Funnel Deep Dive · Jan 2024 → Jul 2026 · Weekly

Where did the funnel decline actually happen?

Two independent trackers, week by week for 30 months: Store-side (Intelligems' pixel running on the Shopify store — independent of Google) vs GA4. Every step measured: sessions → add-to-cart → checkout → purchase. Tracking outages are shaded, not hidden — both trackers have broken windows, and several "trends" people quote are actually tag failures.

GENERATED JUL 14, 2026 · STORE-SIDE: INTELLIGEMS ON-STORE TRACKING (VISITOR-BASED) · GA4 PROPERTY 310751693 (SESSION-BASED) · SHOPIFYQL BRIDGE DOWN SINCE JUN 12 — NATIVE SHOPIFY ANALYTICS UNAVAILABLE

−45%

Session → add-to-cart

GA4 ATC-session rate, Sep '24 peak → 2026

58% → 73%

Checkout → purchase

store-side, Jan '25 → 2026 H1 — improved

54% → 68%

Cart → purchase

store-side, Jan '25 → 2026 H1 — improved

Jun 2, 2025

Sharpest CVR break

both trackers, same week (−19% store / −15% GA4)

Tracking artifacts found

see the data-quality ledger below

The verdict

The add-to-cart → conversion step did NOT decline. The decline is upstream: fewer sessions ever add to cart. Once someone adds to cart, they finish more often than they did 18 months ago — store-side cart→purchase went 54% (Jan '25) → 68% (2026 H1), checkout→purchase 58% → 73%. Both trackers agree on direction. The funnel is not leaking at checkout; it's failing to fill at the top.

Confidence: high — two independent trackers, clean windows only

The decline came in two legs, both at session→ATC. Leg 1 — Oct 2024 → Feb 2025: GA4 ATC-session rate slid 7.7% → 5.0% while ATC→purchase held flat at ~45%; CVR followed (GA4 3.4% → 2.3%, store 4.6% → 2.5%). Leg 2 — May → Jul 2025: both trackers show ATC rate down another ~20%, with the single sharpest week-over-week break in the week of Jun 2, 2025 (matches the May 29 traffic break found in the May-Break investigation). The Oct 2025 trough (GA4 CVR 1.6%) is mostly cold-traffic dilution — sessions spiked +40% that month on paid pushes.

Confidence: leg 2 high (both trackers agree) · leg 1 medium (GA4's baseline sits right after its Aug '24 tag repair)

Your analytics person is right — and it's worse than "one outage." The mid-2024 ATC outage is real and visible in both trackers, but there are seven distinct artifacts (ledger below). Practical rule for the team: never quote an ATC number from Mar–Dec 2024, and never quote GA4 checkout or ATC numbers after Sep 2025 without this caveat. The conversion-rate series is the only metric clean in both trackers for all 30 months.

Confidence: high

Two tags need fixing this week. (1) GA4 begin_checkout has been undercounting since ~May 2025 — by 2026 GA4 counts fewer checkout sessions than purchase sessions, which is impossible. (2) GA4 add_to_cart stepped down ~30% in the week of Sep 22, 2025 with no matching move store-side — a tag change, not shopper behavior. Both are GTM/pixel fixes, not CRO problems.

Confidence: high — internal consistency checks fail, cross-tracker divergence

How to read the team's "Performance Trends (YoY)" sheet against this. The sheet's headline ATC-to-CVR −80% is a measurement artifact, not a funnel change: its prior-year value reads 171% — more purchases than carts, which is impossible — because the prior-year window (early Jul '25) sits inside that pipeline's own add-to-cart outage (the one the analytics team flagged). After the fix, the ratio simply normalized. The sheet's trustworthy rows agree with this report: sessions −29% YoY (sitewide GA4: −37%) and conversion rate roughly flat (+5% there, −5% sitewide). Rule: no YoY comparisons on ATC or ATC-to-CVR until the prior-year window clears the fix date (~mid-2025); revisit ~Aug '26.

