On this pageJump to a section
Count your concepts, not your ads.
Open your ad account and write each live ad's idea in five words or fewer. "Free guide for busy founders." "Client result in 60 days." "Why most calendars stay empty." When you finish, circle the ideas that repeat. If you have fourteen ads and four circles, you do not have fourteen ads. You have four ideas wearing different clothes.
That is the problem a creative bank solves. A bank is a deliberate stock of distinct ideas, filled on a schedule, so you are never one tired ad away from a dry account.
What is a creative bank, and why build one now?
It is a library of ideas with a launch schedule attached.
The problem with the ad you made at midnight
Most accounts run on a single streak: a good ad appears, it works for a while, it fades, and then there is a scramble. The scramble produces the next ad, which is usually a variation on the last one because that is what you know. A bank breaks the cycle by filling the shelf before you need it.
Why the timing is right
On December 2, Meta's engineers described Andromeda as the first step in the ad pipeline (I wrote a full first look at what Andromeda means for small accounts). It takes tens of millions of ad candidates and selects a few thousand relevant ones, and the post says generative AI tools and Advantage+ are expected to grow the number of creatives in the system significantly.1 The post does not tell advertisers how many ads to run or how they are judged to differ. What I take from it, and I flag this as my reading, is that an account offering the system several truly different reasons to show it has more to work with than one offering a single reason several times.
What are the three axes of a concept?
Angle, format and hook. Keep them separate.
Angle: the reason to care
The angle is the answer to "why should this person stop and listen?" A few families that apply to most service businesses:
- Problem: name the pain in the buyer's own language.
- Result: show what changes, honestly and specifically.
- Process: explain how it works so the buyer can see the path.
- Objection: answer the doubt out loud, such as price, time or "I tried this before."
- Story: a real moment from the work, told plainly.
- Myth: correct something your market believes that costs them money.
Format: how it arrives
A person talking to camera. A screen recording. A text-led static. A short demonstration. A carousel. A client's words on a plain background. The same angle in a different format reaches different people, so format counts as variety too. It is not the whole of it.
Hook: the first moment
The first line, the first frame, the first three seconds. A hook is cheap to change and expensive to ignore. Write five for every angle. Most will be mediocre. One will not.
| Ten tweaks | Ten concepts |
|---|---|
| One angle, ten first lines | Five angles across three formats |
| Same promise in every ad | Different reasons for different people |
| Fades together when the idea tires | Gives you a bench when one tires |
| Feels like testing | Is testing |
How do you fill the bank?
From real material, not from imagination.
Mine what you already have
Your best raw material is already in your business. Read through your last twenty sales call notes and your message threads. Write down the exact phrases buyers use for their problem, the objections they raise and the questions they ask twice. Those phrases are hooks that nobody had to invent.
Use your own proof
Collect real client words, with permission. A plain screenshot of a real message from a real client often beats a polished graphic. Do not write testimonials yourself and do not attribute results you cannot document.
Get to thirty
- List six angles that fit your offer, using the families above.
- For each angle, write five hooks from real phrases you have heard.
- Pick two formats you can produce reliably this month.
- Combine them into thirty candidate concepts on a sheet.
- Cut to the ten you can make well in the next two weeks.
- Schedule them across the quarter, two at a time.
Dara Denney shared her rule for splitting a testing budget in a video posted on November 21, 2024 (sponsored by Storyblocks): "You should spend 90% of your tests on net new creatives and only 10% on iterations."3 That is a useful ratio for the bank. Most of what you add should be a new concept, not a tweak of last month's winner.
How do you run the bank week to week?
On a rhythm you can keep.
Weekly cadence
Pick a day. Launch one or two new concepts. Retire nothing in the first week unless it is clearly failing. Review results on a fixed day with the numbers that matter: cost per booked call, show rate and collected cash, not clicks. This is the weekly testing rhythm applied to a bank of ideas.
- Mine real materialSales calls, messages, client words
- Build conceptsAngle, format and hook kept separate
- Launch one or twoOn the same day every week
- Review on a fixed dayCost per booked call, show rate, collected cash
- Rest or recycleKeep the notes, reuse the winners' DNA
Protect the learning
According to Meta's Help Center, learning usually ends after roughly 50 optimization events in the seven days following a significant edit, edits made during learning restart it, and spreading budget over many ads and ad sets means the system learns less about each one. It recommends combining similar ad sets and using realistic budgets.2 So put new concepts into a simple structure with enough budget behind it, rather than giving each its own tiny ad set.
What to track in the bank
A single sheet is enough. Columns: concept ID, angle, format, hook, offer, launch date, status, cost per booked call, show rate and notes. Name each ad with the concept ID so results map back to the bank.
How do you retire and recycle?
Carefully, and with notes.
Retire on evidence
When a concept's cost per booked call rises and stays above your break-even for a full week, move it to "rested" and note what you saw. Do not delete it. A rested concept with a new hook or format can come back later.
Recycle the winners' DNA
Look at what your best performers share. It might be an angle, a type of hook or a format. Use that as a brief for the next batch, not as a reason to make ten copies.
You cannot test your way out of having only one idea. Put more ideas in the bank first.
In the accounts I have run, the ones that stay healthy treat creative like inventory: planned, stocked and rotated. Start your bank this week, even if it only holds ten ideas, so you are ready when January's cheaper auctions arrive. If you want a hand building it, reach out.
Keep building.Ekai
Questions people ask
What is a creative bank for Meta ads?
A creative bank is an organized library of ad concepts, each defined by an angle, a format and a hook, with status and results tracked. It lets you launch new ideas on a schedule instead of scrambling when an ad stops working.
What is the difference between an angle, a format and a hook?
The angle is the reason someone should care, such as a problem, a result or an objection. The format is how it is delivered, such as a person talking or a text-led image. The hook is the first line or moment that earns attention.
How many creative concepts should a small account run?
Meta has not published a number. A sensible starting point is to run a handful of genuinely different concepts at once rather than many near copies, and to add one or two new ones each week.
Does Meta's learning phase limit how much I can test?
It affects how you structure tests. Meta puts the learning threshold at roughly 50 optimization events in the week after a significant edit, and advises against high ad volumes and unnecessary edits. Concentrate budget instead of scattering it.
Sources
- 1
- 2
- 3
