NCA Campaign Analysis — AmazingCo AU

Tab 1 · Paid Campaign Performance

BAU Mystery Picnic vs NCA New Customer — PoP & YoY

Comparing the 47-day performance window before NCA launch (Mar 13–Apr 29) against the 47-day post-launch period (Apr 30–Jun 16), plus YoY context against the same 2025 window.

🚨

Critical Finding: NCA has cost AU market 1,121 purchases at a +38% CPA premium

Switching budget from the BAU mystery-picnic URL to NCA drove purchases down −58.3% PoP (1,924 → 803) while CPA surged from $78.86 → $109.05. The NCA page is optimised for sign-ups, not conversions — the algorithm is paying cold-audience CPMs to capture emails, not transactions. Every dollar over-invested here widens the YoY gap.

Period-on-Period · Same 47-Day Window

Total Cost

$87.6K

−42.3% vs BAU period

BAU: $151.7K

Impressions

6.46M

−57.9% vs BAU period

BAU: 15.3M

Clicks (All)

150,790

+12.3% vs BAU period

BAU: 134,262

Purchases

803

−58.3% vs BAU period

BAU: 1,924 purchases

CPA

$109.05

+38.3% vs BAU period

BAU: $78.86 CPA

PoP Comparison — BAU URL vs NCA URL (47 days each) NCA Under-performing

Period / URL	Cost	Impressions	Clicks	CPM	Purchases	CPA	Conv. Rate
BAU — mystery-picnic Mar 13–Apr 29, 2026	$151,731.88	15,316,579	134,262	$9.91	1,924	$78.86	1.43%
NCA — new-customer Apr 30–Jun 16, 2026	$87,569.23	6,456,135	150,790	$13.56	803	$109.05	0.53%
% Change (NCA vs BAU)	−42.3%	−57.9%	+12.3%	+36.8%	−58.3%	+38.3%	−62.8%

Year-on-Year · Apr 30–Jun 16 Window

⚠️

YoY: We're getting 50.9% fewer purchases from the same budget window — with a worse CPA

In the equivalent period last year (Apr 30–Jun 16, 2025), the mystery-picnic URL delivered 1,636 purchases at $84.91 CPA on $138.9K spend. This year the same calendar window, running NCA, produced only 803 purchases at $109.05 CPA on $87.6K spend. The YoY shortfall is structural, not seasonal.

Purchases YoY

803

−50.9% vs 2025

2025 same window: 1,636

CPA YoY

$109.05

+28.4% vs 2025

2025 same window: $84.91

Impressions YoY

6.46M

−45.3% vs 2025

2025: 11.8M impressions

Conv. Rate YoY

0.53%

−58.2% vs 2025

2025: 1.27% conv. rate

YoY Comparison — Apr 30 to Jun 16 (Same 47-Day Window)

Year / URL	Cost	Impressions	Clicks	CPM	Purchases	CPA	Conv. Rate
2025 — mystery-picnic Apr 30–Jun 16, 2025	$138,911.28	11,803,879	128,336	$11.77	1,636	$84.91	1.27%
2026 — new-customer (NCA) Apr 30–Jun 16, 2026	$87,569.23	6,456,135	150,790	$13.56	803	$109.05	0.53%
% Change YoY	−37.0%	−45.3%	+17.5%	+15.2%	−50.9%	+28.4%	−58.2%

💸

CPM inflation: paying more for less reach

CPM rose from $9.91 → $13.56 (+36.8%) PoP, and from $11.77 → $13.56 (+15.2%) YoY. The NCA page targets a broader cold audience pool with a sign-up CTA, which the Meta algorithm perceives as lower-intent — requiring higher auction bids to compete.

📉

Conversion rate collapsed by over 62%

BAU delivered 1.43% CVR to purchases; NCA delivers only 0.53% — a drop of −62.8%. Even accounting for the different page objective (lead gen vs purchase), the downstream conversion from sign-up to purchase is not visible in these numbers and is not justifying the spend level.

📊

More clicks, far fewer buyers

Clicks actually increased by 12.3% with NCA, but purchases dropped 58.3%. This confirms the NCA landing page is driving browsing behaviour, not purchase intent. Users are arriving, reading, and leaving without transacting — consistent with a lead-gen page design.

