---
title: Ad Account Diagnostic Skill for Claude Code
canonical: https://skillselion.com/skills/mbfinotti/advertising-skills/ad-account-diagnostic
content_type: text/markdown
description: ad-account-diagnostic is a skill for the Operate phase that finds the root cause of an underperforming paid ad account and returns a ranked verdict.
---

# Ad Account Diagnostic Skill

Title: Ad Account Diagnostic Skill for Claude Code

> ad-account-diagnostic is a skill for the Operate phase that finds the root cause of an underperforming paid ad account and returns a ranked verdict.

Canonical: https://skillselion.com/skills/mbfinotti/advertising-skills/ad-account-diagnostic

ad-account-diagnostic is a skill for the Operate phase that finds the root cause of an underperforming paid ad account and returns a ranked verdict with evidence and confidence.

ad-account-diagnostic finds the root cause of a paid ad account that stopped working, instead of defaulting to raising the budget. It interviews first, reconciles platform numbers against backend truth, localizes the failing link in the funnel, then screens seven layers from tracking to external forces. It applies an Evidence Gate, issues a verdict block with confidence, and ranks fixes for handoff. It diagnoses only and never executes a change.

## Key facts

- Installs: 26,779
- GitHub stars: 2
- Author: mbfinotti
- Repository: mbfinotti/advertising-skills
- Works in Grok Bot: yes
- Install command: `npx skills add https://github.com/mbfinotti/advertising-skills --skill ad-account-diagnostic`
- Last updated: October 26, 2026

## What Ad Account Diagnostic does

Find the root cause of an underperforming paid ad account (rising CPA, falling ROAS) and hand off a ranked fix plan.

Ads stopped working and the usual reflex of raising budget or swapping creative treats one unproven guess as the cause.

You get a root-cause verdict with evidence and confidence, ranked fixes with owners, a metric prediction and a re-check date.

Best for: Founders or marketers running paid search, social, video or native whose ads used to work and who need the cause before changing anything.

Not for: Fixing post-click landing pages, writing new creative or making budget and bidding changes, which are handed off to sibling skills.

- Walks seven failure layers in a fixed order, starting with tracking
- Decomposes the failing metric before naming any cause
- Refuses a verdict the data cannot support (Evidence Gate)
- Returns a ranked, owner-tagged handoff plan with a re-check date

## SKILL.md

## Account Diagnostic

Find the root cause of an underperforming ad account before anyone touches it. The whole discipline is refusing the reflex fix: an account that "stopped working" has at least seven candidate failure layers, and the most common responses - raise the budget, swap the creative, blame the algorithm - each treat one unproven hypothesis as a verdict. This skill works the layers in a fixed order, measurement first, because every downstream number is read off the conversion signal: as Adspirer puts it, the checklist is "ordered the way an experienced PPC manager actually works it: measurement first (because every other number is wrong if tracking is broken)".

It decomposes the metric chain to localise the failing link before naming a cause, compares the account against its own history rather than industry benchmarks, and refuses to issue any verdict the data cannot support. The output is a diagnosis with confidence, evidence, and a prioritised handoff plan - never an executed change.

This skill stops at the account boundary; everything past the click hands off elsewhere, and this skill never becomes a budget skill - it exists precisely because "increase the budget" is the wrong default.

- **Past the click** (message match, page friction, the page fix list): `mbfinotti/advertising-skills@paid-landing-page-audit`. This skill only flags "the evidence points downstream of the click" and hands off.
- **Fragmentation verdict** (merge plan): `mbfinotti/advertising-skills@ad-campaign-consolidation`.
- **Creative-layer verdict** (decay analysis): `mbfinotti/advertising-skills@ad-creative-fatigue`, which hands back here when the cause is not creative - the two are reciprocal.
- **Tracking verdict** (fix checklist): `mbfinotti/advertising-skills@ad-conversion-tracking`.
- **Cross-platform discrepancy quantification**: `mbfinotti/advertising-skills@ad-attribution-gap`.
- **Search-term mining**: `mbfinotti/advertising-skills@ad-negative-keywords`.
- **Targeting design**: `mbfinotti/advertising-skills@ad-audience-targeting`.
- **Budget and bidding decisions**: `mbfinotti/advertising-skills@ad-budget-pacing`, `mbfinotti/advertising-skills@ad-spend-allocation`, `mbfinotti/advertising-skills@paid-media-scaling`, and `mbfinotti/advertising-skills@ad-bidding-strategy`.

