Anatomy of /daily-check

/daily-check is the one mureo command most operators run first. It is also the one that best demonstrates the value of an orchestration framework over a collection of platform-specific tools. This post walks through what actually happens between an operator typing the command and seeing a report — and what the manual equivalent would cost in time and fidelity.

The manual daily process

Here is a reasonable version of what a technical marketing operator running four ad platforms has to do every weekday morning:

1. Log in to Google Ads. Check campaign delivery status (any disapprovals? budget caps hit? learning phase?), compare yesterday’s spend to target, note any CPA drift >20%. Export a change history report for the last 24 hours. ~10 minutes.
2. Log in to Meta Ads Manager. Same checks. Meta’s reporting has different attribution windows, different delivery classifications. ~10 minutes.
3. Log in to Google Search Console. Pull the top 20 queries by clicks for the last 7 days, compare to the previous 7. Check whether any previously-top queries have dropped off. ~8 minutes.
4. Log in to Google Analytics 4. Pull session-to-conversion for the organic-paid split, compare to paid-only attribution from step 1, reconcile (they will not match). ~10 minutes.
5. Cross-reference: did the paid non-brand spend yesterday correlate with organic ranking changes? Did a Meta audience exhaustion trigger a Google Ads learning phase spike? ~15 minutes if done well; usually skipped.
6. Write a short status update for the internal team or client. ~5 minutes.

Total, for a competent operator: 50-60 minutes, and step 5 — the cross-platform correlation — is the one step that meaningfully distinguishes good ad operations from tool-operating.

It is also the step most often skipped, because it requires holding four browser tabs open and a mental model of attribution-lag offsets that no single platform surfaces.

What /daily-check actually does

When an operator types /daily-check in Claude Code (or Cursor, Codex, or Gemini with mureo configured), the workflow runs roughly this sequence. All of it is inspectable in the OSS repo; nothing is hidden:

1. Strategy bind

Read STRATEGY.md at the repository root. Extract the current Goals weighting, any active Market context entries (e.g. known seasonal pulse, known competitor action), and the Operation mode. These become the frame for every downstream decision.

2. Parallel fetch across four surfaces

Issue platform queries in parallel:

• Google Ads — last 24h delivery, spend, conversions, CPA by campaign; learning-phase status; any disapproved ads; change history.
• Meta Ads — same shape, with Meta-specific fields for audience exhaustion, creative fatigue score, auction competitiveness.
• Search Console — last 7-day impression/click/position deltas for the top 50 queries.
• GA4 (via MCP) — sessions-to-conversion, attribution split paid-vs-organic, anomaly flags on key events.

Each platform call runs through the validation layer (GAQL for Google Ads; typed request builders for Meta and GA4). Nothing the agent fetches is raw-string injected; nothing leaks the credentials that authenticated the request.

3. Anomaly detection

The combined data set feeds the anomaly detector. Sample-size gates suppress single-day noise. Severity tiers (HIGH / CRITICAL) pick out what actually warrants attention.

4. Cross-platform correlation

The non-trivial step. The workflow cross-references:

• Paid non-brand spend vs organic ranking. If a paid campaign on a keyword cluster reduced spend yesterday while the organic position for the same cluster dropped today, that is a brand-visibility gap and gets flagged.
• Meta creative fatigue vs Google Ads volume. When Meta’s audience shows fatigue signals and Google Ads volume spikes, the likely explanation is demand being displaced to search — not a new demand signal. The workflow will suggest monitoring, not scaling.
• GA4 engagement anomalies vs platform-attributed conversions. A conversion count that looks normal in Google Ads but shows abnormal bounce / time-on-page in GA4 is a conversion-quality problem, not a conversion-quantity one. The workflow recommends investigating attribution before adjusting bids.

No single platform surfaces any of these correlations. Each is specific to ad ops as a discipline; each is in the diagnostic knowledge base that /daily-check consults.

5. Strategy-gated recommendations

For each flagged anomaly, the workflow checks the strategy. A CPA spike on a campaign tagged as brand defense is weighted differently from a CPA spike on a campaign tagged acquisition. The output is not “here are 14 things that changed” — it is “here are 2 things the strategy says you should act on; here are 5 things to monitor; here is 1 thing the data is too thin to call either way.”

6. Report

A single text output to the agent’s chat. Optionally writes a daily-check-YYYYMMDD.md to the account’s state directory for team sharing.

Why this is different from a dashboard

A SaaS dashboard can, in principle, implement most of the above. Two reasons it generally does not:

• Dashboards optimize for displaying data. They do not optimize for refusing to display low-signal data. The sample-size gate — the thing that makes /daily-check actually signal rather than noise — is fundamentally incompatible with the “more widgets good” dashboard paradigm.
• Dashboards do not know the strategy. They aggregate what the platform APIs return. Strategy is operator knowledge that lives outside the platform schema. A dashboard can add a “goals” section that the human has to read; mureo binds the strategy into the execution path of every decision.

/daily-check is not a better dashboard. It is a replacement for the morning process that produced the dashboard.

Time comparison

Step	Manual	/daily-check
Strategy bind	0 (operator carries it in their head)	Automatic
Platform fetch × 4	~40 minutes	~8 seconds, parallel
Anomaly detection	Visual scan; misses	Explicit
Cross-platform correlation	~15 min if done; usually not	Automatic
Strategy-gated filter	Mental model	Explicit
Written report	~5 minutes	Automatic

Roughly a 50–60 minute process compressed into a single command. The value is not the time saved — it is that the cross-platform correlation step, the one that distinguishes competent ad ops from tool-operating, actually gets run every day instead of once a week when the operator has time.

That is the premise of the whole mureo framework: give a disciplined ad operator the workflow coverage that, manually, only the best of them run consistently.