Agency new client audit

New-Client Onboarding Audit

Operating question: What is the state of this new client's account, and what do we fix in week one?

Run this as an evidence-first operating play. Gather the evidence below, separate exact / estimated / partial / unavailable data, apply the decision rules and vetoes, and never invent missing numbers. Be adversarial toward platform self-reported figures — trust commerce truth.

Evidence to gather

• Commerce (Shopify/Woo/BigCommerce/etc.) — total orders and revenue for the last 30 and 90 days, AOV, new-vs-returning split, top products by revenue, and the store's checkout/theme setup. This is the source of truth everything else is reconciled against; without it, the audit cannot run.
• Analytics (GA4) — sessions, conversions, purchase events, and revenue for the same windows, plus the data-stream/tag configuration. Needed to measure tracking drift against commerce.
• Paid (Meta Ads and/or Google Ads) — spend, conversions, conversion value, campaign types (Search / Shopping / PMax on Google; prospecting vs. retargeting on Meta), and branded-search exposure. Needed for the paid-waste scan.
• Targets and context — the client's own stated CPA/ROAS/MER targets if they have them, blended margin or COGS if available, and any known recent migrations, promos, or platform changes.

At minimum this play needs commerce truth plus at least one of analytics / ads / email. With only one platform you can describe a corner of the account, not audit it.

Optional, if available:

• Klaviyo (or other ESP) — which core flows exist, whether they're live and sending, list size, recent campaign send/open/click, and deliverability signals (spam-rate, sending domain authentication). Lets the lifecycle dimension move from "unknown" to a real finding.
• Google Merchant Center — product-feed status and the disapproval breakdown, ideally joinable to top-selling SKUs. Lets the feed-health dimension produce a ranked, revenue-weighted finding instead of a raw disapproval count.
• Search Console / branded-search data — to size how much paid branded spend is cannibalising free clicks.
• Stock cover for top products — a winner that's about to stock out changes which "quick win" is safe to promise.
• Prior-agency context — what the last team last touched, so you don't re-break a deliberate setup.

How to decide — in order

1. Confirm access and scope before anything. List which platforms you actually have, at what scope. If commerce truth is missing, or you have fewer than commerce-plus-one platform, set the missing areas to FIX (access) and audit only what you can defend. Never infer the account's state from a platform you can't see.
2. Establish tracking trust first — it gates everything downstream. Reconcile commerce orders against GA4 purchases and confirm pixel + CAPI present and deduplicated. If drift exceeds ~15% or dedup is broken, mark tracking FIX and treat every paid/analytics number below as unsafe to act on until it's repaired. You cannot judge paid waste on numbers you don't trust.
3. Score feed health against revenue, not volume. A 200-SKU disapproval list matters far less than 3 disapproved hero SKUs that drive 40% of revenue. Rank disapprovals by the revenue of the SKU behind them; raise the revenue-weighted ones as FIX, leave the long tail as WATCH.
4. Audit lifecycle by what's actually sending. A "Live" flow with no recent recipients is FIX, not KEEP. Missing core flows are REFRESH (build them). Deliverability problems (poor authentication, high spam rate) outrank flow gaps — a great flow into the spam folder is worth nothing.
5. Find the obvious paid leaks — conservatively. Only flag zero-return spend you can defend: 0 conversions over 14+ days with real clicks, or branded cannibalisation you can size. Anything ambiguous on a fresh account is WATCH with an access caveat, never KILL on day three.
6. Pick the top 3–5 quick wins, then split the output. Rank findings by impact × confidence × speed. Produce three lists: the week-one fix list (what you'll do now), the do-not-touch-until-mapped list (working setups you don't yet understand), and the kickoff red flags (what the client needs to hear).

The prompt to run

You are my agency onboarding analyst running the "New-Client Onboarding Audit" play.

GOAL: produce a read-only week-one audit of a brand-new client's account across five
areas — tracking trust, feed health, lifecycle/email, paid waste, and the top 3-5 quick
wins — then split the result into a week-one fix list, a do-not-touch-until-mapped list,
and the red flags to raise on the kickoff call.

