# Meta ad account audit checklist (Ben Heath style)

Work through the sections in order. Score each line as one of:

| Score | Meaning |
|---|---|
| ✅ OK | Meets the rule |
| ⚠️ Fix | Needs a change |
| ❌ Critical | Seriously wrong |
| n/a | Doesn't apply |
| Unknown | Required evidence is missing; not a pass or failure |

Every check uses a rule from `framework.md`. The section number is in brackets.
- Look at the last 30 days and compare with the 30 days before.
- For seasonal businesses, compare with the same period last year [§10].
- Don't judge on day-to-day numbers.

---

## 0. Inputs to collect before opening Ads Manager

- [ ] Business model (e-commerce, lead gen, local service, high-ticket, app) and what it sells
- [ ] Target CPA/CPL or target ROAS, and breakeven ROAS. Derive these from margin and costs. [§1]
- [ ] Average order value, customer LTV, repeat rate, refund or cancellation rate
- [ ] For leads: lead→contact, lead→qualified and lead→customer rates, and lead value
- [ ] Monthly budget and the client's appetite for testing
- [ ] Source of truth for sales and leads (Shopify, Salla, Zid, CRM), to compare with Meta
- [ ] Seasonality calendar
- [ ] Any differentiator or offer elements (guarantee, urgency, proof)

**Decision boundary:** without target CPA/ROAS and margins, defer profitability and scaling decisions. Continue checks supported by available evidence; label missing campaign or tracking data as Unknown. CRM traffic sources alone do not establish paid attribution or ROAS.

---

## 1. Tracking and data [§2]

- [ ] Pixel is installed. Conversions API is active alongside it (not one or the other).
- [ ] Automatic Advanced Matching is on. Check event match quality and Events Manager suggestions.
- [ ] Purchase or lead events fire only on true completion, never on page load. Check with Meta Pixel Helper.
- [ ] Event counts in Meta roughly match the backend (orders, CRM leads). Note the size and direction of any gap.
- [ ] Purchase events send value and currency.
- [ ] **Engaged audience** and **existing customers** are defined in Advertising settings → Audience segments. The breakdown shouldn't show "Unknown".
- [ ] Customer and email lists are uploaded (at least 100 people, value column, labels) and refreshed recently.
- [ ] Attribution setting is known (default 7-day click, 1-day view, 1-day engaged-view). The team understands that Meta under-reports long-cycle and recurring revenue.
- [ ] A special ad category is declared where it's required.
- [ ] No boosted posts are running from the apps.

## 2. Objectives and optimisation [§3, §11]

- [ ] Every direct-response campaign uses the **Sales** or **Leads** objective.
- [ ] Performance goal is a conversion event. ❌ if a Sales or Leads campaign optimises for link clicks, landing-page views, reach or impressions.
- [ ] There are no Traffic or Engagement campaigns meant to "warm up the pixel".
- [ ] Optimisation event is the deepest one with enough volume: purchase at 20–25 or more a week, otherwise one step up the funnel.
- [ ] Sales accounts with varying order values use **Maximize value of conversions** (once there's enough data).
- [ ] Lead gen: only *qualified* leads fire the optimised event.
  - Instant forms use conditional logic with an end page for non-leads.
  - Website forms send unqualified people to a page with no lead event.
  - Conversion leads or CRM integration where possible.
- [ ] Instant forms use **Higher intent** plus **SMS one-time passcode**, autofill off for email and phone, and a mandatory phone number (except for lead magnets). Keep forms short.
- [ ] Speed-to-lead: leads are contacted within minutes. Messenger, Instagram or WhatsApp follow-up is set up where relevant.
- [ ] Any Traffic, Awareness or Engagement campaign has a valid reason: no pixel possible, a profile or page-follower goal, or omnipresent content.

