How to Audit Your Media Buyer in 30 Minutes (Without Asking Them)

You Hired a Media Buyer. Did You?

Most founders I talk to genuinely have no idea what their media buyer does day-to-day. The reports come in. The numbers move. Someone says "we increased budget on the winners and paused the losers" on the Friday call. The founder nods. The cycle continues.

This works fine until something breaks. Then the founder discovers their account has been on cruise control for six months and the buyer has been making four budget changes a week, all on Mondays, all 50% in either direction, while charging $12K/month for the privilege.

You don't need to be a media buyer to audit one. You need 30 minutes and the change log.

Why the Change Log Is the Truth

Every action in a Meta ad account leaves a record. Budget changes, creative pauses, audience edits, bid strategy switches — all of it logged with a timestamp and a username.

The change log is the closest thing to a polygraph in performance marketing. Reports can be cherry-picked. Slack updates can be polished. The change log is what actually happened.

When we onboard a new account at Impremis, the first thing we pull is 90 days of change history. The shape of that history tells us more about the previous team's competence than any case study they ever pitched.

How to Pull the Data

This takes about three minutes:

1. Open Meta Ads Manager.
2. Click the clock icon in the top-right of the campaign view (it says "See History" on hover).
3. Set the date range to the last 90 days.
4. Filter by "Edits" to remove auto-system events.
5. Click Export and save the CSV.

What you'll get is a row-by-row record of every human action in the account. Now you can read it like a book.

Four Charts That Tell You Everything

Drop the CSV into a spreadsheet and build these four views. The first one takes 10 minutes; each additional one takes about 5.

Chart 1: Distribution of Budget Change Sizes

Bin every budget change by % size in 5% increments. Plot the histogram.

A healthy buyer shows a fat curve in the 5-25% range, a thin tail at 25-40%, and almost nothing above 50%. Most decisions are nudges. A few decisions are surgical scales or cuts.

An unhealthy buyer shows two giant spikes at "+50%" and "-50%" and almost nothing in between. They're not optimizing. They're guessing, and they only have two guesses.

Chart 2: Changes by Day of Week

Group all edits by weekday. Plot a bar chart.

A healthy account has changes spread Monday through Friday with a slight Tuesday-Wednesday concentration. Weekend edits are rare but exist (when something is on fire, the buyer touches it).

An unhealthy account shows 70-90% of edits clustered on a single day. That's a buyer who logs in once a week, panic-optimizes, and ignores the account in between. We saw this on a $300K/month brand we audited last year — 84% of changes happened on Mondays. Their CAC drift on Wednesday was costing them more per week than the agency cost per month.

Chart 3: Change Velocity vs. Performance

For each week, plot two numbers: number of edits made, and that week's blended CPA. Look at the relationship.

A healthy buyer shows an inverse relationship — more edits in weeks where performance was off, fewer edits when the account was humming. Activity tracks need.

An unhealthy buyer shows a flat line — same number of edits regardless of what the account is doing. Or worse, a positive correlation, meaning they edit more when things are working, breaking what was working.

Chart 4: Predictive Power

For every meaningful change (budget +/- 20% or pause/restart), look at the next-day blended CPA. Compare to the trailing 7-day average.

A healthy buyer shows next-day CPA improving more often than not. They're calling shots that play out.

An unhealthy buyer shows next-day CPA degrading or unchanged. They're moving levers without understanding what the levers do. This is the single most damning signal in an audit.

Red Flags by Severity

| Signal | What it means | Severity | |---|---|---| | 80%+ of changes on one weekday | Account managed in batches, not in real time | High | | Repeated 50%+ budget swings | Buyer doesn't trust their own data | High | | Pause/restart on the same ad within 48 hours | Reactive, not strategic | Medium | | All changes by one user, never a team | No backup, no peer review | Medium | | Change velocity unchanged across performance swings | Going through motions | High | | Next-day CPA worsens after most changes | Buyer cannot predict outcomes | Highest | | Total edit volume below 30/week on a $250K+ account | Account on autopilot | High |

None of these alone is fatal. Three or more in combination is a problem you need to solve immediately.

