# How Long Until UGC Ads Perform, and When Creative Fatigue Hits

> How long until UGC ads perform: the 7-day learning window, the budget needed to exit it, and a creative refresh cadence that fits Naira budgets.

You launched the creator videos on Monday. On Wednesday the dashboard looked flat, so the campaign was paused, and now everyone agrees the content "did not work."

That conclusion is usually wrong, and it is expensive. Paid social has a learning period, a lag in reported conversions, and a fatigue curve. Each one has a different clock, and judging all three on day three guarantees you learn nothing from the spend.

## What the learning window actually requires

Meta's own guidance says an ad set leaves the learning phase after about 50 optimisation events in a week, and that significant edits reset the count. An optimisation event is whatever you told the system to chase: a purchase, a lead, an add-to-cart.

Do the arithmetic before you launch. If your target cost per purchase is ₦8,000, then 50 events costs ₦400,000 in a week, or roughly ₦57,000 a day in one ad set. Run that same campaign at ₦10,000 a day and the ad set will sit in learning for a month, spending the whole time making unstable decisions.

| Target CPA | Weekly spend to reach 50 events | Daily budget in one ad set |
| :--- | :--- | :--- |
| ₦3,000 | ₦150,000 | About ₦21,500 |
| ₦5,000 | ₦250,000 | About ₦35,700 |
| ₦8,000 | ₦400,000 | About ₦57,000 |
| ₦15,000 | ₦750,000 | About ₦107,000 |
| ₦30,000 | ₦1,500,000 | About ₦214,000 |

This is why small Nigerian budgets struggle with Meta. A ₦200,000 monthly test cannot exit learning on a purchase objective, so the account reports "learning limited" and the media buyer blames the creative.

The honest responses are to consolidate budget into one ad set instead of five, to optimise for a cheaper event such as add-to-cart or a qualified lead while you gather signal, or to accept that the first month is a data purchase, not a profit centre.

## The first seven days are for delivery checks, not verdicts

Reported conversions arrive late. The default attribution window on Meta counts a purchase up to seven days after a click and one day after a view, so an ad that ran on Monday can still be credited with sales on Sunday.

Days one to three: check that the ads are delivering, that frequency is sane, and that the creative is being served in the placement you designed for. A 9:16 video running in a 1:1 placement with black bars is a setup error, and no amount of waiting will fix it.

Days four to seven: read thumbstop and hold rate. These are behavioural metrics with almost no lag, and they tell you whether the creative is being watched. If the hook is failing at this point, you already know.

Days eight to fourteen: read CTR, conversion rate and CPA, with enough events to mean something. This is the first honest profitability conversation.

Days fifteen to twenty-eight: scale what works, retire what does not, and plan the refresh. A creative that performs for four weeks in a Nigerian account is a good result, not a disappointing one.

## The signals that separate "not working" from "not enough data"

Kill an ad when the behaviour is bad across enough volume, and be specific about both.

A hook under 10% thumbstop after 10,000 impressions is a failed opening. A hold rate under 25% means the body is losing people before the product earns attention. A CTR under 0.5% with a strong hold rate means the video is watchable and the offer is not compelling. A CPA above twice your target after 30 optimisation events is a losing unit economic, not a slow start.

Do not kill an ad because CPA is high at 900 impressions. At that volume you are reading noise, and the replacement ad will produce its own noise.

One more trap: changing the creative, the audience and the budget in the same week. You will not know which change helped. Change one variable per cycle, even when that feels slow.

## Fatigue is a frequency problem first

Creative fatigue shows up as a pattern, and frequency usually moves before performance does.

Watch frequency on a seven-day window. For cold prospecting, a frequency above 2.5 to 3 means the same people are seeing the ad repeatedly, and the drop in response follows. Retargeting audiences tolerate higher frequency because the intent is already there.

The second signal is cost per thousand impressions rising while CTR holds. That is auction pressure or audience saturation, and a new hook usually fixes it faster than a new audience.

The third signal is CTR falling while CPM stays flat. That is genuine creative fatigue, and the fix is new material.

Nigerian account sizes fatigue faster than the global advice suggests, because audiences are narrower than the population implies. A campaign targeted to crypto-interested men in Lagos Island will burn through a creative in ten days. A broad Lagos campaign on a consumer product can run a month on the same asset. Small verticals are the worst case: ugc.ng lists just 4 creators in crypto and 3 in sports, which tells you how thin specialist audiences and specialist creators both are.

## A refresh cadence you can afford

Refresh in layers, cheapest first. Most fatigue is a hook problem, and a hook fix does not need a new shoot.

| Cadence | What changes | Typical cost | Trigger |
| :--- | :--- | :--- | :--- |
| Weekly | New first three seconds from existing footage | Editor time only | Thumbstop drops 20% week on week |
| Every 3 to 4 weeks | New body and demo, same creator | One brief, about ₦28,750 at the median | Hold rate falls, CTR softens |
| Every 6 to 8 weeks | New creator, same script | One brief plus a new brief fee | Frequency above 3, CPA drifting up |
| Quarterly | New angle or format | A small batch of briefs | The whole concept has plateaued |

The economics are the reason to plan the refresh instead of reacting to it. At the ₦25,000 median rate, a brief costs the brand ₦28,750 once the 15% protection fee is added, with the fee floored at ₦500. Four new videos a month is ₦115,000 in creator cost, which is usually less than the waste created by letting a tired ad run at rising CPMs for another three weeks.

Batch the work with one creator you trust. A creator who has already shot your product can deliver three hook variations from the same footage for an add-on fee, and those variations arrive in two days instead of two weeks.

## How many creatives to run at once

Three or four per ad set is enough. Every additional variant splits the delivery, and with the event volumes most Nigerian accounts generate, five variants in one ad set means none of them reaches significance.

Use the structure that matches your budget. If you can only fund one ad set, run four creatives inside it and let the platform rotate. If you can fund three ad sets, give each a distinct audience and keep the creative set identical across them, so the audience is the only variable.

Advantage+ style campaign setups will mix your assets automatically, which is useful for volume and unhelpful for diagnosis. Keep one manually structured campaign running alongside so you can still read which creative did what.

## When performance drops after a good month

Check four things in order before you blame the creative. Frequency, because audiences saturate quietly. The landing page, because a theme update or a new shipping plugin can cut conversion without touching the ad. Seasonality, because December spending, January cash flow and school fee months move Nigerian consumer behaviour hard. And the auction, because a competitor entering your category raises CPMs for everyone.

Only after those four does the creative become the suspect. Our breakdown of [UGC ad metrics](/blog/ugc-ad-metrics-thumbstop-hold-rate-ctr-and-cpa) shows how to tell a hook problem from a page problem in one pass.

Then brief the next round with a specific job: three new hooks, same body, delivered in four days. Write it as a gig on [ugc.ng](/gigs/new), name the platform and the aspect ratio, and put the refresh date in the brief. A creator who knows the deadline is Friday will shoot on Tuesday, and you will have replacement creative before the frequency number becomes a problem.
