Loading tutorials…
Loading tutorials…
Funnels are the most-used chart type in Amplitude and the most-misused. The defaults assume an attribution window that doesn't match most SaaS sales cycles. Here's how to build funnels that match reality.
Who this is forProduct, growth, and marketing teams analyzing conversion bottlenecks. Especially relevant for SaaS teams running onboarding optimization, paid acquisition analysis, or activation experiments.
What you'll need
Step 1
Amplitude → Charts → "New Chart" → select "Funnel Analysis." This is different from the deprecated "Funnels" tab — use the Charts version (2026 UI).
Navigate to Amplitude → Charts → "New Chart" (top-right). The 2026 UI consolidates Funnels under the Charts surface — the old standalone Funnels page is being deprecated.
Select "Funnel Analysis." You'll see a builder with steps, conversion window, and segmentation options.
Name the funnel descriptively from the start: "Web Signup → Activation → Paid Conversion, 30d, Free Trial source." Generic names ("My Funnel 1") become unsearchable in week two.
Save the chart into a Notebook (Notebooks tab) tied to the project area — e.g., "Q2 Onboarding Funnel Analysis." Free-floating charts get lost.
Step 2
Click "+ Add Step" for each event. Order them as they happen in your real customer journey, not as your code fires them.
Step 1: usually an awareness/entry event. "Page Viewed" on landing page, or "Signup Submitted" for a tighter funnel.
Steps 2-N: intermediate states. "Email Verified," "Workspace Created," "First Feature Used."
Final step: the conversion. "Subscription Started," "Demo Booked," or your North Star action.
Keep it to 4-7 steps. Beyond 7, every step adds ~5-10% noise and the funnel becomes unreadable.
For each step, you can add property filters. Example: "Page Viewed" where `page_path = /pricing`. Use this aggressively — it scopes the funnel to the right journey.
Step 3
"Ordered" requires users to do steps in exact sequence. "Unordered" counts any user who eventually did all steps. The default is Ordered — but Unordered is often more honest.
Find the "Ordered / Unordered" toggle in the funnel settings panel.
Ordered: A → B → C means the user must do A, THEN B, THEN C. If they did C first then A, they don't count. Use for strict sequence analysis (onboarding flow).
Unordered: counts any user who did A, B, and C in any order within the conversion window. Use for "did the user eventually convert?" analysis.
Trap: most onboarding funnels are set to Ordered but your actual UX allows users to skip steps. The funnel undercounts by 15-30% because real users explore non-linearly.
Rule of thumb: for marketing-attribution funnels (Ad → Visit → Sign Up → Pay), use Ordered. For product-usage analysis (did the user adopt features A, B, C?), use Unordered.
Step 4
Default conversion window is 24 hours. For most SaaS that's far too short. Match the window to your actual sales-cycle length.
Find "Conversion Window" in the funnel settings.
B2C impulse purchases: 1-7 days. Defaults work.
SaaS with free trial: 14-30 days. The trial-to-paid transition is the long tail.
B2B SaaS with sales-led close: 30-90 days. A 7-day window will report 5% conversion when reality is 25%.
Enterprise: 90-180 days. Yes, really. Amplitude supports up to 365 days.
When in doubt, build the same funnel at three windows (7d, 30d, 90d) side-by-side. The "right" window is the one where conversion rate stabilizes — adding more days stops improving the number.
Step 5
Click "Group by" or "Filter" to slice the funnel by user property or event property. This is where the actionable insight lives.
Group by `utm_source` to see funnel conversion per acquisition channel. Often the biggest aha — paid channels usually convert 30-60% lower than organic.
Group by `plan_tier` (free / pro / enterprise) to see whether high-intent users convert better. (Spoiler: yes, usually 2-5x.)
Group by `device_platform` (web / iOS / Android) — mobile funnel conversion is usually 20-40% lower than web for SaaS.
Filter by cohort: pre-built cohort like "Active Users L30" applied as a funnel filter shows only that group's funnel. Useful for "do our power users have a different drop-off pattern?"
For paid campaign analysis, group by `utm_campaign` and filter by `utm_source = google`. Compare campaign-level CAC against funnel completion to find the inefficient campaigns.
Step 6
Amplitude shows conversion % at each step. Find the step with the biggest absolute drop and the biggest relative drop — these are different.
Absolute drop = how many users were lost in raw count. Relative drop = % of those who entered that step who left before the next.
Biggest absolute drop is usually early-funnel (most users haven't converted yet, so raw count is large).
Biggest relative drop is the actual bottleneck — the % is the leak rate.
Click on any step → "Users at this step" → drills into a user list. Open 5-10 users in User Lookup to see what they did before dropping off.
For each top drop step, generate 2-3 hypotheses. Then test ONE hypothesis (UX change, copy change, gate removal) and re-measure the funnel after 14 days. Don't change three things at once or you can't attribute the lift.
Step 7
Drop the funnel chart into a Notebook for context. Set a threshold alert on the conversion rate so you find out when it shifts.