Tab 2 · CPA Performance

Cumulative & Singularity CPA — Apr 1 to Jun 15, 2026 vs 2025

Tracking how NCA launch (Apr 30) and the Meta A/B test (May 19) have structurally shifted AU's blended CPA trajectory against the prior year baseline.

📌

Key Callout: NCA created a permanent CPA wedge that hasn't closed since Apr 30

Before NCA launched, CPA tracked broadly in line with 2025 (within ~17%). The moment NCA went live on Apr 30, the gap widened — and the May 19 Meta A/B test further entrenched it. By Jun 15, Blended CPA sat at $67.88 vs $57.65 (2025) — a −15.1% gap. Singularity CPA (our own channels only) tells an even starker story: $72.81 vs $61.99 — a $10.82 premium.

2026 Blended CPA 2025 Blended CPA 2026 Singularity CPA Key Events NCA Launch Apr 30 · A/B Test May 19 · Winter Campaign Jun 11

Blended CPA — 2026 vs 2025 (Includes Marketplace Performance)

Cumulative CPA includes all channels + marketplace bookings. Lower is better.

Singularity CPA — 2026 vs 2025 (Paid Channels Only)

Singularity measures AmazingCo's own paid channel CPA without marketplace contribution. Lower is better.

CPA by Key Period — Phase Analysis

Phase	Date Range	Blended CPA 2026	Blended CPA 2025	YoY Gap	Singularity CPA 2026	Singularity CPA 2025	YoY Gap	Key Driver
Pre-NCA Baseline	Apr 1–Apr 29	$61–$69	$57–$58	−12% to −17%	$66–$76	$58–$61	−10% to −22%	Algorithm in learning; no major event
Flash Sale + Mother's Day Winds Down	May 1–May 18	$45–$60	$42–$56	−7% to −15%	$47–$63	$44–$57	−7% to −15%	Flash sale boosted conversions, tightened gap
NCA A/B Test Launches	May 19–May 31	$49–$54	$44–$49	−9% to −11%	$51–$57	$46–$51	−8% to −10%	Meta A/B split further dilutes purchase volume
Winter Campaign Live + NCA	Jun 1–Jun 15	$63–$68	$56–$58	−9% to −15%	$68–$73	$61–$63	−9% to −15%	Winter campaign + NCA audience overlap inflating CPA

📍

The NCA wedge is now structural, not temporary

Looking at the Singularity CPA trend, the 2026 line has maintained a consistent $10–$15 premium over 2025 from Apr 30 onwards. Prior to NCA launch, the gap was widening but from a cyclical starting point — post-launch it entrenched. The A/B test on May 19 applied more budget to non-converting audience buckets, sustaining the premium through June.

🔑

Singularity CPA peaked at $87.31 vs $69.41 in 2025

June 1 marked the worst single-day gap: Singularity $87.31 vs $69.41 (2025), a $17.90 premium. This coincides with winter campaign ramp-up layered on top of ongoing NCA spend — Meta's algorithm is competing against its own campaigns for overlapping audiences, driving up auction costs. The Blended metric also hit $85.07 vs $66.26 on the same day.

Tab 3 · Meta A/B Test Analysis

New Customer (Modified Audiences) vs BAU Advantage+ (Broad)

Active May 19–Jun 16, 2026. Reviewing campaign-level, ad-set-level, and audience segment data to assess whether we are overpaying for new-audience sign-ups.

🔬

Verdict: BAU Advantage+ outperforms NCA on every purchase metric — despite nearly identical spend

With nearly equal budget ($27.5K NCA vs $27.7K BAU), BAU delivered 196 purchases at $141.22 CPA while NCA delivered 174 purchases at $158.24 CPA — a +12.1% CPA premium for NCA. The A/B test was designed to grow the audience pool for NCA sign-ups, but the data shows we are paying more per purchase with audiences that convert at worse rates. The only bright spot within NCA: the mystery-picnic retargeting segment ($60.30 CPA, 3.22% CVR) — which ironically is the audience best suited to BAU.

