---
name: actcenter_weekly-health-check
description: Run the weekly health check on any Google Ads account — pacing, CPA, lead volume, conversion stability, tracking integrity, structural drift, and the tCPA Health Check framework (5 states: HOLD, RAISE, LOWER, DIAGNOSE, RISK). Use whenever the user says "weekly check", "health check", "sprint check", "como tá a conta do [client]", "review semanal", "checa essa conta", or shares a Google Ads CSV/screenshot and asks for a triage. Built around ACTCENTER's Priority Engine — outputs the top 3 actions for the week, not a 20-page report.
metadata:
  type: skill
  version: 1.0
  owner: actcenter
---

# actcenter_weekly-health-check

The 6-minute weekly read on any Google Ads account. Produces a Priority Engine output: what's healthy, what's drifting, what needs an action this week.

## When to use

Trigger on any of:

- "Weekly check on [client]"
- "Health check"
- "How is [account] doing?"
- User pastes a Google Ads CSV / Excel / screenshot and asks for review
- "Quick read on this account"
- "Should I worry about [account]?"

Do not use for:

- Full account audit (use `actcenter_account-auditor` or equivalent — that's a 200-checkpoint deep dive)
- New account buildout
- Creative-only review

## Inputs the skill expects

Minimum:

1. Last 30 days of account-level KPIs: spend, conversions, CPA, conversion-rate.
2. Last 7 days vs prior 7 days (deltas).
3. MTD spend + monthly budget (for pacing).
4. Bid strategy per campaign.

Nice-to-have:

5. Top 10 campaigns by spend (90 days).
6. Search-term sample (last 14 days, top 50 by cost).
7. Quality Score distribution.
8. Conversion-action list (which actions count, are any importing from offline / CRM).

Ask the user to paste the data or upload the CSV. If the data is incomplete, name exactly what's missing — don't guess.

## The framework — tCPA Health Check (5 states)

Every account, every week, lands in one state:

| State | Definition | Action |
|---|---|---|
| **HOLD** | CPA within ±10% of target. Volume stable. No tracking issues. | Don't touch. Note one observation. |
| **RAISE** | CPA at or below target AND impression share lost to rank ≥ 30%. Headroom for more volume. | Lift tCPA 10–15%. Re-check in 14 days. |
| **LOWER** | CPA ≥ 20% above target for 3+ weeks. Volume not the bottleneck. | Cut tCPA 10–15%. Re-check in 14 days. |
| **DIAGNOSE** | CPA volatile (±25% week-over-week) OR conversion volume dropped > 30% with no campaign change. | Investigate before adjusting. Look at tracking, search terms, auction pressure. |
| **RISK** | Tracking broken, pacing < 50% or > 200%, account flagged, or conversion path obviously compromised. | Stop and fix. No tCPA move until root-cause is known. |

A separate `tCPA Health Check` deep-dive skill validates whether a tCPA *change* is actually justified (`actcenter_tcpa-scale` for growth, `actcenter_tcpa-migration` for moving onto Smart Bidding). This skill identifies *which* state — those decide *whether* to act.

## Output — Priority Engine

Every health check ends with the same three blocks:

```markdown
# Weekly Health Check — [client] — Week of [YYYY-MM-DD]

## State
**[HOLD / RAISE / LOWER / DIAGNOSE / RISK]**

## Pacing
- MTD: $X,XXX of $Y,YYY budget
- Day of month: D / DD
- Pacing %: NN%
- On-pace target: NN%

## Top numbers
| Metric | Last 7d | Prior 7d | Δ |
|---|---|---|---|
| Spend | … | … | … |
| Conversions | … | … | … |
| CPA | … | … | … |
| Conv rate | … | … | … |
| Search IS | … | … | … |
| IS lost (rank) | … | … | … |

## Hidden risks
- [issue 1 with severity tag]
- [issue 2]

## Opportunities
- [opp 1]
- [opp 2]

## This week's 3 actions
1. [action] — owner — by day
2. [action] — owner — by day
3. [action] — owner — by day

## What I'm explicitly not changing this week
- [item 1] — [reason]
- [item 2] — [reason]
```

## How to run

```
1. Confirm: which account, which 7-day vs 7-day window, do you want the file as MD or PPTX?
2. Read the input data. Compute the state.
3. Identify the top 3 actions for the week — no more, no less. Force-rank by expected impact.
4. Write the report. Save to /weekly-checks/{client-slug}/{YYYY-MM-DD}.md
5. If user asked for PPTX, hand off to pptx skill with the same content.
```

## Hidden-risk catalog (what to look for every week)

- **Tracking drift** — conversion volume dropped > 30% with no campaign change → check the tag, GTM, conversion-linker.
- **Search-term cluster drift** — cost/cheap/free queries pulling > 10% of spend → add negatives.
- **PMax cannibalization** — branded queries showing in PMax search-term insights and Brand campaign impressions drop.
- **Ad disapproval cluster** — multiple ads disapproved at once → likely policy update, batch-review.
- **Bid strategy mismatch** — Maximize Clicks running on a conversion-led account → migration candidate.
- **Budget-capped campaign** — campaign at "limited by budget" daily → re-allocate or lift cap.
- **Auction pressure shift** — IS lost (rank) up > 15 pts WoW → new entrant or competitor ramped.
- **Conversion path mismatch** — phone-call conversions firing but no qualifying duration filter.

## Voice & format

Direct. Numbers before adjectives. No hype. Frame each finding as **observation → impact → recommended action**. Avoid the words *amazing*, *concerning*, *interesting*. Use *healthy*, *drifting*, *broken*, *headroom*, *capped*.

The full report is for the operator. If the user wants a client-facing version, hand off to a client-comm skill — this one stays operational.

## Boundaries

- Don't approve or apply changes to the account. Recommend, don't execute.
- Don't recommend a bid-strategy move without invoking `actcenter_tcpa-scale` or `actcenter_tcpa-migration` first.
- If the account is in RISK state, stop and call out the root-cause investigation — don't paper over with bid moves.
- Maintain a 14-day cooldown between tCPA adjustments. Flag if the user is about to violate it.