How to Run a Weekly Creative Review

The format in one table

Most creative reviews fail in one of two ways: they become a slideshow of pretty ads, or they become a media buying meeting where nobody talks about creative. A fixed agenda with fixed inputs prevents both.

Sixty minutes. The strategist runs it. The media buyer and the production lead attend. Everyone gets the inputs the day before.

Inputs to prepare (before the meeting)

1. Ad-level results, one view

Pull every ad that spent in the last 7 days into a single table, sorted by spend: spend, CPA or ROAS against target, hook rate, hold rate, CTR, frequency, days since launch. Add the account's 90-day median for each metric as the top row so outliers are obvious. How to read these together is covered in creative metrics that matter.

2. The test log

Every ad launched last week should already have a written test question, the dimension it changed (angle, awareness, hook pattern, format) and a kill rule. The review applies those rules. It does not invent new ones after seeing the numbers.

3. A market scan

This is the input most teams skip, and the one that keeps the slate from becoming a loop of your own ideas. The question: what started winning in your market recently?

To show what a scan can surface, we compared the newest ads on the Ad Radar board with the rest. On October 9, 2026, 172 ads from 125 brands had cleared $50k in estimated spend in under 60 days. Here is how they differ from the 2,271 older winners:

Shares are among ads carrying each tag. The new group is 172 ads, so treat single-point differences as noise and look at the big ones. Two stand out: fresh winners are more often video, and more often open with a first-person story. This data can't say why the image share is lower, so don't conclude statics are dying. But if your slate is all statics and number openers, the scan is telling you something.

A scan like this takes 10 minutes when the data is already filtered. Ad Radar's Today briefing and winner alerts surface new winners in your niche, and you can follow specific competitors by pasting their Meta Ad Library link. Without a tool, sort the Ad Library by your top five competitors and note any ad that's new and already has several versions.

Block by block

Block 1: last week's tests (15 min)

Go ad by ad, fastest. For each new ad: did it hit the kill rule, the winner rule, or neither? Say the decision out loud and write it. Don't discuss ads that are still inside their read window. They get discussed next week.

The useful question is not "did it win" but "what did it test, and what did we learn?" If a number opener beat a story opener on the same body, write that down. That sentence is worth more than the ad.

Block 2: current winners (10 min)

For each ad in the scaling campaign, look at a 2-week trend of CPA, CTR and frequency. Three possible calls:

• Feed it: stable, so brief 2 to 4 new openers or cuts this week.
• Watch it: wobbling, check again next week.
• Replace its slot: frequency up and CTR down for two weeks. Keep it live while CPA holds, but prioritize a new concept for that audience.

See how to scale a winning ad for what feeding a winner looks like.

Block 3: market scan (10 min)

Show three to five new winners from your niche. For each, name the angle, the awareness level and the opener pattern in one line. Then ask: is any of these a cell we haven't tested?

Never brief a copy. Brief the structure: "first-person story about a person you met, unaware reader, product revealed at second 20."

Block 4: matrix check (5 min)

Open the creative diversity matrix and mark last week's launches in their cells. Name the one empty cell you will test this week.

Block 5: decide the slate (15 min)

Write the slate as a list while everyone watches. Each line has a type, a parent (for iterations), the dimension being tested, an owner and a due date. Leave the meeting with every brief named, even if the brief itself is written later that day.

Block 6: log (5 min)

Three bullets in the learning doc: what we now believe that we didn't last week. Over a quarter this becomes your account's playbook.

The agenda, ready to paste

WEEKLY CREATIVE REVIEW              Date: ______  Facilitator: ______
Attendees: strategist, media buyer, production lead (max 5)
Pre-read (sent day before): results table, test log, market scan, matrix

1. LAST WEEK'S TESTS (15)
   Ad | Test question | Dimension | Rule hit? | Decision | Learning
   ...
2. CURRENT WINNERS (10)
   Ad | 2-wk CPA trend | Freq | CTR trend | Feed / Watch / Replace slot
3. MARKET SCAN (10)
   Ad | Brand | Angle | Awareness | Opener pattern | Cell we lack? Y/N
4. MATRIX GAP THIS WEEK (5)
   Cell: ____________  Reason: ____________
5. SLATE (15)
   # | Type (iteration/concept) | Parent | Dimension tested | Owner | Due
   1 |                          |        |                  |       |
   2 |                          |        |                  |       |
6. LOG (5)
   - We now believe: ___________________
   - We now believe: ___________________
   - Open question for next week: ______

Common ways the meeting goes wrong

• Debating taste. "I don't like the colors" is not a review comment. Point to a metric or a test rule, or let it go.
• Changing the rules after the fact. If the kill rule said 2x CPA and the ad hit 2x CPA with no sales, it's dead. Discuss the rule next quarter, not the ad today.
• No market input. A team that only looks at its own ads runs out of ideas within months. Ten minutes of scanning is the cheapest fix.
• No owner on the slate. A brief without a name and a date doesn't get made.

For the full system this meeting sits inside, see the creative testing framework.

Figures marked as estimated spend come from Ad Radar's model of engagement on public Meta Ad Library ads. They are estimates, labeled as such, and are best used to rank ads against each other.