Funnel Analysis
Also known as: conversion funnel, sales funnel, marketing funnel, onboarding funnel, checkout funnel, funnel optimization
Funnel analysis tracks how users move through a defined sequence of steps toward a goal (purchase, signup, activation) and measures the percentage who complete each step.
The visualization is called a 'funnel' because progressively fewer users complete each subsequent step -- the wide top (many who start) narrows to the smaller bottom (few who complete). Funnel analysis identifies where users drop off, quantifies the size of each drop-off, and guides optimization priorities. Common business funnels: marketing funnel (awareness to purchase), product funnel (signup to activation to habit), and checkout funnel (add to cart to payment complete).
Funnel Analysis, explained at three levels
At home
A funnel analysis tracks how many people complete each step of a process. Imagine an online store: 100 people visit the product page; 60 add the product to their cart; 30 reach the checkout; 15 complete the purchase. This is a funnel: 100 -> 60 -> 30 -> 15. At each step, people drop off. Funnel analysis shows you exactly where the biggest drop-offs happen -- in this example, 40 people dropped off between the product page and cart (40% drop), and 15 dropped off between checkout and purchase (50% of those who reached checkout). Fixing the step with the biggest drop-off will have the most impact on total completions.
At work
Every business process can be represented as a funnel: the hiring process (applications to interviews to offers to hires), the sales process (leads to demos to proposals to closed deals), the customer onboarding process (signup to first use to habit formation). Funnel analysis tells you where you are losing people and by how much. Improving the step with the highest drop-off rate (even a small improvement at a high-volume step can have significant impact on final completions) is the highest-leverage optimization opportunity the funnel reveals.
At home
Funnel analysis requires defining: the steps in sequence (what constitutes Step 1, Step 2, etc.?), a time window (users must complete all steps within X days to be counted as progressing), and the entry point (who is eligible to enter the funnel?). Different time window choices produce different conversion rates. A checkout funnel where users must complete all steps in one session will show lower conversion than one that allows 7 days to complete -- but the 7-day version may count people who visited 5 times over a week, which inflates apparent conversion. Comparing funnel analyses requires ensuring the same definitions are used.
At work
Segmented funnel analysis reveals why conversion varies across user groups: the same funnel may show very different step-completion rates for mobile vs. desktop users, new vs. returning users, different traffic sources, or different geographic markets. A checkout funnel where mobile users drop off 60% at the payment entry step while desktop users drop off 20% points directly to a mobile payment UX problem. Funnel segmentation (breaking the funnel analysis by user attributes) is the diagnostic tool that moves from 'we have a conversion problem at step 3' to 'step 3 is broken specifically for mobile users on iOS.'
At home
Ordered vs. unordered funnels: standard funnel analysis assumes a fixed, ordered sequence of steps. Real user behavior is often non-linear: users may complete steps in different orders, skip steps, or revisit earlier steps. Unordered funnel analysis (sometimes called 'path analysis') counts a user as completing a funnel step if they complete it at any point during the session, regardless of order. This is more permissive but can miss important sequence information (does the order of step completion matter for eventual goal completion?). Path analysis tools (Sankey diagrams, sequence analysis) visualize the actual paths users take rather than forcing them into a predefined funnel.
At work
Funnel analysis and statistical significance: when comparing funnel conversion rates between variants in an A/B test, each step's conversion needs separate statistical testing. A funnel with 5 steps and two variants has 5 test comparisons per step, creating a multiple comparisons problem if all 5 are tested for significance. Best practice: define the primary conversion metric (final funnel completion rate) as the pre-specified primary endpoint; treat intermediate step changes as exploratory insights. Additionally, funnel analysis can show Simpson's Paradox: an overall conversion rate can improve while every step's conversion rate declines, if the mix of users (who tend to convert differently) shifts between variants.
Real-world examples
At home
Nadia was trying to understand why she rarely finished online courses she enrolled in. She tracked her own behavior: enrolled (100% of courses), started Module 1 (80%), completed Module 1 (60%), started Module 2 (30%), completed the course (10%). The biggest drop was between starting Module 1 and starting Module 2 -- she almost always started but rarely continued to the second module. Understanding her personal 'drop-off point' (first module completion) helped her choose shorter courses and set completion reminders after finishing each module.
At work
Thomas ran growth for a B2B SaaS product. The signup-to-paid conversion funnel showed: signup (100%), email verified (75%), first login (60%), core feature used (35%), team member invited (20%), paid conversion (12%). The biggest opportunity was the gap between 'core feature used' (35%) and 'team member invited' (20%) -- users who completed the core feature but didn't invite teammates had much lower paid conversion than those who did. He prioritized adding in-app prompts to invite teammates immediately after first core feature use. The change increased team invitation rate from 20% to 31% and paid conversion from 12% to 15%.
Common confusions
- Many teams assume that optimizing the first step in a funnel (awareness or visit rate) is the highest-priority intervention — The impact of improving any single funnel step depends on both the step's current drop-off rate and the volume at that step. Improving conversion at a step early in the funnel (with high volume) by 5% percentage points may produce more final conversions than improving a late step (with low volume) by 20 percentage points. The correct framework is to calculate the impact of a given improvement at each step: (current volume at step) x (improvement in step conversion rate) x (downstream conversion rate to final goal) = additional final conversions. This calculation often reveals that mid-funnel steps with high volume and moderate drop-off are the highest-leverage optimization opportunities.
- Some assume that all users who drop off at a funnel step are lost and represent failures — Not all funnel drop-offs represent lost conversions. Some users drop off because they completed their goal through a different path (called offline conversion or multi-device conversion -- they converted on mobile after researching on desktop). Some drop off because they are not qualified buyers researching for future consideration. Some drop off because the funnel step exposed a genuine misfit between what the user wanted and what the product offers -- stopping an unqualified user from converting may be appropriate (they would churn quickly or generate support costs that exceed revenue). Drop-off analysis should include exit survey research to understand the reasons behind drop-off before assuming all drop-off is negative.
For analysts
Funnel analysis is a foundational tool for any client with a defined conversion process: e-commerce checkout optimization, SaaS onboarding activation, B2B sales pipeline management, and marketing campaign performance. The diagnostic value is highest when combined with segmentation (which user types convert better or worse?) and exit research (why do users drop off at specific steps?). When setting up funnel analysis for clients, the most important decision is defining the funnel steps and time windows clearly before running the analysis -- ambiguous step definitions produce unreliable data that leads to wrong optimization decisions.
History & origin
The marketing funnel concept (AIDA: Awareness, Interest, Desire, Action) was introduced by advertising theorist E. St. Elmo Lewis in 1898, modeled on the sales conversion process. The digital version became a standard analytics concept with the rise of e-commerce and web analytics in the late 1990s-2000s. Google Analytics and its competitors made funnel visualization a standard feature of web analytics platforms in the 2000s. Product analytics tools (Mixpanel, Amplitude, Heap, Pendo) extended funnel analysis from web analytics to in-product behavior tracking -- enabling analysis of multi-session, multi-device user journeys toward product activation and engagement milestones. Sean Ellis popularized the 'growth hacking' framing of funnel optimization in the 2010s as a core startup metric practice.