What needs to be finished first
The conversion to be analyzed (for example sign-up, purchase, activation) is clearly defined with a start event and target event.
The key stages of the customer journey are documented so funnel steps reflect real user flow.
What needs to be ready before start
BI or product analytics tool with access to event data; table for steps (step, definition, event ID, n, conversion rate, drop); template for segment comparison.
A data analyst with tool access; a product owner responsible for the flow; a marketing or UX representative with hypotheses; optional stakeholder for validation.
Event tracking plan with definitions; time range (for example last 30 or 90 days); segmentation axes (channel, device, plan); known technical drop-offs.
2-3 h
Create funnel report in the tool with defined steps in sequence. Set time window and minimum sample size (at least 500 users per step for stable rates). Add baseline comparison period.
The one question this method answers
At which steps do users lose conversion in each segment, and which hypothesis explains the drop?
Marker: Phase
| Step | Duration | Action | Hint |
|---|---|---|---|
1Phase 1: Verify steps | 30 min | Check step definitions against tracking. Sanity check: step n+1 cannot have more users than step n. Clean duplicates and multi-triggers. | Poor definitions create pseudo drops. If login fires multiple events or pop-up counts as page view, funnel is distorted. |
2Phase 2: Calculate conversion rates | 30 min | For each step, show absolute count, conversion relative to previous step and to start, plus absolute and relative drop. | Show three values at once: conversion to previous, conversion to start, absolute loss. One number alone hides where leverage lies. |
3Phase 3: Segment | 45 min | Split funnel by segment (channel, device, plan, region), and mark top three deviations. Prioritize segment with largest absolute impact. | A weak segment can drag the overall funnel. Mobile iOS can show very different values from desktop Chrome. Without segmentation, you optimize the wrong average. |
4Phase 4: Hypotheses at largest drop | 45 min | Collect three to five hypotheses per top drop with explanation and validation plan (session replay, heatmap, logs, interview). Prioritize hypotheses by ICE. | Formulate testable hypotheses (for example "Users do not understand price information") rather than vague statements. |
5Phase 5: Follow-ups | 20 min | Define owner, validation method, and deadline per top hypothesis. Set funnel report as recurring weekly review. Record tracking changes as tickets. | A funnel without cadence becomes one-off analysis. Weekly review with top-3 drops and hypothesis status keeps it useful. |
What comes out at the end
Funnel report with step list, conversion rates, drop values, segment breakdown, top hypotheses, and follow-up table with owner and deadline. Add baseline period comparison and notes on tracking changes.
Version funnel configuration in the tool (naming convention with date). Archive weekly snapshots as Markdown or PDF. Create a new funnel for tracking changes and archive the old one with reason.
Compact working template for Funnel Analysis with context, input, output artifacts, and next step.
Analyzes conversions and drop-offs between defined steps.
When and for what do we use this method?
Which data, observations, decisions, or materials are available?
Short notes along the runsheet.
Owner, date, and success signal.
Concrete filled scenario, fictional example
Concrete filled scenario, fictional example
| Step | Definition | Users | Conv-to-Prev | Conv-to-Start | Drop |
|---|---|---|---|---|---|
| 1 Visit | Landing page loaded | 18,400 | - | 100% | - |
| 2 Search | Search query sent | 9,620 | 52% | 52% | 8,780 |
| 3 Detail | Workspace detail opened | 6,110 | 64% | 33% | 3,510 |
| 4 Booking start | Booking form opened | 2,040 | 33% | 11% | 4,070 |
| 5 Booking confirm | Booking confirmed | 1,260 | 62% | 6.8% | 780 |
Top drop: Step 3 → 4 (33% conv-to-prev, 4,070 users in absolute terms). Mobile iOS drops to 24%, desktop stays at 41%.
Hypotheses
Next steps: Session replay (10 mobile sessions, Owner @anna, by 22.05.), A/B test for CTA position (Owner @ben, from 27.05.).
Recognize symptoms and steer against them
Later step shows more users than previous step or conversion over 100%.
Enforce event deduplication per session. Recheck tracking plan. If the tool does not offer dedupe, use SQL distinct by user ID per step.
Overall conversion seems acceptable while individual channels or devices show very weak rates.
Require segmentation by channel, device, and plan. Standard funnel report includes at least three segments side by side.
A list of guesses with no one assigned to verify them.
Define validation methods before collecting hypotheses (heatmap, session replay, interview). Require method and owner per hypothesis before adding to backlog.
Users bounce back and forth between steps or skip steps, but linear funnel does not show it.
For branched flows, add path analysis (Sankey or flow report). Keep linear funnel for main path and analyze branches separately.
Rates decline and the team panics; the cause is seasonality or tracking updates.
Choose equal-length and same-window comparison periods. Keep tracking changelog visible in dashboard. Verify whether deployment changed tracked events before alarm.
Done signals checkable in under a minute
Go to the profile for purpose, similar methods, and sources or continue to the next method in the catalog.