NCA Total Spend

$27,533

May 19–Jun 16

Modified audience targeting

BAU Total Spend

$27,680

Nearly equal budget

Broad Advantage+ targeting

NCA CPA

$158.24

+12.1% vs BAU

174 purchases

BAU CPA

$141.22

Better efficiency

196 purchases

Campaign-Level A/B Summary

Campaign	Spend	Purchases	CPA	vs BAU
BAU Advantage+ (Broad audiences)	$27,680.00	196	$141.22	Baseline
New Customer Advantage+ (Modified audiences)	$27,533.24	174	$158.24	+12.1% CPA
% Difference	−0.5%	−11.2%	+12.1%	—

Ad Set Deep-Dive — NCA (Modified Audiences)

New Customer Advantage+ — Ad Set Breakdown Sorted by CPA

Ad Set	Spend	Impressions	Clicks	CPM	LPVs	Purchases	ROAS	CPA	CVR	Verdict
MOFU \| Web V - mystery-picnics 30d	$3,557.93	197,720	1,830	$17.99	809	59	5.15x	$60.30	3.22%	⭐ Best Performer
TOFU \| Lookalikes purchasers	$3,308.11	372,652	2,496	$8.88	858	19	0.92x	$174.11	0.76%	Marginal
MOFU \| ad engagers + video viewers 30d	$3,776.24	309,341	2,404	$12.21	1,151	19	1.73x	$198.75	0.79%	Marginal
MOFU \| Web V - new-customer LP 30d	$6,041.30	267,874	28,344	$22.55	21,936	30	1.47x	$201.38	0.11%	❌ Inefficient
TOFU \| Broad \| 25–65 \| F	$9,946.76	996,903	8,924	$9.98	3,906	45	1.32x	$221.04	0.50%	❌ Over-invested
BOFU \| LP web visits by time spent 30d	$902.90	47,365	3,751	$19.06	2,025	2	0.18x	$451.45	0.05%	❌ Pause

BAU Advantage+ — Ad Set Breakdown Original Broad Setup

Ad Set	Spend	Impressions	CPM	LPVs	Purchases	CPA	Verdict
MOFU \| Web V - mystery-picnics 30d	$3,556.39	197,670	$17.99	809	59	$60.28	⭐ Best Performer
Broad \| 25–65 \| F	$29,118.92	1,960,602	$14.85	32,538	207	$140.67	Marginal
TOFU \| Broad \| 25–65 \| F	$8,491.52	830,562	$10.22	3,203	34	$249.75	❌ Over-invested
MOFU \| Web V - new-customer LP 30d	$6,039.83	267,821	$22.55	21,934	30	$201.33	❌ Inefficient
MOFU \| ad engagers + video viewers 30d	$3,775.76	309,291	$12.21	1,149	19	$198.72	Marginal
TOFU \| Lookalikes purchasers	$3,307.64	372,593	$8.88	858	19	$174.09	Marginal
BOFU \| LP web visits by time spent 30d	$902.88	47,363	$19.06	2,025	2	$451.44	❌ Pause

Click and CVR data not available at this breakdown level — figures reflect spend, impressions, CPM, landing page views, purchases, and CPA only.

Audience Segment Analysis — Prospecting vs Engaged

NCA — Audience Segment CPA Performance

Cost per purchase by audience type within NCA campaign

MOFU mystery-picnics · Prospecting$42.59 CPA · 34 purch.

MOFU mystery-picnics · Engaged$84.33 CPA · 25 purch.

MOFU new-customer LP · Prospecting$112.43 CPA · 8 purch.

TOFU Lookalikes · Prospecting$150.04 CPA · 19 purch.

TOFU Broad · Engaged$131.74 CPA · 11 purch.

MOFU ad engagers · Prospecting$197.01 CPA · 16 purch.

MOFU new-customer LP · Engaged$233.65 CPA · 22 purch.

TOFU Broad · Prospecting$249.75 CPA · 34 purch.

BOFU LP web visits · Engaged$647.76 CPA · 1 purch.

BAU — Audience Segment CPA Performance

Cost per purchase by audience type within BAU campaign

MOFU mystery-picnics · Prospecting$42.59 CPA · 34 purch.

MOFU mystery-picnics · Engaged$84.33 CPA · 25 purch.

MOFU new-customer LP · Prospecting$112.43 CPA · 8 purch.

TOFU Broad · Engaged$131.74 CPA · 11 purch.

Broad 25–65 F · Prospecting$136.83 CPA · 136 purch.

TOFU Lookalikes · Prospecting$150.04 CPA · 19 purch.

Broad 25–65 F · Engaged$151.02 CPA · 60 purch.

MOFU ad engagers · Prospecting$197.01 CPA · 16 purch.

MOFU ad engagers · Engaged$207.86 CPA · 3 purch.

