What funnel analysis is
A funnel is an ordered list of steps you expect people to take before they buy. Funnel analysis counts how many people reach each step, and in what order, so you can see where most of them leave.
A typical SaaS funnel is: visit the site, view pricing, start a trial or sign up, then pay. An online store's is: view a product, add to cart, start checkout, then pay. Every step loses people. The point is to find the step that loses more than it should.
Funnels are strict about order. A person counts at step 3 only if they also completed steps 1 and 2 before it. That is the difference from simply counting each page's visitors, and it's why funnel numbers are smaller and more honest.
Funnel formulas
Step conversion = people who reached this step ÷ people who reached the previous step
Drop-off = 1 − step conversion
Overall conversion = people who finished ÷ people who started
Overall conversion is the product of the step conversions. That is why a small gain at any step improves the whole funnel by the same proportion.
Worked example
A software company tracks four steps over 30 days:
| Step | People | Step conversion | Drop-off | From start |
|---|---|---|---|---|
| 1. Landing page | 10,000 | — | — | 100% |
| 2. Pricing page | 3,000 | 30% | 70% | 30% |
| 3. Sign-up | 600 | 20% | 80% | 6% |
| 4. Paid | 150 | 25% | 75% | 1.5% |
Overall conversion is 150 ÷ 10,000 = 1.5%, which is also 30% × 20% × 25%.
Which step to fix? Don't just pick the biggest drop-off percentage. Ask what each fix would be worth. Raising pricing → sign-up from 20% to 25% gives 750 sign-ups and, at the same 25% paid rate, about 188 customers: 38 more. Raising landing → pricing from 30% to 35% gives 3,500 pricing views, 700 sign-ups and 175 customers: 25 more. At $40 a month, the first fix adds about $1,520 of MRR, the second about $1,000.
Then look at who dropped off, not just how many. If the 2,400 people who left pricing without signing up mostly came from one ad campaign, the problem may be the traffic, not the page.
How to set up funnel analysis
- Map the real path. Use paths to payment data, or watch session recordings, to see the order people actually take. Don't build a funnel around a path nobody follows.
- Track each step as an event: a page view, a button click, a form submission, a payment.
- Keep it short. Three to six steps. Every extra step adds noise and a place for tracking to break.
- Choose a time frame. Decide how long people have to complete it. A same-session funnel and a 30-day funnel answer different questions.
- Break it down by source. Compare the funnel for search visitors with the funnel for paid visitors. Their drop-off points often differ.
Use the conversion rate calculator to check step and overall rates from your own counts.
Common mistakes
- Ending the funnel at sign-up. A step that brings more sign-ups but fewer payments is a loss. End at the payment when you can.
- Ignoring order. Counting page visitors without order inflates later steps with people who arrived there directly.
- Fixing the biggest percentage drop by reflex. Some steps are supposed to filter. Weigh the value of each fix, as in the example above.
- Mixing traffic sources. A flood of low-intent traffic makes every step look worse. Segment before you redesign.
- Overlooking broken pages. A sudden drop at one step is often a bug, such as a JavaScript error on the checkout form, not a design problem.
- Small samples. 40 people at a step can't tell a 20% rate from a 30% one. Wait for more data or widen the date range.
How MetricRaven handles funnels
MetricRaven funnels end where the money is. Open Traffic → Conversions → Funnels and build a funnel from ordered steps: pages, goals and the payment itself. A later step counts only when the same tracked visitor completed every earlier step in order. A site can have up to 50 funnels.
- Who dropped off. Under each step, N dropped lists the visitors who reached it but not the next one, with where they first came from, their country and what they spent. Open a person to see their journey.
- Time between steps. Each step shows the median time since the previous step and since the first one.
- Paid and time to pay. Next to the completion count, N paid shows how many finishers paid in the period and the median time from the first step to the payment.
- Errors at drop-off. With JavaScript errors on, a step shows a warning badge when visitors who left before it hit an error.
- Filters. Every dashboard filter applies, so you can compare the funnel for one channel or campaign with another.
Not sure which steps to use? Paths to payment shows the page sequences buyers actually take, and Save as funnel turns one into a funnel. See Goals & funnels.
Questions
- What is a funnel drop-off rate?
- The share of people who reached one step but didn't reach the next. If 3,000 people viewed pricing and 600 signed up, the drop-off at sign-up is 1 − 600 ÷ 3,000 = 80%.
- How many steps should a funnel have?
- Usually three to six. Fewer hides where people leave; more adds noise and more places for tracking to break. Include the payment as the last step when you can.
- What is the difference between funnel analysis and cohort analysis?
- Funnel analysis follows people through a sequence of steps toward one goal. Cohort analysis follows groups who started at the same time over weeks or months, to see how retention changes.
- Can a funnel include the payment?
- Yes, if your analytics tool knows about payments. In MetricRaven the payment is a step you can add to any funnel, and each funnel also shows how many people who finished it paid and how long it took.