A production schedule and a replacement schedule are different. Weekly deliveries do not require stopping every ad each week. Replace an active ad based on results, the offer and what its replacement is intended to test.
AI Vidia, a studio that makes AI video and image ads for DTC and consumer brands, delivers a new batch of ads every week on its monthly plans. To try one concept before a monthly plan, see the First Batch.
Separate new requirements from declining results
A sold-out product, expired offer or changed package may require correction even while an ad sells. Rising acquisition cost needs investigation before ordering replacements; landing-page and seasonal changes may matter as much as the video.
Give each active ad a status: continue, correct facts, investigate or test a new concept. File age alone should not determine the status. An older ad can remain a useful control.
What a faster refresh costs
Every new ad carries three costs: making it, the spend it needs before its results can be read, and your team's review time.
On Meta, an ad set's learning phase usually ends after about 50 results in the week after its last significant edit, and Meta lists adding a new ad to the ad set as a significant edit. Refresh faster than that and the ad set can keep relearning, so no file collects enough results to judge. What resets learning on Meta and TikTok lists the other edits.
A worked example with made-up inputs: at a cost per purchase of EUR 40, the 50 purchases that end learning take about EUR 2,000 of spend. That spend still buys purchases. Refresh before it is spent, and you pay for a new file without learning whether the last one worked. Production adds the per-video price; AI Vidia's monthly plans work out to roughly EUR 127 per finished video on Start and EUR 99 or less on Studio. Add your team's review hours at your own rate.
TikTok's Smart Creative shows the platform's own pace: it runs the most distinct of your initial videos for 3 to 5 days on no more than 10% of the ad group's budget, then pauses initial videos that show signs of fatigue.
Put your own numbers into the cost per usable ad calculator, check how old your live ads are with the Fatigue Clock, and see how many new ads a weekly cadence needs.
Build a decision calendar
Review results before the next production cycle. Name who can stop an inaccurate ad and who orders a replacement. Keep approved alternatives for products with frequent stock or offer changes. Before a seasonal peak such as Black Friday, plan production backwards from the go-live date, so new ads finish learning before the busiest days.
In an illustrative week, a review may lead to one price correction and one new demonstration. It need not trigger changes to five functioning ads. The calendar should organise responsibilities rather than impose a file quota.
A refresh decision in practice
Imagine acquisition cost rises for a previously stable bag ad. Check offer and stock first. If the popular colour sold out, a new opening alone is unlikely to solve the issue.
With product and checkout unchanged, inspect consistent impression, click and purchase trends. Locate the change before briefing production.
Choose a partial revision or a new explanation for a reason. A new opening keeps the comparison bounded; another concept tests a different question.
Keep the previous version, replacement date and simultaneous changes. Revisit the original hypothesis, allowing an inconclusive status rather than declaring every replacement successful.