### Interview

Ask before opening any data. One question per message; offer multiple-choice answers where possible; skip anything already supplied or visible in the data.

- Which platform(s) does the account run on? (search / paid social / video / native / several)
- B2B or B2C/ecommerce?
- Monthly spend, and roughly how many conversions per month on the event the account optimises to? (Decides whether the Evidence Gate can clear at all.)
- What window is under suspicion - when did performance degrade, and against which prior period is it being judged?
- What changed, and exactly when? (Budget, bids, creative, targeting, optimization event, landing page, site release, consent banner, price, promo calendar - dates matter more than the list.)
- What can you export, and at what granularity? (Per-campaign per-day is the working minimum; per-ad-set and per-ad breakdowns unlock the localisation step.)
- What is the target CPA/ROAS - and is it derived from unit economics or inherited from a dashboard?
- Is CRM or backend revenue data reachable for reconciliation? (Yes, directly / yes, via someone / no.)
- Which attribution window and counting settings does reporting use, and did they change in the window?
- Who will implement the fixes, and what is their effort ceiling - hours available, whose hands (developer, marketing ops, creative studio, page owner), and how reversible a change is allowed to be? (The handoff plan is addressed to them; see Prioritisation for how a ceiling reshapes the queue.)
- By what date does the recovery have to show up in reporting? (A hard deadline promotes fast-acting fixes - tracking repair, unblocking a genuinely budget-capped campaign - and demotes offline-outcome wiring and creative programmes, whose payoff lands a window or two later.)
- One-off recovery, or a compounding asset? (A compounding mandate promotes the offline-outcome feedback loop and structural consolidation above the quick wins; see Prioritisation.)

### Workflow

1. Run the Interview; collect every answer before touching data.
2. **Reconcile before interpreting.** Compare platform-reported conversions against CRM/backend truth over a lag-mature window, against OnlyDeb's practitioner reference points:
   - Gap above roughly 40%: something is broken - stop, record a tracking FAIL, and hand off.
   - Gap within roughly 10%: proceed to the other layers.

   These are attributed starting points, not laws - the account's own historical ratio, and whether that ratio is _stable_ week to week, is the real criterion. If backend data is unreachable, every later finding carries at most medium confidence, and the verdict must say so.

3. **Decompose the failing metric** with [references/metric-decomposition.md](references/metric-decomposition.md). Chain the funnel - impressions → clicks → conversions → revenue - and express the outcome as CPM × CTR × CVR × AOV (Pigeon Digital) to find which single link actually moved. Name no cause before the failing link is localised.
4. **Break down and compare.** Pull the failing metric by campaign, then ad set, then ad, and compare each against the account's own history - never industry benchmarks. AdStellar's rule for CPM generalises: elevated evenly across everything points external; concentrated in one branch points to a targeted internal cause. This step is what separates internal from external causes.
5. **Walk the layers in fixed order** (below) using [references/layer-evidence.md](references/layer-evidence.md). For each layer record one of four states - `pass` (evidence rules it out), `FAIL` (evidence confirms it), `unknown` (evidence missing), `not applicable` - with its own severity and confidence. Never collapse to a binary: `unknown` is not `pass`, and treating an unchecked layer as clean is the most common way audits go wrong.
6. **Apply the Evidence Gate** (below). If it fails, the verdict is `insufficient evidence`: state exactly what additional days, spend, or access would clear it, and stop. No verdict on underpowered data.
7. **Issue the Root-Cause Verdict** (block below; worked versions in [references/examples.md](references/examples.md)). A compound verdict - two unrelated confirmed causes - is legitimate; rank them.
8. **Rank the findings by efficiency** (Prioritisation section) and write the handoff: which sibling skill or owner takes each finding, in rung order.
9. **Log the prediction.** Every verdict must state which metric should move, in which direction, by roughly how much, once the recommended fix ships - this is the raw material for the Measuring section.
10. Set a re-check date one full comparison window (at equal attribution-lag maturity) after the fix ships.
11. If your harness has persistent memory, memorise the account's baselines, the per-layer screen, the verdict, fixes applied, predictions, and re-check dates, so the next run starts from history instead of re-deriving it. If it does not, emit a short state block the user can paste into the next session.

If your harness can read the exports or run the calculations, compute every step directly; otherwise emit the exact export steps and spreadsheet formulas (columns, ratio, delta `(current - baseline) / baseline`) for the user to run and report back.