MOFU new-customer LP · Engaged$233.65 CPA · 22 purch.

TOFU Broad · Prospecting$249.75 CPA · 34 purch.

BOFU LP web visits · Prospecting$255.12 CPA · 1 purch.

BOFU LP web visits · Engaged$647.76 CPA · 1 purch.

⭐

Mystery-picnic audience is the gold standard

Across both campaigns, the MOFU | Web V — mystery-picnics 30d audience dramatically outperforms all others: prospecting at $42.59 CPA, engaged at $84.33 CPA. These users have already demonstrated intent for the exact product — routing them through NCA's sign-up flow wastes that intent. This audience belongs on the BAU purchase URL.

❌

Are we overpaying for sign-ups? Yes.

The NCA new-customer LP retargeting audience (people who visited the NCA page) converted at $201–$234 CPA. We are paying to retarget people who already visited a sign-up page, with more ads to sign up. Within BAU, TOFU Broad prospecting at $249.75 CPA is similarly inefficient — cold audiences on a broad targeting setup simply don't convert at the rate the mystery-picnic retargeting segment does.

🔁

BOFU LP time-spent audiences should be paused immediately

BOFU LP web visits by time spent delivered only 2–3 total purchases across both campaigns at a CPA of $451–$648. This audience type, intended to recapture high-intent visitors, is fundamentally incompatible with a sign-up CTA page — the intent signal is "browsed a purchase page" but the ad sends them to a lead-gen form.

Tab 4 · Website / Landing Page Performance

BAU vs NCA Landing Page — All Traffic & Meta Traffic

GA4 session and transaction data comparing the NCA new-customer page (Apr 30–Jun 16, 2026) against the BAU mystery-picnic page (same period, 2025). Includes both all-traffic and Meta-only views.

🖥️

Key Finding: NCA drives 70.8% more sessions but converts 87.3% fewer of them to purchases

The NCA page attracts significantly more traffic (100,635 vs 58,919 sessions YoY) but produces only 74 purchases vs 340 — a −78.2% decline in transactions. The session-to-purchase rate collapsed from 0.58% to 0.07%. This isn't a traffic quality problem alone — it's a fundamental page objective mismatch. NCA is optimised to capture emails, not drive purchases.

All Traffic — Landing Page Comparison (Apr 30–Jun 16)

Metric	NCA 2026 /promo/new-customer	BAU 2025 /promo/aus/mystery-picnic	% Change	Direction
Ad Spend	$104,831.18	$191,555.83	−45.3%	Lower spend (NCA)
Sessions	100,635	58,919	+70.8%	More traffic, less quality
Engaged Sessions	32,233	27,913	+15.5%	Marginally more engagement
Engagement Rate	32.03%	47.38%	−32.4%	❌ Much lower engagement
Bounce Rate	67.97%	52.62%	+29.2%	❌ Far higher bounce
Avg. Session Duration	4m 04s	1m 18s	+213%	⚠️ Browsing, not buying
Transactions	74	340	−78.2%	❌ Critical shortfall
Total Revenue	$16,014.98	$65,860.82	−75.7%	❌ Revenue collapse
Session → Purchase Rate	0.07%	0.58%	−87.3%	❌ Conversion funnel broken
Engaged → Purchase Rate	0.23%	1.22%	−81.2%	❌ Even engaged users don't buy
AOV	$216.42	$193.71	+11.7%	✅ Better basket size
Cost per Session (CPS)	$1.04	$3.25	−68%	✅ Cheaper traffic, less valuable

Meta Traffic Only — Landing Page Comparison (Apr 30–Jun 16)

Metric	NCA 2026 Meta → /promo/new-customer	BAU 2025 Meta → /promo/aus/mystery-picnic	% Change	Direction
Meta Spend	$87,569.23	$138,911.28	−37.0%	Lower Meta spend
Sessions (Meta)	93,921	38,221	+145.7%	Far more sessions from Meta
Engaged Sessions (Meta)	31,080	13,803	+125.2%	More engaged sessions, less converting
Engagement Rate	33.09%	36.11%	−8.4%	Slightly lower
Bounce Rate	66.91%	63.89%	+4.7%	Marginally higher bounce
Avg. Session Duration	4m 19s	57s	+356%	⚠️ Probing, not purchasing
Transactions (Meta)	70	115	−39.1%	❌ Fewer Meta-driven purchases
Revenue (Meta)	$15,265.00	$23,518.52	−35.1%	❌ Revenue decline
Session → Purchase Rate	0.07%	0.30%	−75.2%	❌ Meta traffic converting at quarter rate
Engaged → Purchase Rate	0.23%	0.83%	−73.0%	❌ Engagement not leading to purchase
AOV	$218.07	$204.51	+6.6%	✅ Higher basket when they do buy
Cost per Session (Meta)	$0.93	$3.63	−74.4%	✅ Very cheap traffic, very low quality

