Cohort Funnels
Cohort funnels break funnel performance down by acquisition source or time window, revealing whether your conversion rate is genuinely improving or just being masked by traffic mix shifts.
Cohort Funnels
Funnel conversion performance broken down by acquisition cohort, signup window, or campaign — so trends aren't hidden by traffic-mix shifts.
A cohort funnel is the same step-by-step conversion funnel you already track (view → add-to-cart → checkout → purchase), but segmented by the cohort a visitor belongs to: the week they first landed, the channel that acquired them, the campaign they came from, or whether they're new or returning. Each cohort gets its own funnel; you compare them side by side.
The value is in what it surfaces. A flat sitewide funnel can stay at 2.1% conversion for six months while your paid-social cohort silently collapses from 1.8% to 0.9% — masked by an improving email cohort. Cohort funnels expose that split so you can act on the cause, not the average.
Cohort funnels are a specific lens inside broader funnel analytics. Where a standard funnel answers "how many of everyone who hit step 1 reached step 4," a cohort funnel answers "how many people from THIS group reached step 4 — and how does that compare to last month's group?"
The cohort marker depends on what you're investigating. Acquisition source is the most common (Meta vs Google vs organic vs email), but signup-week cohorts catch product-quality drift, campaign cohorts isolate creative performance, and device cohorts surface mobile-only regressions after a theme update.
Cohort step conversion = Cohort users at step N / Cohort users at step 1
Cohort users at step N
Users from cohort reaching step N
Count of unique users belonging to the cohort who completed step N of the funnel within the observation window.
Cohort users at step 1
Cohort entry size
Count of unique users in the cohort who completed the funnel's entry step (usually a landing-page view or session start).
A Shopify apparel store compares its September Meta-acquired cohort against August's Meta cohort on the checkout step.
August Meta cohort — sessions: 18400
August Meta cohort — reached checkout: 644
September Meta cohort — sessions: 21000
September Meta cohort — reached checkout: 504
→ August: 3.5% → September: 2.4%
Sitewide checkout-reach was flat at 3.1% both months, hiding a 31% relative drop in the Meta cohort. The likely cause is creative fatigue or a broader audience expansion pulling in lower-intent traffic — not a checkout problem.
Read cohort funnels as ratios, not absolutes. A cohort with 10,000 entries and 1.5% checkout-reach contributes more lost revenue than one with 800 entries at 0.6% — but the smaller cohort's drop is what tells you something has changed. Volume tells you where the money is; rate tells you where the diagnosis is.
Typical funnel performance by acquisition cohort — online retail, AOV €40-€120
| Acquisition cohort | Add-to-cart | Reach checkout | Complete purchase |
|---|---|---|---|
| Direct / branded search | 12-16% | 6-9% | 3.5-5.0% |
| Organic search (non-brand) | 8-11% | 3.5-5.0% | 1.8-2.6% |
| Email (existing list) | 14-20% | 8-12% | 4.5-7.0% |
| Paid search (Google) | 9-12% | 4.0-5.5% | 2.0-2.8% |
| Paid social (Meta, prospecting) | 5-8% | 1.8-2.8% | 0.7-1.4% |
| Paid social (Meta, retargeting) | 11-15% | 5.5-7.5% | 2.8-4.0% |
Use the gap between cohorts as your diagnostic. If your paid-social prospecting cohort sits at 0.5% purchase rate while retargeting clears 3.5%, the funnel works — prospecting audience quality is the constraint. If retargeting also collapses, the checkout itself is the problem. Cohort funnels turn "conversion is down" into a specific, testable hypothesis.
Cohort funnels FAQ
A regular funnel aggregates every visitor into one curve. A cohort funnel splits that curve by a cohort marker — acquisition source, signup week, campaign, device — and shows one funnel per cohort so you can compare them directly.
Acquisition source. It's the variable most likely to drive funnel performance differences (intent quality varies massively between branded search and cold paid social) and it's the easiest to act on once you spot a problem.
Match the window to your purchase cycle. For most online stores, weekly cohorts work for tactical monitoring and monthly cohorts for strategic trends. Daily cohorts are too noisy below ~5,000 sessions/day.
Aim for at least 1,000 sessions and 30+ conversions per cohort before drawing conclusions. Below that, normal variance can look like a real trend. For smaller cohorts, widen the time window rather than trust noisy data.
No — they complement it. The sitewide view tells you total revenue impact and overall direction; cohort funnels tell you which group is driving the change. You need both to prioritise fixes.
They're a segmentation lens applied on top of standard funnel analytics. The funnel steps are the same; the cohort dimension is what's added. Most modern analytics tools support cohort funnels natively once your event tracking is clean.
Traffic-mix shifts. If a high-converting cohort grows while a low-converting one shrinks, the average can stay flat even when both cohorts are individually getting worse. Cohort funnels are the only way to catch this.
For the test itself, no — randomisation handles cohort mix. But after a test wins, cohort funnels are useful for post-hoc analysis: does the lift hold across all acquisition sources, or only in retargeting traffic?
Returning visitors usually convert 2-4x better than new visitors on the same store. Splitting them is essential — a shift in new/returning mix (e.g. after a paid-spend pullback) will move sitewide conversion without anything on-site actually changing.
Comparing cohorts of very different sizes without accounting for variance, and reacting to one week's data. Always check whether the gap between cohorts is larger than the week-to-week noise within a single cohort before declaring a trend.
Get an AI expert review of your site
Paste your URL — Metricuno's AI runs the same heuristic checks a senior CRO consultant would, scoring your page and prioritising the fixes that'll move conversion fastest.