I will paste: commerce orders/revenue, GA4 conversions, Meta/Google spend tables, GMC
feed diagnostics, and Klaviyo flow status. Some of these may be missing.

RULES:
- PRE-FLIGHT: First list which inputs I have access to vs. missing. If I lack access to the
  client's commerce data plus at least one of analytics/ads/email, STOP and list exactly
  what access to request — never guess the account's state.
- Establish tracking trust first. If commerce orders and GA4 purchases diverge >15%, or
  pixel/CAPI dedup is broken, mark tracking FIX and treat all paid/analytics numbers below
  as unsafe to act on until repaired.
- Score feed disapprovals by the REVENUE of the SKU behind them, not by raw count.
- Judge Klaviyo flows by whether they are actually SENDING, not just "Live". Deliverability
  problems outrank missing flows.
- Flag paid waste conservatively: only zero-return spend you can defend (0 conversions over
  14+ days with real clicks, or sized branded cannibalisation). Ambiguous = WATCH with an
  access caveat, never KILL on a fresh account.
- Never recommend changing anything before the account is mapped. This run is read-only.
- Every finding must carry: a number, its source, the time window, and a confidence level.
- Separate exact / estimated / partial / unavailable evidence. Do not invent missing data.

RETURN:
1. A 3-sentence executive read of the account's overall state.
2. A findings table using exactly this header row:
   | Area | Finding | Severity | Week-1 action | Owner | Confidence |
   Use "—" for any cell you cannot fill. Do not add or drop columns, and do not replace the
   table with prose.
3. A do-not-touch-until-mapped list (working setups you don't yet understand).
4. The top 3 red flags to raise on the kickoff call.
5. What evidence is blocked and exactly what access I should request to upgrade it.

Decision rules

• FIX — tracking drift between commerce and GA4 > ~15%, pixel/CAPI dedup broken (no shared event_id), a disapproved SKU in the top-revenue decile, a "Live" flow with 0 recent sends, or broken sending-domain authentication. These are the week-one repairs.
• REFRESH — a core lifecycle flow is missing entirely (no welcome, no abandoned-checkout), or feed attributes are weak but not disapproved. Build or improve, don't just toggle.
• KILL — paid spend with 0 conversions over ≥14 days with ≥300 real clicks and confirmed-working tracking, or branded-search/PMax spend you can prove is cannibalising free organic clicks. Only after access and tracking are verified.
• WATCH — anything ambiguous on a freshly-granted account: a 0-conversion column you suspect is an access/attribution artefact, a flow toggled live the day you got access, a feed in "pending review."
• KEEP — a setup that is demonstrably working and within the client's stated targets; leave it alone and note it as a baseline.
• Every recommendation carries a number, source, time window, and confidence level — on a handover, an unsourced claim is how you set a target you can't hit.

Vetoes — stop if any apply

• Do not change, pause, or edit anything before the account is mapped. This is a read-only audit; week one is for understanding, not surgery.
• Do not infer the account's state from a platform you can't see — flag the missing access explicitly and audit only what you can defend.
• Do not promise a win you can't yet evidence. "We'll lift ROAS 30%" before you've confirmed tracking is how you lose trust in month two.
• Do not call paid spend "wasted" until you've confirmed the conversion-tracking link actually reports into the account you can see.
• Do not judge a flow KEEP on "Live" status alone — require evidence it is sending to real recipients.
• Do not raise a "pending review" feed or a 24–48h-old GA4 gap as a red flag; those are normal fresh-access artefacts.

Output

A findings table, ranked by severity then impact:

Area	Finding	Severity	Week-1 action	Owner	Confidence
Tracking	GA4 purchases trail Shopify orders by 18% (30d)	FIX	Repair data layer; add CAPI dedup	Analytics	High

Followed by: a do-not-touch-until-mapped list, and a top-3 kickoff red flags list.

---

Run this on your live store data in one prompt — start a free 14-day trial of ShopMCP (no credit card). It connects your store, ads, analytics, email/SMS, and finance and runs this play read-only by default, applying changes only on your approval.