US Performance Diagnosis Dashboard

Executive summary

When it started: CPA began drifting up from a healthy ~$55–$60 baseline in mid-May, then broke sharply on June 1 (~$68 → $79.87) — the same week an Advantage+ targeting setting issue across US/CA/UK (identified May 28, fixed June 1) flooded reporting with junk sessions, and a $100 tCPA cap was deliberately added to Southern Occasions (June 9). CPA continued climbing through June, peaking in the low-$80s.

How it's going: the June 30 sale end was the second shock — a mass automated pause/re-activation, new creative assets still in early learning (Summer NH refresh, built Jun 29–30), and a genuine conversion-rate decline in Southern/Northern search campaigns pushed CPA into the $80–$90 range through early-to-mid July. Manual responses (tCPA raised to $104.5 on Jul 6, ongoing search-term and audience clean-up) are lagging indicators reacting to the spike, not yet a resolved fix. Meta shows the same pattern one step behind, with CPA roughly doubling from baseline into July.

Pathway to recovery: three things need to land before CPA can realistically return toward $58: (1) the new creative set needs to clear its learning phase (typically 1–2 weeks post-launch); (2) the tCPA increases need time to normalise bidding rather than being adjusted again mid-cycle; and (3) the Meta UTM/tracking gap (see Channel performance) needs fixing so channel-level decisions are made on real data rather than partially-blind reporting. Absent a new shock, a gradual pull-back toward the $65–$70 range looks achievable within 3–4 weeks; a full return to $58 likely needs a further round of Southern/Northern campaign restructuring given how much manual pausing has already been done at the keyword level.

Blended US CPA vs target

Daily CPA against the $58 target, with major account changes and seasonal activity overlaid. Hover for exact values.

Actual CPA Target ($58) Change / seasonal event

Key events

May 28 – Jun 1: Advantage+ targeting issue across US/CA/UK caused the GA4 session spike (7–10k sessions/day, 80%+ bounce) — identified May 28, fixed June 1.
Jun 1 CPA inflection point, same week as the targeting issue and a $100 tCPA cap added to Southern Occasions (Jun 9).
Jun 16: tCPA reduced from $100 to $95 on Southern Occasions in response.
Jun 23–26: US/CA/UK prime sale live — CPA holds in the high $70s.
Jun 30 – Jul 1: sale ends; mass automated pause, manual re-activation, NH summer campaigns live Jul 1.
Jul 6: tCPA increased +10% ($95 → $104.5) on both Northern and Southern Occasions.
Jul 9–12: iPhone Chrome checkout anomaly (dev brief sent to Scott's team) — site-side, not visible in Google Ads change history.

Seasonal & NH/US timeline

Filtered to activity that affects the US or NH markets only (AU/NZ-only seasonal moments excluded)

Date	Activity

Full Google Ads change log

Every logged US account change, April–July. Search by campaign or keyword.

Date	Campaign	Activity

CPA by channel: Google Ads vs Meta

Hover to see exact CPA for either channel on any day

Google Ads CPA Meta Ads CPA

Why CPA spiked — by channel and period

Period	Google CPA	Meta CPA	Likely driver

CPS by channel: Google Ads vs Meta

Cost per session, using real session counts filtered to amazingco.me/us pages by source/medium (google/cpc vs meta/facebook-pd) — corrected from the previous GA4 channel-grouping approach, which was undercounting Meta sessions by roughly 700×.

Google Ads CPS Meta Ads CPS

Core insight

With real session data, Meta is actually cheaper per session than Google — average CPS across the period is roughly $0.61 for Meta vs $0.96 for Google. The previous CPS read (using GA4's default channel grouping) was badly wrong: it classified almost none of Meta's real traffic as "Paid Social" because Meta's UTM tagging (source=meta, medium=facebook-pd) isn't recognised by GA4's default channel-grouping rules, even though the sessions are very much there (164k Meta sessions vs 146k Google sessions over the period, once matched on source/medium directly). This changes the read on Meta: its clicks genuinely are cheaper than Google's, and it's converting a comparable volume of sessions — so a higher Meta CPA than Google's isn't explained by expensive traffic, it's explained by lower on-site conversion quality per session, which does point back to creative/audience quality rather than a tracking artefact. The tracking issue is real and worth fixing (it was masking true channel volume), but it's separate from the creative-quality issue — both are true at once.

Recommendations

Google Ads: hold the current tCPA discipline (Occasions at $104.5) and keep the search-term exclusion work running — it's the main lever moving CPA back toward target. Prioritise Empire State Building and Shopping generics for tROAS re-tuning given the multiple reductions already made.
Meta: since cheap sessions are confirmed real (not a tracking gap), the creative/audience mix is the actual lever — review the current creative set for what's attracting volume but not converting, and consider tightening audience targeting rather than broadening further.
Still worth asking Scott to add Meta's UTM tagging convention to GA4's channel-grouping rules — not because the sessions are missing, but so future reporting (and anyone using the default channel view) doesn't repeat this misread.

GA4 sessions (US)

Bounce rate & engagement rate

Insights & recommendations

The May 28–30 session spike is fully explained: it's the Advantage+ targeting setting issue across US/CA/UK, identified May 28 and fixed June 1. Sessions jumped from ~2-3k/day to 8-10k/day while bounce rate spiked to 80%+ and engagement rate collapsed to ~16-27% — classic signature of junk/misdirected traffic, not real demand. Already resolved, but worth confirming with Scott that no residual mistagged traffic is still leaking into US reporting.

