Weekly Funnel Audit

Week of Aug 10 – Aug 16

Every step of the acquisition funnel, 13 weeks deep, with this week's CAC change traced to the step that caused it. Generated Aug 17, 9:42 AM PT.

CAC rose $2.71 to $124.60 as site conversion fell to a 13-week low of 1.281%, only partly offset by cheaper media at $28.26 eCPM.

NC CVR erosion drove CAC up +$11.04 to $124.60, only partly offset by cheaper media (eCPM -$7.76); leverage (-$0.70) and CTR (+$0.13) were negligible.

Callouts

🔴 Homepage CVR broke its band — diagnose this week. Warm homepage dropped 2.24% → 1.815% (-19%, 13-week low) on 19.3k sessions, adding the bulk of the $11.04 CVR hit to CAC. Pull homepage module/AB changes shipped Aug 10–16 and roll back anything post-Aug 9 that touches above-the-fold.
🔴 Rise-2 lost 11.6% CVR on 34.9k sessions — refresh the cold LP now. 1.352% → 1.195% while paid sessions grew 15.3%; cheaper Meta inventory is landing on a page that can't hold it. Ship a new rise-2 variant this week or reallocate the 42% paid share to listicle (1.387%, flat).
🟡 Hold Meta budget flat at $160K — click CAC $232 vs survey $217. Two lenses agree Meta is the pressure point, but the driver is site not media (eCPM at 13-week low). Do not cut until homepage/rise-2 fixes ship; re-read Meta CAC next week against a repaired funnel.
🟡 YouTube stays in at $21K — click CAC $303, survey-implied $102. 11.5% survey share on 812 responses is consistent with the 12-month 0.46× pattern. No cut on clicks alone; hold spend and revisit on the 4-week view.
⚪ Ignore the product-page and checkout-step alarms. Product CVR 4.504% → 3.138% reverts last week's spike to the 13-week median. ATC +42%, checkout -13.6%, CO2P -15.8% sit on GA4 tracking degraded since May 18 — directional only.

Checkout-funnel steps unreliable since GA4 begin_checkout broke May 18; NC CVR (Shopify NCs ÷ GA4 sessions) is trustworthy.

Watch next week

Homepage CVR back inside its 2.0–2.5% band on ≥18K sessions — if it holds sub-2% a second week, the break is structural, not a shipping artifact.
Rise-2 CVR vs listicle at 42%+ paid share — if rise-2 stays below 1.25% while paid sessions hold above 60K, the cold-LP refresh is the pacing item, not Meta cost.
NC CVR direction off 1.281% — a third consecutive weekly print below 1.35% resets the baseline and forces the CAC-target conversation regardless of eCPM tailwind.

Decomposition

Why CAC moved this week

Exact arithmetic, not attribution: blended CAC is the product of these four steps, so their contributions sum to the week's CAC change to the cent. Method in the appendix below.

Step by step · 13 weeks

The acquisition chain

Step by step · 13 weeks

On-site: where sessions become customers

Cross-checks

Measurement health

Agent analysis

The story of the week

CAC rose $2.71 WoW to $124.60, but the headline understates the churn underneath: site conversion added $11.04 while eCPM saved $7.76. This is an offsetting week — media got cheaper (eCPM $30.10 → $28.26, a 13-week low) but the site converted worse (NC CVR 1.402% → 1.281%, also a 13-week low). CTR held flat at 0.712% and leverage barely moved at 2.486x, so the story is entirely media-cost vs site-conversion.

Tracing the CVR drop: rise-2 fell 1.352% → 1.195% (-11.6%) on 34.9k sessions, and homepage collapsed 2.24% → 1.815% (-19.0%, a 13-week low breaking its ~2.0–2.5% band). Product-page CVR also fell 4.504% → 3.138% (-30.3%), but that reverts a one-week spike and lands near the 13-week median — noise, not signal. Listicle was flat at 1.387%. So the CVR damage concentrates in rise-2 (paid cold LP) and homepage (warm).

The traffic-quality split confirms both pools weakened on their own terms: paid CVR 1.36% → 1.28% and warm CVR 3.08% → 2.11% (per lenses.trafficQuality). Warm's drop is the larger surprise — homepage-driven — and cannot be blamed on paid mix. On the media side, cheaper eCPM with flat CTR is consistent with the July–August regime of lower-cost, lower-quality inventory; clicks rose 5.0% and paid sessions rose 15.3% on the same spend, so the funnel top is doing its job.

Regime context: CAC has bounced in a $108–$136 band for 13 weeks with no trend; this week ($124.60) sits mid-band. But NC CVR at 1.281% is the lowest print in the series and extends a slow bleed from 1.60% (Jun 29) — worth watching, not yet a break.

Confounders checked

Traffic mix: paid share of sessions rose 38.4% → 42.0%, which mechanically drags blended CVR down since paid converts at ~1.28% vs warm ~2.11%. But both buckets fell against their own prior week (paid -5.9%, warm -31.5%), so mix explains only part of the blended drop — the warm-pool decline is real.
Measurement — GA4 begin_checkout degraded since May 18: checkout rate (-13.6%), ATC (+42%), and CO2P (-15.8%) are all flagged but unreliable and cannot be used to localize the CVR loss to a funnel stage. NC CVR itself uses Shopify D2C NCs ÷ GA4 sessions, so the top-line CVR number is trustworthy.
Bounce snap-back: paid bounce jumped 11.13% → 19.36% and warm 15.78% → 23.13%, both reverting toward the pre-July norm after unusually low July prints. Consistent with the LP CVR softness but does not prove causation.
Product-page noise: the -30.3% product CVR drop follows a +35% spike the prior week and lands at the 13-week median — treat as reversion, not a new problem.

Regime changes

Events ledger

Numbered flags appear on every chart above. To add an event, edit reports/funnel-audit-events.json in the marketing-pulse repo — anything that could move a funnel step belongs here (site changes, tracking changes, promos, data outages).

Method appendix — step definitions & decomposition

The identity: CAC = spend/NC = (spend/impressions) × (impressions/clicks) × (clicks/sessions) × (sessions/NC) — so a change in CAC is exactly the sum of the four steps' log-changes, converted to dollars with the log-mean. Contributions reconcile to the cent by construction; the generator refuses to publish if they don't.
Spend & New Customers: Daily Spend Tracker — the canonical blended-CAC source (spend ÷ Shopify D2C new customers). Never a sum of Northbeam channel CACs.
eCPM & CTR: Northbeam paid media (Clicks Only model, cash accounting), Amazon excluded. eCPM uses tracker spend over NB impressions, so it's an effective blended rate, not a platform-reported CPM.
Traffic leverage: GA4 sessions ÷ NB paid clicks. Rises when organic/warm traffic carries more of the load per paid click; falls when the site depends more on paid.
NC CVR: Shopify D2C new customers ÷ GA4 sessions. Never Omni Total Orders (includes renewals).
Paid vs warm: Paid = rise-2 + listicle landers (cold prospecting); warm = everything else. Different intent pools — each is compared to its own history, never to each other.
Checkout steps: GA4 begin_checkout degraded since May 18, 2026 — those two panels are directional only until tracking is repaired.