Cohort Funnels
Cohort funnels break conversion rates down by acquisition window or source, revealing whether funnel performance is improving or quietly decaying over time.
Cohort Funnels
A funnel view that segments conversion performance by acquisition cohort, channel, or time window — exposing trends sitewide funnels hide.
A cohort funnel takes the standard step-by-step conversion funnel (landing → product → cart → checkout → purchase) and splits it by a cohort marker: the week a visitor first landed, the channel that acquired them, the campaign they clicked, or the device they used. Each cohort gets its own funnel, and you compare the drop-off rates across them.
The purpose is diagnostic. A sitewide funnel showing 2.4% checkout conversion is an average — it can stay flat for months while paid-social cohorts collapse from 1.8% to 0.9% and organic cohorts climb from 2.6% to 3.5%. Cohort funnels surface that movement before it shows up in monthly revenue.
Cohort funnels are a specific lens within funnel analytics. Where a standard funnel answers "how does the average visitor move through checkout?", a cohort funnel answers "is this week's traffic behaving differently than last week's — and where exactly does the gap open up?"
The most useful cohort markers for an online store are acquisition week (catches creative fatigue and seasonality), acquisition channel (isolates paid traffic quality from organic), first-touch campaign (links funnel health to a specific ad set), and device type (separates mobile checkout pain from desktop).
Cohort Step Conversion = Users_in_cohort_reaching_step_N / Users_in_cohort_reaching_step_N-1
Users_in_cohort_reaching_step_N
Users at step N
Distinct users from the cohort who reached the current funnel step.
Users_in_cohort_reaching_step_N-1
Users at previous step
Distinct users from the same cohort who reached the previous step.
A Shopify apparel store compares the May 6 paid-social cohort to the May 13 paid-social cohort at the add-to-cart → checkout step.
May 6 cohort reaching add-to-cart: 4,200
May 6 cohort reaching checkout: 1,470
May 13 cohort reaching add-to-cart: 3,900
May 13 cohort reaching checkout: 1,053
→ May 6 step conversion = 35.0%. May 13 step conversion = 27.0%.
Cart-to-checkout dropped 8 percentage points week-over-week for paid-social traffic — a creative or audience shift, not a sitewide problem. The sitewide funnel barely moved because organic and email cohorts held steady.
Read cohort funnels by step, not by total conversion. The interesting question is which step the cohorts diverge at. If two cohorts have similar product-page conversion but split at add-to-cart, the issue is product perception or price for one of them — not checkout UX.
Typical step-conversion ranges by acquisition cohort — Shopify apparel store, monthly cohorts
| Cohort | Landing → Product | Product → Cart | Cart → Checkout | Checkout → Purchase |
|---|---|---|---|---|
| Organic search | 58-68% | 14-18% | 55-65% | 68-75% |
| Branded paid search | 62-72% | 16-22% | 60-70% | 70-78% |
| Paid social (cold) | 38-48% | 6-10% | 30-42% | 55-65% |
| Email / SMS | 70-80% | 20-28% | 65-75% | 75-82% |
| Referral / affiliate | 50-60% | 10-14% | 45-55% | 65-72% |
The pattern that matters: paid-social cold traffic and email cohorts can both finish at roughly the same final purchase rate, but they get there through completely different step profiles. Optimising the checkout for the average visitor improves neither — you need cohort-specific fixes.
Cohort funnels FAQ
A regular funnel reports one set of step conversions across all visitors. A cohort funnel runs that same calculation separately for each group — by week, channel, campaign, or device — so you can compare the step-by-step shapes against each other instead of looking at an average.
Acquisition channel. It almost always produces the widest divergence in funnel shape, and it maps directly to where your media spend goes. Once you've split by channel, layer in acquisition week to catch creative fatigue inside each channel.
Three to five. Two cohorts can't tell you whether a gap is a trend or noise; more than five becomes a wall of numbers. For weekly cohorts, the trailing four weeks plus the same week from last year is a strong default.
Aim for at least 1,000 sessions per cohort reaching the first funnel step, and at least 30-50 conversions at the final step. Below that, the step-conversion ratios swing wildly and you'll chase noise. Smaller stores should widen the cohort window from weekly to bi-weekly or monthly.
Partially. GA4's Funnel Exploration supports a single breakdown dimension, which gives you a basic cohort split. It struggles with multi-dimensional cohorts (channel × week), retroactive cohort definitions, and stable user counts across long time windows due to sampling and thresholding.
Retention cohorts track returning behaviour over time (week 1, week 2, week 4 repeat purchase). Cohort funnels track a single conversion journey, segmented by who entered it. They're complementary — retention cohorts tell you which acquisition cohorts are worth keeping; cohort funnels tell you why some cohorts convert worse on day one.
For a stable channel like organic, expect ±2-3 percentage points week to week on a mid-funnel step. Paid social can move ±5-8 points as creative rotates. A swing larger than that — especially one that persists for two consecutive weeks — is usually a real change, not variance.
First-touch for acquisition-quality questions ("is the traffic we're buying getting worse?"). Last-touch for conversion-path questions ("which channel closes the sale?"). For cohort funnels diagnosing funnel decay, first-touch is the default — you want to hold the entry source constant across the journey.
They tell you whether a winning variant won uniformly or only inside a specific cohort. A checkout test that lifts conversion 8% sitewide might be +14% for paid-social and flat for organic — meaningful for how you'd roll it out and which audiences to expand.
Weekly for paid-channel cohorts, where creative and audience shifts move fast. Monthly is enough for organic and email cohorts. The point isn't a recurring report — it's catching a divergence in the week it starts, before it becomes a quarter of lost revenue.
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.