Daily GA4 funnel data, July 2025 → June 2026. The decline isn't a checkout problem and it isn't one event — it's a structural shift in September that weakened the top of the funnel, compounded by a traffic step-down on May 29.
Since last July: visitors reaching a product page −10%, add-to-cart rate −29%, traffic −35%. The funnel loses people before the cart, not after. Orders are the product of all three declines.
Cart-to-purchase and checkout completion have been roughly stable for 12 months. Bottom-of-funnel CRO is not where the recovery is.
In September, collection pages went from 20% to 45% of sessions — and collection landers click through to product pages at half the old rate (40% → 21%). The funnel never recovered its July shape.
Weekly view. Watch the purple line (share of sessions on collection pages) cross above the gray line trajectory while product-page reach falls. That crossover in September 2025 is the origin of the upper-funnel weakness — it predates every later break.
Daily zoom, mid-April → June 11. Two distinct breaks, matching the May Break investigation exactly: CVR breaks May 10, traffic breaks May 29 (~33k → ~26k sessions/day, −20%, no recovery since).
Collection pages go from ~20% to ~45% of all sessions in six weeks. Product-page reach falls 45% → 30%; add-to-cart falls ~4.5% → ~3%. Whatever drove this (landing-page strategy, site navigation, paid destinations), the funnel has been structurally weaker ever since.
Transactions exceed add-to-carts (impossible). The sheet's own "Bad Removed" column blanks these days.
ATC 5.3%, CVR 3.4% peak week. The funnel can still convert when the offer is strong — evidence the leak is motivational/structural, not mechanical.
Sessions +40% for one week at 1.23% CVR (the year's worst), PLP share spiking to 45%. It ended and metrics snapped back — a preview of how PLP-heavy cold traffic behaves.
~253k weekly sessions early April → ~225k by month-end, while funnel rates hold. This is a volume story, not a conversion story.
2.95% (May 9) → 1.81% → 1.48%. Coincides with PLP share peaking at 50% of all sessions. Matches the May Break investigation's first break date.
From this date the "begin checkout" event frequently undercounts — completed purchases exceed recorded checkouts (110–130% on many days; 251% on May 27, whose 4.49% "CVR spike" is an artifact). Something changed in checkout tracking around May 18 and we are partially blind at the bottom of the funnel. Needs a trace.
Sessions step down from ~33k/day to ~26–27k/day overnight, with no recovery since. Matches the May Break investigation's second break date.
~26.9k sessions/day (vs 41.5k last July, −35%), ATC 3.16% (vs 4.45%, −29%), CVR 1.72%. Every lever is below last summer simultaneously.
| Funnel stage | July 2025 | June 2026 | Δ |
|---|---|---|---|
| Sessions / day | 41,472 | 26,900 | −35% |
| Reach a product page | 41.7% | 37.4% | −10% |
| Collection → product page click-through | ~40% | ~21% | −48% |
| Add to cart | 4.45% | 3.16% | −29% |
| Session conversion rate | 1.87% | 1.72% | −8% |
| Checkout completion | stable | stable* | ~flat |
*Checkout completion unreadable after ~May 18 due to the begin-checkout tracking break; stable through mid-May.
Source: "MUD: Performance Deep Dive" Google Sheet (GA4 export), Funnels tab — daily sessions, PLP/PDP/ATC/checkout/transaction sessions, July 1 2025 → June 11 2026. All dates fully processed (no intraday data used).
Known-bad ranges excluded or flagged: Oct 1–6 2025 (transactions > ATC), May 18+ begin-checkout undercount, May 27 CVR spike (artifact).
Related: The May Break investigation (break-date forensics) ·
CVR Pulse: Intelligems shutoff (hourly, June 11–12).