What Did Your Marketing Operator Actually Change This Week?
If you cannot tell what your marketing partner changed this week, ask for a decision log. It should show the evidence behind the decision, the person responsible, the work that shipped and the next review point. A dashboard shows movement. The log explains what someone decided to do about it.
I am Timofei Tskhovrebov. My agency-replacement offer is direct operating ownership: media buying, AI creative, landing pages and tracking within an agreed scope. That makes accountability concrete. You can speak to the person making the account decisions and ask to see the work behind them.
Here is the weekly review template I would put in front of a founder. It separates implementation from business results, so you can assess the work while a test or customer cohort still needs time. Use it to evaluate an independent operator or the paid-social agency you already pay.
Ask for decisions, not a list of activity
“Optimized campaigns” does not tell you much. Which campaign changed? What evidence justified it? Was the change actually implemented? What would make the operator reverse it?
Equally, leaving a budget alone can be a deliberate decision. If recent trial users have not had enough time to become paying customers, a lower trial cost is not enough to justify more spend. The log should explain why the budget stayed put and when that decision will be reviewed. An empty change history and a documented hold are different things.
I would keep the review centered on decisions that affect spend, the offer, the conversion path or the reliability of the evidence. You do not need an entry for every file export. You do need a record of the choices that determine where your money goes.
Copy this seven-field marketing decision log
Use one entry per meaningful decision. A shared document or spreadsheet is enough. Keep links to the actual campaign record, creative version, page or test evidence beside the explanation.
| Field | What to record | What you should be able to check |
|---|---|---|
| 1. Decision and date | A short ID, date and the choice being made. | Which entry to return to next week. |
| 2. Evidence and window | The source, metric definition and observation dates. | Whether the evidence is comparable and old enough for the question. |
| 3. Reason | Why this action follows from the evidence; what remains uncertain. | The reasoning behind the move. |
| 4. Owner and dependency | Who implements it, who approves it and what input is missing. | Who can move the work forward. |
| 5. Change and proof | The exact change, version and link to the finished work. | Whether it shipped or is still a proposal. |
| 6. Next review | The expected signal, review date or agreed evidence threshold. | When to keep, change or stop the action. |
| 7. Outcome and status | What was observed; what is open, completed or reversed. | Whether the result supports the next decision. |
Keep two statuses when necessary: implementation complete and outcome unresolved. A page can be live and checked while its commercial effect is still unknown. Combining those statuses into “done” hides a useful distinction.
For creative, link the agreed brief, selected asset versions and launch record. The number of generated variations does not establish which buying idea was tested. My brief-to-campaign workflow explains that production sequence; this log connects it to the subsequent decision.
Three hypothetical entries: hold, ship and investigate
The following examples are fictional. They show how to fill the template, rather than describe a client account or promise a reporting cadence.
EX-01: Hold the media budget while the cohort matures
Evidence: a recent group of trial users is still inside the conversion window agreed for this business. The trial cost looks lower, but the eventual paid-customer result is not yet available.
Decision: keep the current authorized budget. The operator retains a dated budget snapshot and the cohort report. The business owner confirms the customer definition and the observation window. Record the actual next review date in the entry.
Status: the hold is implemented; the acquisition outcome is pending. At review, use the same cohort and agreed cost definition. If the evidence still cannot answer the question, record the remaining gap and choose the next action explicitly. “Needs more time” should not renew itself indefinitely.
EX-02: Ship a landing-page clarification
Evidence: a review finds that the page describes the offer differently from the ad. The business owner confirms which description is accurate.
Decision: the operator publishes the approved clarification. The entry links the URL, previous and new copy, page version, and device and form checks. If a developer owns the release, name that dependency instead of marking an unshipped edit complete.
Status: implementation is complete after the correct page and working conversion path are verified. Commercial impact remains unresolved. A shipped clarification is not evidence of a conversion-rate increase, and simultaneous changes can make later movement hard to attribute.
EX-03: Investigate a missing tracking event
Evidence: an agreed test transaction is missing an expected event from the verification record. Preserve the test reference and the event definition without putting customer details into a broadly shared report.
Decision: the operator documents the gap and works with the integration owner to investigate it. Access and the site's consent configuration may be dependencies. Attach the repair record and retest evidence as the work progresses.
Status: open until the agreed test and reconciliation pass. A ticket marked closed does not establish a working measurement setup. Record any remaining limitation; repairing a test event does not make every advertising and backend report identical.
Keep the decisions you reverse
Suppose the page change in EX-02 introduces a reproducible form defect. Create a reversal entry linked to EX-02. Restore the verified version within the agreed authority, record the failed check and successful retest, and reopen the original change.
Do not erase the first entry. You want to see what was believed, what changed and why the response made sense. In this hypothetical example, the reason for reversal is a verified form defect. It is not a guess that a small conversion-rate movement proves the copy failed.
The same discipline applies when an ad test disappoints. Retain the hypothesis and evidence, then state what the result does and does not support. The point is to improve the next decision, not defend the previous one.
Review the work alongside the business result
A complete log can establish delivery and reasoning. It cannot substitute for acquisition economics. Keep the agreed business measures beside it, with their sources and periods intact.
In my anonymous AI SaaS account case, December spend was $17,852.93 at $139.48 per purchase. May spend reached $148,907.37 at $131.08. Those are actual account results with an original export. They show the scale and purchase CPA I managed; the fictional entries above illustrate a review tool.
For your review, I would ask four questions: what changed, what evidence is ready, what decision needs your input, and what happens next? If an item is blocked, leave it visible with a named owner. If an assumption failed, update the plan. If results miss the commercial target, use the log to inspect the decisions rather than treating a long activity list as success.
Direct ownership should make the work easier to inspect
Replacing a paid-social agency with one senior operator only helps if responsibility becomes clearer. Media, creative, pages and tracking must connect, while specialist work and your team's approvals remain explicit. My agency-replacement responsibility map explains that scope. If you are still evaluating candidates, start with the operator work sample.
My full-stack growth operation costs 5% of monthly ad spend or a $3,000 monthly minimum, whichever is greater. Media spend is separate. We agree the work, access, dependencies and review expectations on the call; the separate AI Systems Build is not automatically included.
If you want the person making the decisions directly on the account, bring the account to me. Show me the current report and the work waiting behind it. I will explain what I can own and which decisions need better evidence.