Growth Diagnostic · Trailing 90 Days

Why are signups down and CAC up?

Mar 13 – Jun 11, 2026 · Northbeam · Intelligems · Meta · Microsoft Clarity · Shopwatch. Snapshot as of Jun 11, 2026.

Headline: No single smoking gun — three things degraded at once and compounded, so CAC crept rather than spiked. (1) Media auction got pricier (eCPC: FB +10%, Google +44%) with CTR roughly flat — a CPM story, not creative. (2) The blended Rise 2 CVR fell in May→June — but a clean cut (§2B) shows the canonical /pages/rise-2 page is stable (CVR 1.84%→1.87%, bounce improved); the drag is worse-converting redirect variants whose share grew 7%→11%. The real page isn't broken. (3) Site speed/flicker is chronically poor (CLS 0.72, LCP 3.4s, ~45% in-app webviews) — a real baseline tax, though it did not worsen month-over-month, so it's a fix-the-floor issue, not the recent trigger.

01Signups, CAC & media efficiency

Prior 30d (Mar 13–Apr 11) vs recent 30d (May 13–Jun 11), blended from Northbeam. attribution: NB default / lifetime — see note

New customers / day

446

was 527 · target 600

−15% · 26% under target

Blended CAC

$95.17

was $91.76 · target $90

+3.7% (recent edge worse)

Facebook CAC

$181

was $149 · 5,067 new cust

+21%

Facebook eCPC

$4.02

was $3.67 · clicks −15%

+10%

CAC by channel, prior vs recent 30d. Google CAC +32%, Facebook +21%.

Is it a CTR problem? (you asked)

Meta media metric	Prior	Recent
CTR (avg daily)	1.56%	1.67%
CTR (last 2 wks trend)	—	1.9% → 1.5%
FB effective CPC	$3.67	$4.02
Google effective CPC	$1.57	$2.26
FB clicks	268k	229k

CTR did not structurally drop (30-day avg up). But CPC rose anyway → cost driver is CPM / auction inflation, not click-through.

Attribution note

CAC/new-customer figures use Northbeam's default (lifetime) window, not your 7-day-click standard, so absolute CAC may shift modestly under 7-day-click — the prior-vs-recent direction holds. CTR/CPM/CPC are unaffected by attribution.

02The Rise 2 funnel — where it actually leaks

Your funnel: LP → PDP → ATC → Checkout → Buy. The May→June CVR decline is concentrated at one step.

2-week window	LP→PDP	PDP→ATC	ATC→Checkout	Checkout→Buy	CVR
Apr 24 – May 7	25.3%	13.7%	93.8%	59.3%	2.14%
May 8 – 21	21.9%	13.6%	93.3%	59.2%	2.00%
May 22 – Jun 4	16.5%	14.1%	94.3%	59.6%	1.78%
Jun 5 – 11	15.7%	15.2%	92.0%	51.6%	1.54%

LP→PDP collapsed 25%→16% (−38%) across May–June. PDP→ATC held steady (~14–15%). The damage is at the top of the funnel — Rise 2 is sending a shrinking share of visitors to a product page.

The diverging step: LP→PDP vs PDP→ATC

PDP→ATC is flat. LP→PDP is the line that broke — right as test load peaked.

Where Rise 2 sessions leak (recent 30d)

Of 220.6k landings: 57% bounce on the LP; ~90% of 2nd-pagers exit before ATC. 3.4% add to cart, 1.82% convert.

The "LP→ATC dip" is really an LP→PDP dip

The broken link is getting people off the Rise 2 landing page onto a product page. Once on the PDP, add-to-cart behaves normally. Fix points at the Rise 2 LP — its routing/CTAs (the redirect tests) and render — not the PDP.

2BClean cut: canonical Rise 2 vs redirect variants

Is canonical /pages/rise-2 actually degrading, or is worse redirect-variant traffic dragging the blended number down? Decomposed by exact landing path (canonical = blended − rise-2-* variants).

Segment	Window	Landings	LP→PDP	ATC%	CVR	Bounce
Canonical /pages/rise-2	Prior	230,008	20.1%	3.57%	1.84%	65.2%
Canonical /pages/rise-2	Recent	195,852	18.0%	3.44%	1.87%	56.3%
Redirect variants rise-2-*	Prior	18,349	13.8%	2.45%	1.24%	70.3%
Redirect variants rise-2-*	Recent	24,800	17.5%	2.86%	1.46%	62.4%

Canonical = "landing path contains rise-2" minus "contains rise-2-". Exact-match on /pages/rise-2 returns 0 in Intelligems (it isn't keyed to that literal path), so subtraction is the reliable read. Redirects sent to the homepage are logged under home and never enter this bucket at all.