Confidence: high — a funnel ratio above 100% is self-refuting; session and CVR rows cross-validate

ATC-to-CVR by channel, YoY — outage-free version

The corrected version of the team's YoY bar chart. Windows: Jan 1 – Jul 12 of each year — chosen because the store-side pixel is fully healthy in both (no outage days to remove; the team's version compared against a 2025 baseline whose add-to-cart events were undercounted, which inflated the 2025 bars and manufactured the "−41%"). Source: store-side on-site tracking, visitor-based. GA4 cannot produce this view — its own ATC tag has been broken since Sep '25, poisoning all of 2026. TikTok is omitted (channel effectively dead in 2026: zero tracked conversions). AppLovin isn't a channel in the store-side taxonomy (folds into "Other").

ATC-to-CVR rate — clean windows, both years

Of visitors who added to cart, the share that purchased. Up in every channel.

2025 (Jan–Jul 12)2026 (Jan–Jul 12)

Channel	ATC '25	ATC '26	CVR '25	CVR '26	ATC→CVR '25	ATC→CVR '26	Δ

Every channel's ATC rate fell (that's the real decline — fewer visitors carting), while cart completion rose. The team's sheet showed the opposite because its 2025 ATC denominator was undercounted during the outage: artificially few carts → artificially high 2025 ATC-to-CVR → fake YoY collapse.

The funnel, step by step — two trackers

Store-side in violet, GA4 in green. Gray bands = windows where a tracker's tag for that metric was broken or suspect — lines inside a band should not be trusted. Levels differ by design (store-side counts unique visitors, GA4 counts sessions); read the shapes, not the gaps between lines.

Sessions / week

Raw traffic volume. GA4 sessions vs store-side sessions.

Store-sideGA4outage

Bands: store-side pixel outage Mar 11–31 '24; store-side undercount Jun 8–22 '26. Note traffic is ~25% higher in 2026 than 2024 — while conversion fell. Cold paid traffic (listicles/rise-2) dilutes every sitewide rate below.

Add-to-cart rate — % of traffic

Share of sessions (GA4) / visitors (store) that added to cart. This is where the decline lives.

Store-sideGA4outage

Bands: GA4 ATC tag broken Mar 11–Aug 11 '24 · store-side ATC undercounting Jul–Dec '24 (healing gradually) · GA4 ATC tag stepped down Sep 22 '25 → present (undercounts vs store).

Reached-checkout rate — % of traffic

Share of traffic that began checkout. Store-side cart→checkout runs ~95% and stable: almost everyone who carts starts checkout.

Store-sideGA4outage

Bands: store-side checkout tag broken before Apr '24 · GA4 checkout tag broken Mar 11–Aug 11 '24 · GA4 begin_checkout degraded May '25 → present (undercounts; by 2026 it counts fewer checkouts than purchases — impossible).

Conversion rate — purchases ÷ traffic

The one series that is clean in both trackers for all 30 months. Both agree: two-leg decline, Oct '24–Feb '25 and May–Jul '25.

Store-sideGA4outage

Bands: store-side pixel outage Mar 11–31 '24 · GA4 purchase event over-firing Apr–mid-Jul '24 (counts 1.5–1.8× store orders — inflated) · GA4 one-week purchase spike Sep 29 '25 (12.1k vs ~5k neighbors).

Add-to-cart → purchase

Shane's question: of those who added to cart, how many bought? Flat, then improving. GA4 held ~45% through May '25; store-side rose 54% → 68% into 2026.

Store-sideGA4outage

Inherits every ATC outage (bands). The 2026 store-side rise is partly real, partly arithmetic: when fewer casual browsers cart, the ones who do are more committed. Either way — this step is not the leak.

Add-to-cart → checkout

Of those who added to cart, how many reached checkout? Store-side: ~93–96%, dead flat — carting and starting checkout are nearly the same act on this site.

