Why weekly, and why now
Meta's Andromeda engine matches a winning ad to its best audience faster than the old system did, and exhausts that audience just as fast.
In 2023 a good ad could carry an account for two months. In 2026 the same ad finds its buyers in days and runs out of them in weeks. The fix is not changing ads more aggressively. It is having the next concepts already live before the current winner turns. Read the full story in Meta Andromeda.
Director by Sprites · private beta
Keep the pipeline full. Every Monday.
Director is a full creative team for your Meta ads. Every Monday you get finished, on-brand video ads, reviewed by a human. You pick the winners. Sprites ships the campaigns.
- Strategy, hooks, scripts, shooting and editing, end to end
- New concepts before the current winner fades
- No prompting, no endless revisions
When to swap a Meta ad
Swap when the trend turns, not on a calendar: rising frequency with falling click-through and a CPA that has climbed for a week.
- Frequency keeps climbing while reach flattens. The engine is showing the ad to the same people.
- Hook rate drops (3-second views divided by impressions). Fewer people stop for it.
- CPA rises for 5 to 7 days in a row against the ad's own best week, not against the account average.
- CPM rises with no change in competition. Meta charges more to reach people who care less.
Track all four per ad with creative analytics. One bad day is noise. A week is fatigue. More on the signs in creative fatigue.
What not to change
Do not edit a live winner. Add new ads next to it and let spend move on its own.
- Editing the creative, targeting or optimization event of a live ad set counts as a significant edit and can send it back into the learning phase.
- Pausing a winner the moment a new ad launches throws away proven spend. Let Advantage+ shift budget as the new one earns it.
- Swapping a headline on the same video is not a new ad to Andromeda. Ship a new concept instead.
Questions
How often should I change my Facebook ads in 2026?
Add new concepts every week and expect a winner to last 2 to 4 weeks at meaningful spend. Swap when frequency, hook rate and CPA turn for a full week, not on a fixed calendar.
How long should a Facebook ad run?
As long as it keeps hitting your CPA. At scale that is usually 2 to 4 weeks in 2026. On small budgets an ad can run for months because frequency builds slowly.
Does adding a new ad reset the learning phase?
Adding or editing creative in an ad set can count as a significant edit and restart learning. Batch new ads together once a week instead of trickling them in daily.
Should I refresh ad creative on a schedule?
Launch new concepts on a schedule, weekly. Retire old ones on their numbers. The schedule is for supply, the data is for removal.
Is it better to edit an ad or launch a new one?
Launch a new one. Editing a live winner risks resetting learning, and a small edit gives Andromeda nothing new to match.
Never run out of the next ad
Director delivers finished, human-reviewed video ads every Monday. You pick the winners. Sprites ships the campaigns.