## 3. Structure [§4]

- [ ] Count campaigns and ad sets. List **optimised conversions per ad set per week**.
  - ⚠️ if most ad sets get fewer than about 50 a week and could be consolidated.
  - ❌ if many ad sets get fewer than 10 a week.
- [ ] Learning Limited ad sets: fix by consolidating, not by raising the budget on an unprofitable campaign.
- [ ] One campaign per product range, offer or market. Not split by variant, colour or interest.
- [ ] **No duplicate campaigns** created to scale.
- [ ] Cold and warm aren't split by default. Any split has a messaging reason (warm-only offer, upsell).
- [ ] No separate top-, middle- and bottom-of-funnel campaigns. Funnel-stage ads sit together in one ad set.
- [ ] No audience A/B ad sets (interest vs interest, interest vs lookalike). Only location splits are valid.
- [ ] A testing campaign exists only if the account does 50 or more optimised conversions a week.
- [ ] Budgets are daily, not lifetime. Lifetime only where scheduling or sequencing needs it.
- [ ] Any exception (omnipresent content, location tests, product ranges) is deliberate and documented.

## 4. Targeting [§5]

- [ ] Controls hold only:
  - Location (where the business can serve)
  - Minimum age, only if a law requires it
  - Language, only if it matters
  - Necessary exclusions
- [ ] Suggestions include **all** custom audiences: website 180d, video 3s 365d, IG and FB engagers 365d, lead forms 90d, customer and email lists.
- [ ] Age and gender suggestions are used only where there's a real skew (gender about 90/10).
- [ ] "Further limit the reach" (manual or hard targeting) is used only for a stated reason. If it is used, warm and cold are in separate ad sets.
- [ ] No reflexive exclusions of past purchasers or engaged audiences.
  - To focus on new customers, use Customer Lifecycle Strategy, not exclusions, and only for big-share or one-purchase brands.
- [ ] Lookalikes, if used:
  - Sized sensibly (US 1%, UK 2%, small countries larger %).
  - Treated as cold.
- [ ] **Value rules** are configured where segment values differ (age, gender, location, device, OS, placement, conversion location).
  - Percentages come from real CRM or LTV data, not Meta's +50% default.
  - Rules are applied at the ad set level.
- [ ] Gender exclusion need: done with a −90% value rule (hard gender targeting was removed in Nov 2026).

## 5. Placements [§5]

- [ ] Sales and Leads campaigns with conversion goals use **Advantage+ placements**. No trimming based on the breakdown.
- [ ] Non-conversion campaigns (reach, traffic, engagement):
  - Audience Network, Messenger, Threads, in-stream and search are removed or bid down: account-level placement controls, or a placement value rule at −90%.
  - "Allow limited spend to excluded placements" is unticked where it's available.
- [ ] Check the placement breakdown for outliers. Act through value rules, not hard exclusions, on conversion campaigns.

## 6. Creative [creative.md]

- [ ] Active ads per main ad set:
  - ⚠️ if only 1–3 on a decent budget.
  - The goal is many high-quality, *different* ads (20+ is fine after Andromeda; start with at least 4 genuinely different ones).
- [ ] Diversity is present across:
  - Formats: static, video, carousel.
  - Styles: UGC, creator partnership, founder, demo, testimonial, animation.
  - Angles, avatars and awareness stages.
  - Background or colour tweaks don't count.
- [ ] Both statics and video are running. The 2026 trend favours statics, but don't drop video for new concepts or cold expertise selling.
- [ ] One angle per ad. Ads aren't feature-stuffed.
- [ ] Headline states the offer clearly. Primary text has a call-out, benefits and a CTA, and its length matches the ask.
- [ ] Copy variations sit inside each ad (up to 5 texts or headlines), not in separate ads.
- [ ] Proof is in every ad: testimonials, results, reviews, work photos.
- [ ] Hooks:
  - Hook rate is measured. Above 10% is good, below 5% is poor.
  - Several hooks exist per winning video.
- [ ] Creator or partnership ads are used, or there's a plan for them (Creator Marketplace, boost permission, 2–4 creators).
- [ ] Native look:
  - No repurposed TV or billboard creative.
  - No hashtags.
  - No AI-looking statics.
  - Clear CTA.
  - Assets fit feed, vertical and landscape placements.
- [ ] **Creative enhancements:** check every ad. Meta re-enables them after duplicate, edit or preview. Turn off off-brand overlays and AI image variations, and leave them off in regulated industries.
- [ ] **Personalized destinations:** off, unless it's pure e-commerce with deliberate intent.
- [ ] Landing page matches the ad (product, offer, colours, imagery). Product-specific pages, not the homepage. Fast load: link click → landing page view drop under about 20–50%.
- [ ] Fatigue: winners with rising frequency and falling ROAS have replacements ready.