What Good Looks Like

A strong media buyer running a six-figure-per-month account looks like this in the change log:

• 50-150 edits per week, varying with account state
• Most budget changes between 5% and 25%
• Edits distributed across at least 4 of 5 weekdays
• Larger changes (40%+) reserved for clear outliers, with a visible ROI follow-through
• Naming conventions consistent across ad sets and campaigns
• Pause decisions sticky — they don't unpause an ad they killed two days ago
• One or two trusted team members making changes, with clean handoffs

This is what active management looks like. It is not glamorous. It is not what most agency pitch decks describe. It is what works.

What I Look for in 90 Seconds

When I do a fast read on an account, before any chart-building, I scan three things:

1. Total edits in the last 7 days. Below 20 on a >$100K/month account is a yellow flag, below 10 is red.
2. Most recent edit timestamp. If the last edit was three days ago and ROAS is moving, the buyer isn't watching.
3. Distribution of "who." If it's all one username and that person took a vacation last week, the account ran itself for seven days. That's not management. That's hosting.

These aren't comprehensive. They are diagnostic. They tell you whether to spend the next 30 minutes on the deeper audit.

The Five-Question Conversation

Once you've built the charts, set up a 20-minute call with the buyer. Don't accuse. Ask:

1. Walk me through the three biggest budget changes you made last month and what you expected to happen.
2. What's the largest single change you'd be comfortable making, and when?
3. How do you decide when to pause vs. when to wait?
4. When was the last time you predicted a change wouldn't work but ran it anyway?
5. Show me one ad set you're worried about right now and tell me why.

A real operator will answer these in specifics. A bad one will speak in generalities. The gap between the two answers, in five questions, will tell you everything.

For the broader operator framework around metrics that matter, see the metrics that matter more than ROAS. For why ROAS is a dangerous summary statistic, see why ROAS isn't the goal.

FAQ

Do I need to be technical to do this audit?

No. The change log exports to a spreadsheet. The charts are pivot tables. If you can run a basic SUM-IF, you can do this in an afternoon.

What if my buyer says they manage performance "strategically" with fewer edits?

Fewer edits is fine if the account is performing. The Chart 3 test settles it — change velocity should track account need. A truly stable, scaling account can run on lower edit volume. A drifting account on low edit volume is being neglected and called "strategic."

How often should I run this audit?

Quarterly at minimum. Monthly if you've recently switched buyers or if performance has shifted in either direction. The change log doesn't lie, and trends only show up over time.

Can my buyer hide changes from me?

No. Every action in an ad account is logged at the platform level by user. They can leave the account, but they can't erase the trail. If your buyer pushes back on you pulling change history, that itself is the audit answer.

What if my buyer is internal, not an agency?

Same audit. The questions are the same. Internal hires get the same scrutiny external agencies do, often more, because there's no contract escape hatch — only an HR conversation.

What about automated rules and Advantage+?

Filter system events out for the human-action audit. Then run a parallel audit on the rules themselves. "Set it and forget it" rule logic is the most expensive form of laziness in modern Meta accounts.

How does this relate to AdFuse?

AdFuse exists because doing this audit by hand for 100+ accounts a month is impossible. The platform surfaces these patterns automatically — buyer cadence, change distribution, predictive accuracy — without anyone having to build pivot tables. But the principles are the same whether you do it manually or with software.

The Real Lesson

The goal of the audit isn't to fire your buyer. It's to know whether you have one.

A real operator will welcome the audit because it surfaces what they're already doing. A weak operator will dodge it because it surfaces what they're not.

You paid for active management. The change log tells you whether you got it. Read it. Trust what you see. Adjust accordingly.

Keep reading

Pieces I've written on related topics that pair well with this one:

• How to Audit a Media Buyer's Work Without Micromanaging Their Account — Most agency audits check ROAS and miss everything that matters.
• Media Buyer vs. Growth Marketer: The Hiring Decision That Defines Your Next Stage — Hiring the wrong role at the wrong stage costs more than the salary. Here's the stage-based framework for when to hire a media buyer vs.
• What I Got Wrong About Hiring Media Buyers — Most media buyer hires fail because agencies optimize for platform skills instead of judgment, communication, and commercial thinking.
• How to Build a Paid Media Playbook That a New Hire Can Execute Without You — Build a paid media playbook your team can run without you — KPI systems, attribution across platforms, and creative workflows that actually scale.
• The Funnel Audit: How to Find Where Paid Traffic Is Leaking Before You Scale — Scaling into a leaking funnel just buys more waste. Here's the four-stage audit framework that diagnoses where paid traffic is lost before you add bud…