The short answer
Competitor ad research works best as a fixed 30-minute weekly slot: 5 minutes to scan for changes, 15 to log and tag what survived, 10 to write one brief. Weekly is the right speed because winners surface slowly: on our board, most ads only become clearly visible as winners between week 7 and week 10.
Why weekly, and why 30 minutes
Research that happens "when there is time" doesn't happen. Research that takes a whole afternoon gets skipped the first busy week. A fixed half hour on the same day every week survives both.
Weekly is also the right speed for the data. On the Ad Radar board (2,443 active Meta ads with $50k+ in estimated spend, October 2026), here is how many ads sit in each weekly age band:
| Days live | Ads on the board |
|---|---|
| 30 to 36 | 10 |
| 37 to 43 | 24 |
| 44 to 50 | 49 |
| 51 to 57 | 82 |
| 58 to 64 | 70 |
| 65 to 71 | 144 |
| 72 to 78 | 120 |
Very few ads clear $50k in estimated spend in their first five weeks. The weekly bands fill up from about week 7 onward. That means a competitor's winner usually becomes recognizable around two months in, and you have weeks, not hours, to notice it. Checking daily adds noise. Checking monthly means you see winners late. Weekly is the balance.
Before week one: the setup session
The first session is not 30 minutes. Block two hours once and do three things:
- Pick the watchlist. Five to ten brands: direct competitors, adjacent brands selling to the same buyer, and one or two volume leaders in your category. The legit way to spy on competitor ads covers how to choose them.
- Log every active ad once. This is your baseline. Without it, you can't tell what changed.
- Find their other Pages. Search each brand and product name as a keyword to catch creator and partner Pages. Add those Pages to the watchlist under the brand.
The 30-minute agenda
Minutes 0 to 5: scan for changes
Open each competitor's Page in the Meta Ad Library, active ads only. You are only looking for three kinds of change since last week:
- Survivors: ads that crossed 30 days live.
- Scaled ads: ads whose version count went up, or the same creative showing up on new Library IDs.
- Removals: ads that were in last week's log and are gone. A long-running ad that stops is news too.
Mark them, don't analyze yet.
Minutes 5 to 20: log and tag what changed
For each marked ad, add a row to the sheet: start date, versions, format, the first sentence verbatim, landing page type. Then three tags: hook pattern, angle and awareness level. Our swipe file guide has the fixed vocabulary, so your tags stay consistent week to week.
Click through to the landing page of any new survivor. Screenshot the offer above the fold. If the page type changed (product page to quiz, for example), note it in capitals. Funnel changes predict the next round of ads.
Minutes 20 to 30: write one action
Look at the rows you added and answer one question: what is the one thing worth testing because of this? Write it as a brief, a hook to test, or a page change. One action per week is 50 a year, which is more than most teams ship from research.
The template
Copy this into a doc and fill it every week. The whole thing should fit on one screen.
| Section | What goes here |
|---|---|
| Week of | Date |
| Watchlist checked | Brands, with a check mark |
| New survivors (30+ days) | Brand, hook verbatim, tags, landing type |
| Scaled ads (versions up) | Brand, old and new version count, what changes between versions |
| Stopped ads | Brand, days it ran, your guess why |
| Pattern of the week | One sentence across the rows above |
| Action | One brief, test or page change, with an owner |
A worked example: the cleaning niche
To show what a filled report looks like, we ran the process on the cleaning niche of our board: 151 winning ads from 62 brands. Two brands hold a quarter of them, Earth Breeze with 22 ads and Uproot Clean with 17. Nearly every ad from both is video. The niche median is 113 days live.
New survivors. One Uproot Clean ad is 63 days old and already estimated in the hundreds of thousands. It opens on a warning aimed at one specific buyer:
If you are a pet parent whose towels still come out smelling like wet dog, here are the three mistakes I see pet owners make every single week.
- Est. spend
- $364K
- Days live
- 63
- Format
- Video, 142s
A callout to one buyer (pet parents), one sensory problem (wet-dog towels) and a number (three mistakes). It sells a washer cleaner without naming it in the first line.
Scaled ads. Another Uproot Clean ad, live 127 days, runs three versions. Same brand, different door into the same problem: a family story instead of a warning.
My grandpa used to fix washing machines for a living.
- Est. spend
- $467K
- Days live
- 127
- Format
- Video, 414s
- Variants
- 3
A seven-minute story that earns authority through a relative's job. Three versions means the brand is betting on this one now.
Controls. Earth Breeze's oldest active ad has run 483 days and is pure offer, delivered in a creator's voice:
I basically got all of this for free from EarthBreeze because their four-pack is 40% off right now, plus a bunch of free gifts
- Est. spend
- $585K
- Days live
- 483
- Format
- Video, 26s
Twenty-six seconds, offer first, gifts listed on camera. Sixteen months live says the offer is the control, and new hooks get tested around it.
Pattern of the week. The newer winners in this niche open on a specific person with a specific smell or mess (pet parents, wet towels). The old control opens on the deal.
Action. Brief two hooks for our own product that call out one buyer and one sensory problem, and keep the offer close to the opener for the bottom of the funnel.
That is the whole report. It took one pass through two Pages and three ads.
Common ways the process breaks
- The watchlist grows to 30 brands. Cap it at ten. Rotate one out when you add one.
- You log everything, not changes. After the first session, only changes go in.
- Nobody reads the report. Put the action at the top, assign an owner, and review last week's action before starting this week's.
- Video hooks go unrecorded. The library can't search speech, so transcribe the first sentence every time. Our search tricks guide explains why.
- Old controls get ignored. New ads are exciting. The ad that has run for a year is the one the competitor trusts most.
Making it faster
The slowest parts are noticing changes and transcribing hooks. Both can be automated. Automating competitor ad monitoring covers the options, from alerts to scheduled Claude prompts. In Ad Radar you can follow a brand by pasting its Ad Library link, get the Today briefing of what changed, and see each ad already tagged and transcribed, which turns minutes 0 to 20 into about five. The 10 minutes of thinking at the end stay yours. If you need the signals for judging what counts as a winner, start with how to find winning Facebook ads.
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.