### The Layer Order

Fixed order, because each layer's evidence is only readable if the layers before it hold:

- **Tracking** corrupts every number.
- **Structure** corrupts the data pooling and learning that targeting and creative reads depend on.
- **Bidding** reads assume the auction inputs above it are sane.
- **Offer & downstream** sits past the click.
- **External** causes are a diagnosis of exclusion, claimed only when internal evidence rules the others out - the uniformity test in step 4 can fast-path there, but never skip the tracking check to get to it.

Baker's 8-layer framework states the why: auditing creative on an account whose pixel feeds 56% wrong data "produces conclusions that look right but are mathematically meaningless" - and across B2B SaaS accounts spending $3K-$250K/month he reports 80% of underperformance tracing to pixel health and creative diversity, not bidding or budget. This is a sequence, not a menu: the layers are not alternatives to pick between, so they carry no efficiency ranking - every one gets screened, in this order. The ranking lives one step later, over the _fixes_ the screen produces (see Prioritisation).

1. **Measurement / tracking** - is the conversion signal real, deduplicated, and consented?
2. **Structure** - fragmentation, overlapping campaigns self-competing, budget traps, wrong optimization event.
3. **Targeting** - audience overlap, saturation, too narrow/broad, wrong intent.
4. **Creative** - resonance decay; the differential itself runs in `mbfinotti/advertising-skills@ad-creative-fatigue`.
5. **Bidding / budget** - strategy-volume mismatch, learning state, impression share lost to budget vs rank.
6. **Offer & downstream** - price, promo, page, checkout; flag and hand to `mbfinotti/advertising-skills@paid-landing-page-audit`.
7. **External** - auction inflation, seasonality, competitor entry; confirmed by uniform elevation plus market evidence.

Per-layer confirming evidence, ruling-out evidence, and the check that settles each: [references/layer-evidence.md](references/layer-evidence.md).

### Evidence Gate

Refuse a verdict the data cannot carry. All four checks must pass before any layer FAIL becomes a verdict:

- **Conversion volume.** Enough conversions in both windows that the observed delta exceeds noise (band ≈ `p ± 2 × sqrt(p × (1-p) / n)` on the relevant rate). Attributed reference points, not laws: Google recommends evaluating automated bidding over periods with at least 30 conversions for Target CPA and 50 for Target ROAS (Google Ads Help), and OnlyDeb notes campaigns under roughly 15-30 conversions/month lack signal for the algorithm at all - low-volume accounts (most B2B) gate on leading metrics instead and say so in the confidence line.
- **Learning-phase state.** No verdict from data collected while delivery is recalibrating. Reference points: Meta ad sets typically exit learning after roughly 50 optimization events in 7 days (Meta Business Help Center, as relayed by AdStellar), and Google Smart Bidding can take up to three weeks or 1-2 conversion cycles (Google Ads Help). Niblin: ROAS swinging 20-50% day-to-day is normal during learning - judge only after exit or 7+ days.
- **Window comparability.** Baseline and comparison windows must have equal duration, equal attribution-lag maturity (recent days always under-report; a trailing window always "looks like" decline), matched day-of-week composition, and no known outages. Jyll Saskin Gales audits all active campaigns together over a 90-day window because campaigns "interact with each other" - never judge one campaign in isolation over a cherry-picked week.
- **Attribution consistency.** Never sum or compare conversions across differing attribution windows, counting methods, or models - a 7-day-click figure plus a 30-day-click figure is not a total, it is a category error. Report side by side until definitions reconcile; a settings change mid-window invalidates the comparison entirely.

When the gate fails: verdict `insufficient evidence`, plus the arithmetic of what clears it - the additional days or spend needed for `n` to shrink the noise band below the observed delta at current traffic, or the specific access (CRM export, ad-level breakdown) that is missing.

### Root-Cause Verdict

Deliver one block per diagnosis:

```
ROOT-CAUSE VERDICT - <account>, <date>
platform(s)    : <...> | model: B2B | B2C
window         : <comparison window> vs baseline <baseline window> (lag maturity matched: yes/no)
volume         : <spend>, <conversions> conversions in window | reconciliation gap: <x% vs backend | unverified>

decomposition  : <which link moved: CPM / CTR / CVR / AOV - direction and size vs own history>
localisation   : <uniform across account | concentrated in: campaign/ad set/ad>

layer screen
  measurement/tracking : pass | FAIL | unknown | n/a - <evidence>   [severity, confidence]
  structure            : pass | FAIL | unknown | n/a - <evidence>   [severity, confidence]
  targeting            : pass | FAIL | unknown | n/a - <evidence>   [severity, confidence]
  creative             : pass | FAIL | unknown | n/a - <evidence>   [severity, confidence]
  bidding/budget       : pass | FAIL | unknown | n/a - <evidence>   [severity, confidence]
  offer & downstream   : pass | FAIL | unknown | n/a - <evidence>   [severity, confidence]
  external             : pass | FAIL | unknown | n/a - <evidence>   [severity, confidence]

confidence     : high | medium | low - <basis: gate math, reconciliation state, unknowns count>
verdict        : <root cause layer(s), or insufficient evidence - with the one-sentence causal story>
evidence       : <the two or three observations that carry the verdict>
findings       : <ranked by efficiency - each with severity, confidence, outcome bought, effort, owner>
prediction     : <metric expected to move, direction, rough size, by when>
handoff        : <sibling skill or owner per finding, in sequence>
re-check       : <date - one full window after fix, lag-mature>
```