Click "Save to Notebook" → pick the right Notebook (e.g., "Onboarding Funnel Tracking"). Add a 1-3 sentence note explaining what the funnel measures and what "good" looks like.
In the chart settings → Alerts → Add alert: "Notify me when conversion rate drops by 20% week-over-week." Slack + email options.
Annotation feature: add an annotation on the chart for every product change (Settings → Annotations → "New annotation"). Six months later you can see "Why did conversion drop in week 7?" — and the annotation says "Removed credit-card-up-front in trial."
Share the Notebook link in #product or #growth Slack channels. Charts that are seen by 5+ people get maintained; private charts rot.
Common mistakes
Using Ordered funnel for non-linear product flows
What goes wrong: Real users explore your product non-linearly. An Ordered funnel undercounts conversion by 15-30%. You make UX decisions based on a metric that overstates the drop-off — typically wasting 40-80 hours of eng time on the wrong problem.
How to avoid: For product-adoption funnels, default to Unordered. Only use Ordered when the sequence is genuinely required (legal forms, multi-step wizards).
Conversion window too short for the sales cycle
What goes wrong: B2B SaaS with 60-day sales cycle measured on a 7-day window shows 4% conversion. Reality is 22%. CMO concludes the funnel is broken and rebuilds onboarding — wasting $10K-30K in eng time fixing a measurement artifact.
How to avoid: Match window to actual cycle. When unsure, build at three windows side-by-side (7d, 30d, 90d) and pick the one where the rate stabilizes.
Mixing platforms (web + mobile) in one funnel
What goes wrong: Users who signed up on web and converted on mobile look like two separate users (if identity is broken) OR like one user doing the same event twice (double-count). Either way the funnel is wrong by 10-25%.
How to avoid: Filter the funnel to one platform OR fix identity resolution so cross-platform users merge. The latter is a 4-8 hour SDK fix.
Not grouping by acquisition source
What goes wrong: You see 18% funnel conversion overall and think it's fine. Hidden: organic converts at 35%, paid at 6%. You keep spending on paid because aggregate looks acceptable. Typical waste: $2K-15K/mo depending on spend.
How to avoid: Always view funnels grouped by `utm_source` for any product with paid acquisition. The aggregate number hides the actionable insight.
No annotations for product changes
What goes wrong: Six months later, conversion has shifted twice and nobody remembers why. Was it the pricing change in March? The onboarding redo in April? You re-run analysis trying to figure it out — typical investigation cost: 8-20 hours of analyst time.
How to avoid: Every product/marketing change gets an Annotation on the relevant chart. Takes 30 seconds. Saves days of forensic work later.
Building too many steps
What goes wrong: A 10-step funnel is unreadable. The eye glazes over and the chart stops getting checked. Effectively the funnel doesn't exist in the team's decision-making.
How to avoid: Keep funnels to 4-7 steps max. If you need finer granularity, build a second drill-down funnel from one of the steps.
Recap
Done — what's next
How to set up Amplitude event tracking the right way
Read the next tutorial
Hand it off
Reading funnels is a skill that compounds. The first 10 funnels you build will be 60% right; the next 90 will be 90% right. A specialist who has built 500+ funnels brings that learned judgment immediately — typically $30-100 per funnel build, or $300-700/mo for ongoing analysis at $14-16/hr.
See specialist rates
Three usual culprits: (1) conversion window too short — Stripe sees the user pay 45 days after signup, Amplitude's 30-day window misses them. (2) Identity resolution broken — same user has multiple `user_id`s. (3) Server-side payment events not firing to Amplitude. Audit each.
Yes. As long as server events are sent via the Node SDK with the correct `user_id`, Amplitude treats them identically to browser events in funnels. Common pattern: web "Checkout Started" → server "Subscription Started" (from Stripe webhook) → web "Onboarded."
Funnels measure conversion through a defined sequence. Pathfinder (separate Amplitude feature) shows ALL paths users take, not just predefined ones. Use funnels for hypothesis testing; use Pathfinder to discover unexpected user flows.
Add `revenue` or `mrr` as an event property on your conversion event ("Subscription Started"). Then in the funnel, switch the "Measured by" dropdown from "Uniques" to "Sum of property" → choose `revenue`. Now the funnel shows $ converted, not just users.
At least one full conversion window of data. If your window is 30 days, wait 30 days post-launch before reading the funnel. Earlier readings have right-truncation bias — recent users haven't had time to convert.
Amplitude
Bad event tracking is the most common reason Amplitude projects fail. Here is the naming convention, the SDK code, and the Data Guard rules that keep your taxonomy clean for years — not weeks.
Amplitude
Retention is the single most important SaaS metric. But Amplitude has three retention models and they answer different questions. Pick the wrong one and you'll think you're growing when you're actually leaking users.
Amplitude
Funnels show you the path you expected. Pathfinder shows you the paths users actually take. The gap is where 60% of product insight lives — and where most teams never look.
Amplitude
DIY Amplitude is a great idea — until your taxonomy gets out of control or your charts disagree with reality. This is the honest framework for when the math flips toward hiring.