I currently run more than $80K a month in ad spend across Google, Meta, Microsoft and Amazon. Over 22 years the lifetime figure has passed $8M, spread across 75 completed Upwork contracts and my own agency’s clients.
None of that spend is managed by inspiration. It is managed by a weekly rhythm that has barely changed in five years, even as the tooling underneath it changed completely. This is what the rhythm looks like.
Monday belongs to measurement
Every week starts with the same unglamorous question: do I still believe the numbers? Conversion tracking breaks silently. A consent banner update, a checkout replatform, a tag manager publish by someone on the client’s side; any of these can cut the data feeding every bidding algorithm, and the platforms will keep spending confidently either way.
I have watched a purchase event die quietly on an e-commerce account after a site update while the campaigns kept optimizing toward a signal that no longer existed. Since then the rule is absolute: no budget increase on an account whose tracking I have not verified that week.
The measurement stack matters because the platforms reward it. Google’s own enhanced conversions documentation reports that advertisers implementing the feature see an average conversion rate improvement of 5% on Search and 17.1% on YouTube (Google, 2021). Server-side tagging, enhanced conversions, Meta’s Conversions API and a properly configured Consent Mode are not tactics anymore; they are table stakes. Feed the algorithms better data and they bid better. Starve them and you pay for their guessing.
Hygiene, or fighting for the other 64 cents
The ANA’s programmatic transparency study is still the most sobering document in this industry. After tracing $123 million of spend across 35 billion impressions, its conclusion was blunt.
Only about 36 cents of every dollar entering a demand-side platform effectively reaches a consumer (ANA, 2023).
That study covered open-web programmatic, not search or social, but the lesson generalizes: waste is the default state of ad spend, and hygiene is the discipline that fights it. On my accounts hygiene has a fixed weekly slot:
- Search term review on every Google and Microsoft search campaign. New negatives every single week, without exception. Stop for a month and match-type drift quietly funds queries you never chose.
- Placement and channel review on Performance Max and Demand Gen, using every exclusion lever Google exposes.
- Amazon search term reports mined in both directions: negatives out, new exact-match winners in.
- On Meta, frequency and fatigue checks, plus a hard look at where Advantage+ placements are actually spending.
None of this is clever. All of it compounds.
Creative gets half the calendar because it drives half the outcome
NCSolutions’ meta-analysis of roughly 450 campaigns found that creative quality drives 49% of incremental sales, more than targeting, reach and recency combined (NCSolutions, 2023). My accounts keep confirming the direction: when performance moves sharply, the cause is usually a creative change, not a bid change.
So creative runs as a pipeline, not a scramble. A standing batch of new angles is always in production, every account keeps a testing slot, and kill rules are agreed before anything launches, not argued after. AI image and video generation has collapsed the production cost of testing. It has not changed what wins: an honest offer stated plainly still beats a beautiful ad about nothing.
Budget rules, agents, and where judgment stays human
Budget reallocation follows written rules. Money moves toward marginal performance, not average performance; the best campaign on average is often past the point of diminishing returns while an unloved one has headroom. I move budgets on a fixed cadence, in measured steps, never mid-learning-phase, and never off a two-day window.
Since 2024 I have run an AI automation practice alongside the media buying, and agents now do real work inside this rhythm. They draft the weekly account summaries, pre-mine search term reports, flag tracking anomalies, watch pacing daily and catch disapprovals. Some of it runs on local open-weight models on my own hardware, which keeps client data off third-party clouds.
What agents do not do: move money, change bid strategies, or talk to clients. Every recommendation lands in a queue that a human clears. The failure mode of full automation is confident action on broken data, and as Monday’s ritual keeps proving, data breaks constantly.
This is also my answer to the fee question. Industry pricing guides put typical agency management fees at 15% to 30% of monthly ad spend (HawkSEM, 2026). At $80K a month that is real money, and what it should buy is the rhythm described above, executed every week by someone accountable for it. If a proposal cannot describe its weekly operating cadence in concrete terms, the fee is buying a dashboard login and a monthly PDF.
The week, compressed: Monday is measurement and the numbers I am willing to defend. Tuesday and Wednesday are hygiene and budget moves, executed against the written rules. Thursday is creative review and launching the next test batch. Friday is client communication, drafted by agents and rewritten by me, because the judgment about what matters this week is the part clients actually pay for.
Twenty-two years in, the honest summary is that managing serious ad spend looks less like genius and more like flight checklists: verify your instruments, remove waste, keep testing, follow the rules you wrote when you were calm. The machines keep getting better at executing the checklist. Deciding which rules belong on it is still the job.