### Prioritisation

Rank findings before handing off - a flat list of twenty observations is a pitch deck, not a diagnosis. Rank by efficiency: the outcome each fix buys per unit of effort it costs. Never by cheapest, and never by the layer screen's own row order - the screen is a diagnostic sequence, the ranking is a work queue.

Default fix-class order, highest value/effort ratio first:

1. **Tracking repair** - buys back the readability of every other number in the account and puts delivery on a real signal again; costs an hour to half a day of one developer, most of it end-to-end testing. Always ships first: every later fix is measured through it.
2. **Optimization-event / offline-outcome fix** - buys a platform that chases qualified outcomes instead of cheap volume (Koda: the offline feedback loop "consistently improves lead quality more than any targeting adjustment"); costs a couple of days of marketing-ops and CRM wiring, most of it field mapping and testing, then keeps paying without further work.
3. **Structural consolidation** - buys starved units the conversion volume they need to exit learning; costs a couple of days of rebuild plus one learning window of volatility, and un-merging later is expensive.
4. **Targeting redesign** - buys reach on audiences that are not yet saturated; costs about a day, and reverses cleanly.
5. **Offer & downstream fixes** - the outcome can beat everything above it, but the work sits outside the account: costs coordination with whoever owns the page, the price or the checkout, so their calendar sets the pace instead of the auditor's.
6. **Creative replacement** - buys the click back where decay is confirmed; costs a production cycle and never ends - a standing job, not a fix.
7. **Budget and bid moves** - buy nothing until every rung above passes, and each significant edit spends a learning window (see "Why 'Increase the Budget' Is the Wrong Default").

The axes disagree, which is exactly why the reflex fix keeps winning arguments:

- efficiency: tracking > event fix > consolidation > targeting > offer & downstream > creative > budget moves
- value: offer & downstream == tracking > event fix > consolidation > creative > targeting > budget moves
- effort: offer & downstream > creative > event fix == consolidation > targeting > tracking > budget moves

Excerpt. The rest of the file is on the canonical page and in the source repository.

## Ad Account Diagnostic FAQ

### Why does ad-account-diagnostic refuse to say increase the budget?

It treats the budget reflex as one unproven hypothesis. The skill lists mechanisms such as learning resets, lost impression share to rank, diminishing returns and platform-claimed versus incremental revenue.

### What layers does ad-account-diagnostic check?

Seven layers in fixed order: measurement and tracking, structure, targeting, creative, bidding and budget, offer and downstream, and external. Each gets pass, FAIL, unknown or not applicable.

### What happens when the data is too thin?

The Evidence Gate returns an insufficient evidence verdict and states the extra days, spend or access, such as a CRM export, that would clear it. No verdict is issued on underpowered data.

### Is Ad Account Diagnostic safe to install?

skills.sh reports 3 of 3 security scanners passed. Review the Security Audits panel on this page before installing in production.

## Machine access

- [Sitemap](https://skillselion.com/sitemap.xml)
- [llms.txt](https://skillselion.com/llms.txt)
- [llms-full.txt](https://skillselion.com/llms-full.txt)
- [Developer resources](https://skillselion.com/developers)
- [OpenAPI spec](https://skillselion.com/openapi.json)
- [MCP handshake](https://skillselion.com/.well-known/mcp)
- [API catalog](https://skillselion.com/.well-known/api-catalog)
- [AI catalog](https://skillselion.com/.well-known/ai-catalog.json)
- [Pricing](https://skillselion.com/pricing.md)
- [Auth](https://skillselion.com/auth.md)
- [AGENTS.md](https://skillselion.com/AGENTS.md)
- [Contact](https://skillselion.com/contact)
- [Home](https://skillselion.com/)