Combined Insights — Vivian's Findings + Analytical Context

💡

Insight 1: Revenue attribution underplays BAU's real impact

As Vivian notes, the BAU mystery-picnic page funnels users directly to product booking pages, where the actual revenue is attributed. GA4 records the session on mystery-picnic but the conversion on the product page — meaning the $65.8K revenue figure is an undercount of BAU's true contribution. NCA's $16K revenue is also potentially underattributed (email sign-ups may later convert via email), but given only 74 recorded transactions across 100K sessions, the downstream conversion from email is not visible at scale in this window.

💡

Insight 2: Long session time signals scepticism, not engagement

The NCA page averaged 4m 04s session duration vs BAU's 1m 18s. Vivian's read is correct: users are probing, not buying. A purchase-focused page resolves quickly — user sees product, clicks, buys. A sign-up page creates friction (form fill, email verification, promo code delivery) that extends session time without completing a transaction. The high bounce rate (68%) confirms most users still leave without converting, even after spending time on page.

💡

Insight 3: Cheap CPS is a vanity metric here

NCA's $0.93–$1.04 cost per session looks attractive vs BAU's $3.25–$3.63, but the metric is inverted from what matters. BAU converted sessions at $3.25/session × 0.58% CVR, meaning each purchase cost ~$560 in session traffic. NCA converts at $1.04/session × 0.07% CVR, meaning each purchase costs ~$1,486 in session traffic. Cheap sessions from an incompatible page design are not an efficiency win.

✅

Insight 4: AOV shows buyers have stronger intent when they do convert

Despite the conversion collapse, when NCA does drive a purchase, the AOV is $216–$218 vs BAU's $193–$204 — a 7–12% uplift. This suggests the users who persist through the NCA funnel (sign up, receive code, then purchase) may represent a higher-value customer segment. The hypothesis is worth testing at a smaller budget scale, but does not justify the current 80%+ budget allocation at these purchase volumes.

Tab 5 · Recovery Proposal

Budget Restructure & Campaign Recovery Plan

A phased proposal to restore AU's CPA performance, backed by data across 5 analysis sections. Recommendation: realign spend toward purchase-intent audiences and reactivate the proven BAU structure.

🎯

The core problem: 85% of budget is being directed at a sign-up page, using a retention/conversion budget

NCA was designed as a lead-gen experiment, but it's absorbing the lion's share of AU's acquisition budget ($57.4K / 85% of Meta spend over 30 days). The data is unambiguous: the mystery-picnic BAU URL delivers purchases at $78.86 CPA vs NCA's $109.05 CPA, and the YoY CPA wedge has persisted for 47 days. The recovery path is clear — rebalance budget in favour of the URL that converts.

Current Budget Setup (Daily)

Campaign	Daily Budget	% Weighting	Verdict
Melbourne - Advantage+ Winter	$678.18	—	Learning
Brisbane - Advantage+ Winter	$251.11	—	Learning
Sydney - Advantage+ Winter	$236.23	—	Learning
City Winter Campaigns Subtotal: $1,165.52 / 36.8%
New Customer Campaigns (both)	$1,900.26	60.1%	❌ Over-weighted
BAU Dynamic Remarketing	$98.91	3.1%	Hold
Total	$3,164.69	100%	—

Proposed Budget Setup (Daily)

Campaign	Daily Budget	% Weighting	Action
Melbourne - Advantage+ Winter	$678.18	—	Hold
Brisbane - Advantage+ Winter	$251.11	—	Hold
Sydney - Advantage+ Winter	$236.23	—	Hold
City Winter Campaigns Subtotal: $1,165.52 / 36.8%
New Customer (mystery-picnic audience only)	$285.04	9.0%	−85% Reduction
New BAU Advantage+ (mystery-picnic URL)	$1,615.22	51.0%	✅ New Campaign
BAU Dynamic Remarketing	$98.91	3.1%	Hold
Total	$3,164.69	100%	Same total spend