Bounce rate trended down and engagement up through June (bounce ~55% in April/May down to ~40-45% by late June), even as CPA was rising — meaning the traffic quality on-site was fine; the CPA problem sits in bidding/targeting/tracking, not landing page relevance.

Recommendations

Confirm with Scott the Advantage+ fix is permanent and add an alert/QA step so a repeat spike is caught same-day rather than after a few days of skewed reporting.
Since engagement quality is genuinely good, don't spend more CRO effort chasing bounce/engagement — focus remaining time on the tracking and channel-attribution issues (GA4 funnel gap, Meta channel-grouping gap) that are actively masking real performance.
Use late-June's stronger engagement rate as the baseline to monitor against going forward — any regression back toward 50%+ bounce is worth flagging immediately given how it foreshadowed the May 28 issue.

Sessions by device

Purchases by device

Purchase conversion rate by device

Insights & recommendations

Desktop converts at a meaningfully higher rate than mobile for most of the period, despite mobile carrying 3–4× the session volume — a fairly typical pattern, but worth quantifying against benchmark before assuming it's fine. Tablet volume is too low to read reliably (often single-digit sessions/day) and should stay a footnote, not a decision driver.

Mobile conversion rate dipped visibly from late June into July, around the same window as the sale ending and the account-wide CPA spike — consistent with the broader Southern/Northern conversion rate degradation already diagnosed. The iPhone Chrome checkout anomaly (Jul 9–12, flagged separately to Scott) sits inside this same window and mobile devicecategory.

Recommendations

Prioritise the mobile checkout experience for review first — it's the highest-volume, most CPA-sensitive surface, and the iPhone Chrome findings already point to a specific technical fault there.
Once GA4's add_to_cart/begin_checkout gap is closed (flagged to Scott), re-run this view split by funnel step, not just purchases — right now this can only show the final outcome, not where in the funnel mobile users are dropping.
Don't over-index on tablet numbers for any budget or bidding decisions — the sample size is too thin for the swings to mean anything.

Path to purchase: period comparison

Jun 1 – Jul 13, 2026 (current) vs Apr 19 – May 31, 2026 (previous, same 43-day length) vs Jun 1 – Jul 13, 2025 (year-on-year). US only. Channel bounce-rate rows shaded grey.

Metric	Jun–Jul 2025 (YoY)	Apr–May 2026 (prev)	Jun–Jul 2026 (curr)	% chg (prev→curr)	% chg YoY

Insights & recommendations

Volume is down across the board, but efficiency per session is up. Impressions (-29%), clicks (-46%), sessions (-32%), PDP views (-22%), add to cart (-20%), and purchases (-24%) all declined — yet purchase conversion rate actually improved (2.01% → 2.24%, +12%), sessions-to-PDP rate improved (57% → 65%, +14%), and overall bounce rate improved (52% → 46%). This is the fingerprint of the pruning work in the change log: pausing underperforming keywords/ads, adding negative keywords, and narrowing geo/audience targeting reduces reach and volume but raises the quality of who's left in the funnel.

The volume drop is the CPA story, not a quality problem. Fewer purchases against a similar or higher spend base is mechanically what pushes CPA up, even while every quality metric (bounce, PDP rate, purchase rate) is moving the right direction. This reframes the diagnosis: the account hasn't gotten worse at converting the traffic it has — it has less traffic to convert, partly by design (pruning) and partly from external shocks (Advantage+ issue, sale ending, creative learning phase).

Google and Meta bounce rates are essentially unchanged period-over-period (Google ~17.4% both periods, Meta ~55.6-56.1% both periods) — confirming the Meta creative/audience quality gap is a persistent characteristic of that channel, not something that shifted recently. It's a standing issue to address, not a new regression.

This isn't a seasonal issue — sessions are actually flat-to-up year-on-year (+4.7%), but purchases are down 7.7% and purchase conversion rate is down 11.8%. If this were purely seasonal, traffic and conversion rate would move together. Instead, a similar volume of traffic is converting meaningfully worse than it did in the same window last year, and bounce rate is also worse YoY on both channels (Google 13.5% → 17.4%, Meta 40.7% → 55.6%). That points to a genuine year-on-year decline in traffic quality or on-site experience, not just a quieter time of year.

Recommendations

Reframe the recovery plan around rebuilding qualified volume, not just protecting CPA — the pruning has done its job on quality; the next lever is finding volume that holds the same bounce/PDP/purchase rates rather than reverting to broader, lower-quality targeting.
Use the current period's purchase conversion rate (2.24%) as the new efficiency floor when evaluating any volume-expansion test — if expanded targeting drops this rate back toward 2.0% or below, it's reintroducing the same low-quality traffic that was pruned out.
Since Meta's bounce rate gap versus Google is stable and structural, prioritise the creative/audience review on Meta now rather than waiting — it's not going to self-correct.
The LP view count for the Mystery Picnic page fell 20%, roughly in line with sessions — worth checking this specific product page isn't losing disproportionate visibility in PMax/Shopping given how central it is to the offer.
Given the YoY bounce-rate decline predates this year's specific issues, it's worth checking whether last year's site/checkout experience, creative set, or audience definitions were meaningfully different — something has structurally shifted beyond this year's known events (Advantage+ bug, sale timing, creative refresh).