Store-sideGA4outage

Inherits both the ATC and checkout outages (bands). GA4's line is only meaningful Aug '24–Apr '25 (~65–75%); after May '25 its broken begin_checkout tag drags it down artificially. Trust the violet line: this step never moved.

Checkout → purchase

Of those who reached the checkout page, how many completed? Improved — store-side 58% (Jan '25) → 73% (2026 H1).

Store-sideGA4outage

Bands: store checkout broken pre-Apr '24 · GA4 checkout broken Mar–Aug '24 and degraded May '25 → present (GA4's 2026 "~100%" is the broken tag, not reality). Trust the violet line from Jan '25 on.

Phase table — what each period actually looked like

Period averages, clean windows only (— means the tracker's tag was broken for that metric in that period).

Last 13 weeks — week over week

Store-side / GA4 side by side. Δ = change vs prior week (store-side).

Data-quality ledger

Every artifact found, so nobody re-discovers these the hard way. "Fixed" = confirmed healthy after that date.

#	Tracker	What broke	Window	How it shows up	Status
1	BOTH	Site-wide pixel outage (likely a tag migration)	Mar 11–31, 2024	Store-side sessions collapse to ~1k/wk; GA4 ATC + checkout begin falling the same week	Fixed Apr 1 '24 (store)
2	STORE	Checkout-begin tracking	Jan 1 – Mar 31, 2024	Checkout rate below conversion rate — impossible	Fixed Apr 1 '24
3	GA4	add_to_cart + begin_checkout events	Mar 11 – Aug 11, 2024	begin_checkout hits literal zero in April '24	Fixed Aug 12 '24
4	GA4	purchase event over-firing	Apr – mid-Jul, 2024	GA4 purchases 1.5–1.8× store orders (normally ~1.1×)	Fixed ~Jul 21 '24
5	STORE	ATC undercount (the outage your analytics team flagged)	Jul 2024 – ~Dec 2024, healing gradually	Cart-abandonment math goes negative Jul–Sep '24; ATC→purchase reads an impossible 80–113%	Healthy by Jan '25
6	GA4	begin_checkout undercount	~May 2025 → present	By 2026, GA4 checkout sessions ≈ purchase sessions (checkout→purchase reads ~100%)	STILL BROKEN — fix
7	GA4	add_to_cart step-down (tag change, not behavior)	Week of Sep 22, 2025 → present	GA4 ATC rate drops 4.2% → 2.9% in 2 weeks; store-side moves only 3.9% → 3.3% (traffic mix)	STILL BROKEN — fix

Also: one-week GA4 purchase spike (12.1k, wk of Sep 29 '25) and a store-side session undercount Jun 8–22 '26. Native Shopify Analytics (ShopifyQL) has been unreachable since Jun 12 '26 — the store-side series here is Intelligems' on-store pixel, which tracks the same funnel independently of Google.

Method & guardrails

• Weeks are Monday-start, both trackers aligned to identical buckets. GA4 pulled daily from the Data API and re-aggregated; store-side pulled weekly from Intelligems sitewide analytics (visitor-based — rates use unique visitors as the denominator, so levels sit above GA4's session-based rates by construction).

• "ATC→purchase" = conversion rate ÷ ATC rate within the same tracker. "Checkout→purchase" = conversion rate ÷ reached-checkout rate. Cross-tracker level gaps are definitional; only within-tracker trends are meaningful.

• Confounder warning (standing guardrail): traffic mix shifted cold through this period — listicle/rise-2 paid traffic converts far below home/shop warm traffic. Part of the sitewide ATC-rate decline is dilution, not a worse site. The Weekly Funnel Audit tracks the mix-controlled version of this story; the two-leg timing (Oct '24–Feb '25, May–Jul '25) survives in both.

• Change-points found by 8-week before/after mean-shift scan, cross-checked between trackers; year-boundary "breaks" (BFCM → January) were discarded as seasonal.