Canonical vs variant, recent 30d

Variants convert ~22% worse (CVR 1.46% vs 1.87%) and bounce more — but they're only 11% of traffic.

CVR over time + variant traffic share

Canonical CVR is flat (1.84%→1.87%); variant share grew 7.4%→11.2%, adding a small blend drag.

Verdict: the real Rise 2 page is healthy — a traffic-mix story, not a broken page

Cleanly isolated, canonical /pages/rise-2 CVR held flat (1.84%→1.87%) and bounce improved (65%→56%). The blended decline came from (a) worse-converting redirect variants whose share grew ~50%, and (b) volatility in the partial last week. Don't rebuild Rise 2. The signup/CAC miss lives upstream — media auction inflation + spend↓→traffic↓ — and in test hygiene, not the landing page. Chronic CLS/flicker is still worth fixing as a floor.

03Intelligems test load — your hypothesis

A/B test launches per 2-week window. Velocity ran ~0–2 in Mar–Apr, then 7 and 10 in late May / early June — a ~4–5× ramp, concentrated on Rise 2 redirects + PDP edits.

~8–12 tests ran concurrently in late-May→June, skewed to content/url redirects and content/onsiteEdits DOM edits. The LP→PDP decline coincides with this ramp.

The redirect tests are the most likely LP→PDP culprit

The break lines up in time with a burst of Rise 2 redirect tests ("Rise 2 vs Rise 2 X", "Coffee Rise 2 vs Home/Super LP"). Splitting paid traffic across variant landing pages — some of which route to home/listicle instead of a PDP — would mechanically lower LP→PDP. Worth auditing where each live variant sends users.

04Clarity — friction & speed, prior vs recent 30d

You asked for the before/after. Here it is — and the story is "chronically bad, not newly worse."

Friction & engagement (sitewide)

Quick-backs nudged up (8.35%→8.82%); dead-clicks flat-to-better; pages/session improved. No acute friction spike.

Core Web Vitals & environment

Metric	Prior	Recent	Verdict
Perf score	67	66	—
LCP (load)	3.5s	3.4s	needs work
INP	300ms	310ms	needs work
CLS (flicker)	0.72	0.71	POOR
FB+IG in-app	45.3%	44.6%	slow webviews
Quick backs	8.35%	8.82%	—

CLS 0.71–0.72 is ~7× the "good" bar (<0.1) — serious, persistent layout shift/flicker. 459k JS errors, mostly "myelement is not defined" (an injected snippet erroring).

Flicker is chronic, not the recent trigger

Your hypothesis is half-right: the site does have a real flicker/speed problem (CLS 0.72 "poor", LCP 3.4s, ~45% in slow FB/IG in-app browsers, 8.8% quick-backs) that taxes conversion every day. But it did not deteriorate month-over-month — so it's not what newly broke in May–June. Fix it to raise the floor; look to the redirect tests (Section 3) for the recent LP→PDP drop.

05Diagnosis & what I'd do next

Ranked by contribution to the current signup/CAC miss:

Media auction inflation + spend cut (biggest $ driver). CPM/CPC up with CTR flat — paying more per click for the same pull — and spend↓ pulled traffic↓. Refresh creative & audiences. Per §2B, the Rise 2 page itself is healthy, so this (not the LP) is the main miss.
Tighten redirect-test hygiene (don't rebuild the page). Canonical /pages/rise-2 CVR is flat (1.84%→1.87%); redirect variants convert ~22% worse and their share grew to 11%. Stagger tests, protect a clean control, and route every variant to a real PDP — never send paid traffic to home/listicle.
Fix the chronic CLS/flicker to raise the floor. 0.72 is poor and ~45% of traffic is in slow in-app browsers. Won't explain the May–June dip, but it's a standing conversion tax — and clear the "myelement is not defined" JS error.
Re-pull CAC at 7-day-click to match your standard, and pull the Meta creative-fatigue view to see what's behind the last-2-week CTR slide.

Sources: Northbeam (new customers, CAC, spend, visits, CTR/CPM/CPC — 30d windows, lifetime attribution), Intelligems (Rise 2 conversion funnel ×7 windows + experiment history), Meta Ads (daily CTR/CPM), Microsoft Clarity (friction, Core Web Vitals, browser mix — two 30d windows), Shopwatch (theme change log). Snapshot: Jun 11, 2026 · internal — password protected.