🔧

Recovery Action Plan — 3 Moves to Restore CPA Performance

Recommended immediate actions backed by the data in this analysis

Move 1: Relaunch BAU Advantage+ campaign pointing to the mystery-picnic URL (~51% of budget)
The mystery-picnic URL delivered $78.86 CPA PoP vs $109.05 for NCA — a 27.6% improvement. YoY it delivered 1,636 purchases in the equivalent window vs NCA's 803. Create a new BAU Advantage+ campaign with the original URL https://www.amazingco.me/promo/aus/mystery-picnic, allocating $1,615/day (~51% of total budget). Use Advantage+ Shopping targeting with a purchase objective. This directly addresses the structural CPA wedge that has persisted since Apr 30.
Move 2: Reduce NCA to <10% budget, target MOFU mystery-picnic audience only with tCPA bidding
Vivian's recommendation is data-validated: the only NCA ad set delivering acceptable CPA is MOFU | Web V - URL contains mystery-picnics - 30 days at $60.30 CPA prospecting / $84.33 CPA engaged. Retain this audience exclusively at $285/day (~9%), exclude past converters from the last 30 days, and switch to a tCPA bid cap to prevent the algorithm paying a premium for sign-up micro-conversions. Pause TOFU Broad, BOFU LP time-spent, and MOFU new-customer LP retargeting immediately — these three ad sets account for $16.8K of NCA spend and deliver $201–$647 CPA.
Move 3: Maintain winter city campaigns in learning phase — do not disrupt
Melbourne ($678/day), Brisbane ($251/day) and Sydney ($236/day) Advantage+ winter campaigns are in the algorithm learning period. Any budget changes here will reset the learning window and reduce efficiency. Hold spend constant. The data shows winter campaigns launched June 11 — at minimum 14 days of learning data (by June 25) should be accumulated before any optimisation decisions are made on these campaigns.

💬

Feedback on Vivian's Proposal — What the Data Supports

Analytical validation and additions to the proposed budget restructure

✅ Strongly Agreed: Mystery-picnic retargeting for NCA is the right call

The data is unambiguous — MOFU mystery-picnics is the only audience delivering sub-$100 CPA within NCA. Running NCA at <10% with only this audience is the right scope. These users have already shown commercial intent for the product, making them the most likely to respond to a sign-up offer with the expectation of receiving a discount and then converting.

✅ Strongly Agreed: Revert BAU to mystery-picnic URL

The mystery-picnic URL delivered materially better economics in every comparable period: PoP (+38% CPA improvement), YoY ($84.91 vs $109.05), and Meta-specific (115 vs 70 transactions in the same 47-day window at lower spend). The recovery plan's core thesis — that we over-pivoted away from a working URL — is fully supported by the data.

➕ Addition: Exclude past converters across BAU too, not just NCA

Vivian proposes excluding past converters (30 days) from NCA — this should also be applied to the new BAU Advantage+ campaign. Excluding customers who have already purchased prevents wasted spend on retention objectives through acquisition campaigns, and ensures budget is directed at genuinely new audiences for the mystery-picnic product.

➕ Addition: Set a 14-day review gate before further NCA scaling

With NCA reduced to <10% budget, the algorithm will need a 7–14 day re-learning period on the new, narrower audience. Recommend a formal review at the 2-week mark (around July 1) to assess: (a) whether the mystery-picnic NCA audience's CPA holds below $80; (b) what proportion of sign-ups from this period convert to purchases; (c) email attribution data from the NCA funnel to build a true LTV case for the NCA strategy.

The Core Insight From This Analysis: 85% of our acquisition budget has been quietly redirected to a page designed not to sell — reverting it is a one-line fix, not a strategy overhaul, and it alone accounts for the gap between 1,924 purchases at $78.86 and where we are today.

📈

Expected Recovery Outcome (Next 47-Day Window)

Based on the BAU mystery-picnic PoP performance benchmark: reverting ~90% of budget to a purchase-objective URL should restore purchases toward 1,500–1,900 range and bring CPA back toward $79–$90 — narrowing the YoY Singularity CPA gap from −15% to within −5%. The NCA test continuing at <10% budget provides ongoing learnings without structural CPA damage.