the brief
press enter
Three forces compounded: media auction inflation (paying more per click, CTR flat), a Meta pullback with Google reallocation, and a test program that is rerouting the funnel — the canonical Rise 2 page now sends 24% of engaged visitors to other content pages instead of products (was 10% in April).
Daily blended CAC (black, right axis) and new customers per day (indigo, left axis), 7-day averages, with every known intervention marked. The two adjudicated break dates — May 10 (CVR) and May 29 (traffic) — each sit within a day of a known event.
Every bar is 100% of the traffic that hit the rise-2 URL in a two-week window. Teal = got the real page and stayed on the expected path (includes the ~10.2% who always clicked to other content by choice — shown as its own slice). The growing rose + purple wedge is the diversion: redirected to a variant URL before load, or pushed off the real page after load — the surge tracks the URL tests exactly. Explore it interactively →
Each of these was a plausible suspect. Each was checked against clean pulls and cleared — useful because every one of them narrows where the real problem can live.
flat-to-better on canonical Rise 2 (62% → 56-58%); ad traffic quality did not collapse
ATC→checkout ~92-95% flat all quarter (checkout→buy dipped to 51.6% only in the final partial week — watch, small sample)
per-PDP-visitor ATC propensity unchanged (13.6-15.4%)
no structural drop (30d avg rose 1.56%→1.67%; only a 2-week slide 1.9%→1.5% at the tail worth a creative-fatigue look)
CLS 0.72 / LCP 3.4s are chronically POOR (a standing conversion tax, ~45% of traffic in slow FB/IG in-app browsers) but did not worsen in May-June, so they don't explain the recent dip
Three data-ranked hypotheses from the diagnostic, each with its evidence and its honest counter-argument — plus two confirmed organizational misfires surfaced in the Head of Growth pre-read. Confidence is stated explicitly; nothing here is causally proven until the kill-switch read comes back.
Turned off for the Strawberry Matcha LTO launch by the brand side without growth sign-off (team silo). Hidden because the new Alia popup elevated total email/SMS capture, masking that PDPs were no longer feeding the list.
Program review optimized for ROAS over scale; media-buyer affiliates (higher CPA than coupon affiliates) were cut May 9. Cost ~1,000 NCs in May.
Hold April’s routing mix (25.3% of rise-2 visitors to product pages, 10.2% to content) and keep June’s own per-path conversion rates — defensible because per-path rates were stable in all seven windows. This isolates what routing alone is costing.
Connor’s full pre-read covers the media-side program: what the Q1 audit flagged and fixed, and what new issues Q2 surfaced. Status as of Jun 11. Read the original in Notion →
| problem | action + status | |
|---|---|---|
| Meta attribution on 1-day click — model built on ~20% of the journey, optimized toward existing customers | Reverted to 7-day click; Northbeam (7DC + deterministic views) is the single source of truth | |
| Budget decisions driven by in-platform Meta reporting (credited organic converters to paid) | Northbeam is now the record for all CAC and optimization decisions | |
| Meta placement mix drifted to low-intent inventory; Search/Explore/Profile cut to near-zero | Feed recovered to 43–47% of spend, FB Reels down to 25% from 36% peak; Search prioritized via its own placement campaign | |
| GeistM and TQE driving cheap, non-converting traffic (organic cannibalization) | Both cut immediately (off completely mid-May); budget reallocated to Google/Meta | |
| Podcast spend cut 57% YoY without performance rationale | Normalized to $3–4K/day | |
| YouTube stuck in perpetual learning phase from erratic budget changes | Relaunched with clean structure, back to $5K/day | |
| Google Shopping offline since 2025 (Merchant Center disapproval: health-claims language, cordyceps species unidentified) | Listings rewritten ingredient-first, species identified, trigger language removed, citations added; resubmitted — target reinstatement end of June | |
| CRM list contaminated by Q1 traffic; welcome-flow RPD collapsed $0.50 → $0.09 | Jan–Mar unengaged suppression lands in June; welcome + abandon flows rebuilt (mostly complete 6/11) | |
| Affiliate channel over-indexed on BOF — attribution noise + cannibalization | Two erroneously-paused media buyers reinstated; six high-volume affiliates added in June |
| problem | action + status | |
|---|---|---|
| Meta CPM inflation across ALL placements at once — blended CPM $42 in May, an 18-month high; CPO $118 May → $125 June | Creative rebuild underway: new UGC creator roster, revised briefs. Primary lever, but slow to show in data. Influencer creators scaling well; statics lagging | |
| Creative burning out faster than it can scale — winners exhaust their matched audience, Meta moves them down the intent curve; more spend accelerates burnout | Briefs rebuilt around buyer identity (wellness-stack, not price-sensitive); ABO audience tests from 1P data launching this week | |
| ASC (Advantage+ Shopping) over-leveraged; performance deteriorated | Spend shifted to ABO where audience inputs can be controlled (ASC still active) | |
| PDP view rate fell 45.2% in June vs 59.7% March peak (GA4) — a growing share of sessions never reaches a product page | Email capture added to LPs and PDPs to recover non-converters; A/B testing the impact on bounce + CVR | |
| Demand Gen / YouTube under threshold — regression puts $15K/week as the floor for organic brand-search lift | YT spend raised and tested against an audience stack, Shorts, etc. |
Also in flight on media: Relaunched AppLovin (Axon) 6/11 · Launching Attentional (Meta test) · Re-launching TikTok (TT Shop offline for months) · Launching other product funnels: nourish, coffee, matcha.
URL tests went dark Jun 11 evening. If routing is the driver, these move within days — in this order. If they don’t move in ~10 days, the routing hypothesis takes real damage and media-side is carrying more of the miss than modeled.
The first domino. Should start recovering within 1–2 days of the pause.
Should fall back to the organic-click baseline if the surge was test redirects.
Direct check that the pause actually took — if this isn’t ~0, something is still redirecting.
Follows routing with a short lag. April’s peak was 2.32%.
The counterfactual’s +66/day. Above that needs the media side too.
Routing fix gets to ~$101; the last $11 is CPMs, creative, and Google Shopping.