |
| Entry criterion | First event, any occurrence | Changes who is counted at the top |
| Segmentation | Device, channel, country, cohort | Reveals where averages hide differences |
| Exclusion steps | Events that disqualify a path | Removes users who took an alternative route |
The conversion window deserves particular care. A session-based funnel with a 30-minute window suits a quick checkout, while a business-to-business onboarding funnel may need 14 to 90 days. A window that is too short undercounts slow but genuine conversions; one that is too long attributes unrelated later behavior to the original visit.
Funnel analysis compared with related analytics methods
Funnel analysis answers where users stop within one process, which distinguishes it from methods that track users over time or summarize performance. It is often combined with these neighboring methods rather than used alone.
| Method | Core question | Typical output |
|---|---|---|
| Funnel analysis | Where do users drop out of a sequence | Step and overall conversion rates |
| Cohort analysis | How do groups behave over time | Retention curves by start period |
| Path analysis | Which routes do users actually take | Flow diagrams of event sequences |
| KPI tracking | Is a target metric on course | Single values against goals |
| Attribution modeling | Which channels deserve credit | Weighted credit per touchpoint |
A practical pattern is to use funnel analysis to locate the leaking step, path analysis to see where those users went instead, and cohort analysis to check whether a fix also improves longer-term retention rather than only the immediate conversion.
Using funnel analysis to improve conversion
Funnel analysis is most valuable as a diagnostic starting point for product and marketing decisions. The usual workflow is to define the funnel, establish a baseline, segment it to find where drop-off concentrates, form a hypothesis about the cause and then validate a change with a controlled experiment.
Segmentation is where most insight appears. An overall drop-off of 40% at checkout may hide a rate of 25% on desktop and 60% on mobile, which points to a device-specific usability problem rather than a pricing issue.
| Observed pattern | Likely cause | Typical next step |
|---|---|---|
| Sharp drop at a form step | Too many fields or validation errors | Review field-level error logs |
| Drop concentrated on one device | Layout or performance defect | Test the step on that device |
| Drop at the payment step | Unexpected costs or missing methods | Compare quoted and final prices |
| Gradual decline across all steps | Weak intent from the traffic source | Segment by acquisition channel |
| Sudden change after a release | Tracking or product regression | Verify event instrumentation first |
Limitations and common mistakes
Funnel analysis shows where users leave, not why they leave. Explaining the cause requires qualitative research, session observation or experiments, and treating a funnel chart as a causal explanation is the most common analytical error.
- Broken instrumentation: a missing or duplicated event creates a false drop-off or an impossible conversion above 100%.
- Mixing counting units, such as sessions at the top and users at the bottom, which inflates or deflates the rates.
- Forcing a linear path on a process that users complete in varying orders, which hides real conversions.
- Ignoring sample size in small segments, where a few users swing the rate by many percentage points.
- Optimizing one step in isolation, for example removing a verification step that later raises fraud or churn.
Privacy controls and consent choices also matter. When some users decline tracking, the funnel describes only the observable population, and comparisons across periods with different consent rates can be misleading.
In everyday language
Think of a shop with several rooms that customers must walk through before reaching the cash register. Funnel analysis counts how many people enter each room and how many leave before the next one.
If a thousand people enter the shop but only two hundred reach the second room, the problem is probably near the entrance. If almost everyone reaches the register but many walk away there, the problem is at the register. The value of the method is that it points to the exact room where people give up, so a team knows where to look first instead of guessing.
Seen in the real world
A subscription meal service tracks a five-step signup funnel: landing page, plan selection, account creation, delivery details and payment. Over four weeks, overall conversion is 4.2%. Segmenting by device shows that mobile users complete the delivery details step at only 48%, compared with 81% on desktop.
Session recordings reveal that the address field rejects apartment numbers entered in a common format, and the error message appears below the visible area of small screens. The team fixes the validation and moves the message next to the field, then runs an A/B test for two weeks. Mobile completion of the step rises to 76%, and overall mobile conversion increases by roughly a third without any change to pricing or advertising spend.