## 7. Testing [§7]

- [ ] There's a testing process, and new ads actually receive spend.
- [ ] The mechanism fits volume:
  - Push delivery or new ads in the same ad set when below about 50 a week.
  - Creative Testing tool at dozens to hundreds a week.
  - A testing campaign only at 50 or more a week.
- [ ] Creative Testing tool comparison metric is CPA, CPL or ROAS, not cost per engagement.
- [ ] Tests isolate one variable, start from a hypothesis, and follow the priority order: offer → angle → format/style → hook → small details.
- [ ] Tests run long enough: enough conversions, about 85–90% significance, usually 7–10 days or longer at low volume.
- [ ] The best ad wasn't switched off to force spend onto tests.
- [ ] Testing share of budget fits the situation:
  - 10–20% when very profitable.
  - 20–50% normally.
  - Up to 100% when nothing works.

## 8. Budgets, change cadence and scaling [§7, §8, §9]

- [ ] Change log shows edits no more than about once every 7 days per ad set, made in batches. ⚠️ for daily tinkering.
- [ ] Small budgets (under $3,000 a month):
  - One offer, one campaign, one ad set.
  - No awareness or traffic spend.
  - Niche messaging.
- [ ] Scaling method:
  - Either automated rules: +3% a day under target and −3% over a higher threshold, 7-day window, lifetime impressions above 8,000, cap and floor.
  - Or stepped manual increases: wait 5–10 days between steps, with smaller percentages at higher spend.
- [ ] No 10x jumps. No horizontal duplication.
- [ ] ROAS headroom exists before scaling (e.g. 4–5x when the minimum is 3x).
- [ ] The scaling ceiling is known: the spend at which ROAS hits the minimum.
- [ ] Spend follows seasonality: more in strong periods, less in weak ones.

## 9. Metrics review [§10]

- [ ] The north-star metric (ROAS or CPA/CPL) is agreed. Nobody is killing ads on CTR or CPM.
- [ ] The column preset includes results, CPR, value, ROAS, reach, impressions, frequency, CPM, hook rate, average play time, link clicks, landing page views, CPC (link) and CTR (link).
- [ ] Frequency is about 2.5 or less on cold, and 6–8 on warm for most accounts.
- [ ] Link CTR: above 1% good, above 2% very good, below about 0.65% poor.
- [ ] E-commerce click-to-purchase rate: under 2% means you should investigate the landing page.
- [ ] Breakdowns have been reviewed:
  - Audience segments (new, engaged, existing)
  - Placement and platform
  - Demographics (change messaging for any skew)
  - Video retention curves
- [ ] Ads that get spend but few direct conversions are checked for a top-of-funnel role before being turned off.

## 10. Beyond the buttons [§0, §13]

- [ ] Offer strength against the value equation:
  - Dream outcome
  - Likelihood (proof, guarantee)
  - Time delay
  - Effort
- [ ] Clear differentiator against competitors (check the Ad Library for their long-running ads).
- [ ] LTV levers: upsells, bundles, subscriptions, repeat purchase flows, referrals.
- [ ] Brand assets: founder or creator presence, reviews, social proof volume.
- [ ] Strategy fits the business. Consider omnipresent content only for LTV of about $1,000 or more, expertise or high-ticket businesses.

---

## Diagnosis and report

1. Place the account in the 5-step diagnostic:
   - No conversions → offer
   - Low hook rate and CTR → creative
   - High CTR, low conversion → landing page
   - Many conversions but unprofitable → business model
   - Profitable → scale
2. List ❌ items first, then ⚠️. Order by leverage:
   1. Offer and optimisation event
   2. Tracking
   3. Structure consolidation
   4. Creative
   5. Budget and scaling hygiene
3. For each fix, give:
   - The rule behind it
   - The expected trade-off (e.g. "CPL will rise but lead quality and close rate should improve")
   - When to review it, based on conversion volume
4. Propose a 30-day plan:
   - Week 1: tracking and optimisation fixes
   - Week 1–2: consolidation
   - Ongoing: creative batches and tests
   - Fixed review dates
   - No changes in between
