The Weekly Optimization Loop

An optimizer running fifteen accounts loses ten to fifteen hours a week to report-pulling before any thinking begins. This agent works eight fixed stations across every account and delivers one thing: an approval queue your specialist clears over coffee. The human judges; the agent looks.

Get this skill — free in the pack MageCloud growth services

Year-ahead planning with Roger from A&E Leisure

What it takes off your team’s plate

The Monday grind, station by station:

Health sweepsDisapprovals, feed rejections, billing alerts, and tracking anomalies — the blockers that outrank everything.

Pacing mathMonth-to-date versus plan per campaign, with the projected month-end number.

Search-term triageThis week’s waste, grouped into negative list proposals with the savings attached.

Anomaly explanationCPC spikes, CTR collapses, geo weirdness — each with a cause hypothesis, not just a red arrow.

How a run works

Eight stations, in order, every week:

Health sweep first. Broken tracking halts the run — no optimizing on corrupted data.
Pacing and scale candidates. Deviation over 10% flagged; capped winners listed with impression share lost to budget.
Search terms and negatives. Waste proposed into shared lists; converting queries flagged for exact-match expansion.
Products and assets. Thirty-day zero-converters, week-over-week movers with likely causes, PMax assets rated Low for two-plus weeks.
Bids, budgets, hygiene, anomalies. Target sanity with 30-conversion evidence minimums, reallocation math shown, expired promos and price mismatches caught, and one paragraph per anomaly with a hypothesis.

✋ The approval gate

Caps and boundaries, non-negotiable:

No live change without approval of that specific item — approving #3 does not approve #4
Bid and target moves capped at ±15% per week; budget shifts at 20%
No campaign paused without sign-off, regardless of performance
Negatives removal, attribution settings, and conversion actions are human-only surfaces

What you get

WEEKLY-[client]-[date].md — three sections: FYI, approval queue, strategic flags

Each item: metric + analysis + risk + APPROVE / EDIT / SKIP

Strategic flags separated from mechanical approvals

A run log of time spent and approval outcomes — how the loop learns

At a Google workshop on using AI in marketing

Roger at A&E Leisure put it best: with every agency before us it was a pound in, fifty pence out — with us the ratio flipped. Ratios flip through relentless weekly hygiene, not heroics. The loop is that hygiene, made cheap enough to never skip.

I have written about this from the trenches:

Works together with

Run it yourself, or bring my team in

The skill is free in the pack — install it and start this week. If your store does seven figures and you want this running inside a full revenue-optimization engagement, my team at MageCloud handles it end to end.

Download the pack Book a call