Choose one home for it
Pick a single tool and stop there. A Notion database, an Airtable base or a Google Sheets file all work. I would pick the one your team already opens every day.
One database, one row per ad. Screenshots in folders and bookmarks in browsers do not count, because nobody can search them.
Decide what a winner is
You need a rule before you start saving ads, or everything looks like a winner. For your own ads, try this: at least 5,000 impressions and a cost per qualified lead at least a quarter below your account average.
Write the rule at the top of the file. If your volume is lower, use leads rather than impressions. For example, ads with at least ten leads and the best cost per lead.
Record the same fields every time
Make these columns:
- Screenshot or video link.
- Platform and format: LinkedIn Ads single image, Meta Ads reel, and so on.
- The first line of the copy, word for word.
- Angle: pain, outcome, proof, curiosity or offer.
- Audience and the offer shown.
- Dates it ran, spend and click-through rate.
- Cost per qualified lead, not only cost per lead.
- One sentence: why do you think it worked?
The last column is the one people skip, and it is the most useful. A guess you write down today can be proved wrong later, and that is how you learn.
Collect other people's ads too
Your own winners are a small sample. Add ads from companies selling to the same buyers. Meta, LinkedIn and Google all publish public libraries of the ads currently running on their platforms.
Save ads that have stayed live for months. Nobody keeps paying for a loser for long, so a long run is a hint that the ad works. Treat it as a hint, not proof. Record the date you saw it.
Tag by angle and hook, not by company
Add a tag for the angle and another for the hook type: a question, a number, a claim, a quote from a customer. Sorting by competitor tells you little. Sorting by angle shows you that, for example, proof-led ads beat curiosity-led ads for your audience.
Use it at the start of every brief
Before you write new ads, filter for your audience and format. Read five rows. Write down the pattern you see. Then write six variants, two for each of three angles, and send them into your next test.
Link the file to your testing routine in A/B test ad formats systematically and your monthly refresh in Run monthly creative refreshes.
Feed it on a schedule
Put a ten minute slot in the calendar every Friday. Add any ads that crossed your winner line this week, and any you saw that made you stop scrolling. Once a month, delete or tag as "tired" any entry whose performance has fallen for four weeks.
Common mistakes
- Saving ads without the numbers, so the file becomes a mood board.
- Copying a competitor's ad word for word. Use the structure, not the sentences.
- Only keeping winners from last month, and losing the older proof.
- Setting up a clever database nobody updates.
- Skipping the "why it worked" line.
How you know it works
A new brief takes you minutes to start rather than an hour. Your file holds at least twenty entries with numbers after two months. When someone asks "what angle worked for this audience?", you answer from the file and not from memory.
