How small is too small for a retargeting audience?

The correction first

Meta publishes no minimum custom audience size. Not 100. Not 1,000. Not for website custom audiences.

The figures in circulation trace to a Meta Business Help page titled Troubleshoot website custom audience size, at facebook.com/business/help/237515166435276. That URL now redirects to a different help page, whose guidance on the point is entirely non-numeric. Verbatim, typo included:

“If your audience is size is still too small, consider: Waiting until you have several hundred people in your website custom audience before using it…”

“Several hundred” is advice. It is not a minimum, it is not enforced, and it is not a number you can plan against. Meta's About custom audiences page does not contain a minimum either — no floor, no recommended size, nothing but an account-level cap on how many audiences you can have.

Search engines still index the old URL under its old title. That is the whole explanation for why a number nobody can source keeps appearing in guide after guide: the page looks alive in search results, and almost nobody clicks it.

How we checked this

A page whose argument is “everyone is repeating a source they never opened” owes you its own method. So:

• The redirect and the quoted text were read in a browser on 27 July 2026. Meta Business Help pages carry no revision date, so an accessed date is the only correct citation form for them.
• They could not be re-confirmed automatically. facebook.com/business/help blocks automated fetching — a server-side request returns an empty document rather than the page. If you want this verified, a browser is the only tool that works, and both links above are right there.
• The indexing half was re-confirmed today. Search still returns the old URL under the title Troubleshoot website custom audience size, which is the mechanism by which the dead figure keeps propagating.

If Meta later publishes a documented minimum, this page is wrong and we will change it. That is a different thing from being unsourced.

Why “how small is too small” is the wrong question

A retargeting audience does not have a cliff. It has three things that get gradually worse as it shrinks, and none of them announces itself.

Asking for a threshold assumes there is a size at which the campaign switches from working to not working. There is not. There is a size at which the campaign becomes a bad use of your money, and that size depends on your budget, your creative, and how long you intend to run — not on a number Meta was never publishing.

What actually degrades

1. Cost per impression rises

This is the least understood of the three and the most expensive.

A narrow audience means fewer available impressions. Your budget does not shrink to match. To spend it, the delivery system has to win a larger proportion of the auctions that exist for that small group of people — and it is bidding against every other advertiser who also wants them, in a pool where there are far fewer alternatives to fall back on.

Andrew Southworth is, as far as we can tell, the only person writing in the music vertical who names this at all: a small custom audience, he notes, “can cause higher auction rates”. He is right, and the fact that almost nobody else mentions it is why artists conclude their creative is bad when the actual problem is that they pointed a full daily budget at a very short list.

There is no published curve for this. Meta does not release one, and any specific cost-per-thousand figure you find attached to an audience size was not measured on your account. What matters is the direction, and the direction is reliable: narrower pool, higher pressure.

2. Frequency compounds fast

The same budget divided among fewer people means each person sees the ad more often. On a small warm pool this compounds within days rather than weeks, and the first symptom is not fatigue — it is cost, because a diminishing response to the same creative pushes your cost per result up while the audience count on the screen has not changed at all.

3. The optimiser stops learning

Delivery optimisation improves with conversion volume. A small audience produces few conversions, so the system has less to work with and takes longer to get anywhere — and on a short campaign it may never get there. This is why a small audience and a short flight are a bad combination even when each would be survivable alone.

It is also why the events you send matter more when the audience is small. If you are optimising for a shallow event, you are spending a scarce signal on something that does not predict much.

The size/freshness trade-off

The retention window is the control everyone reaches for when an audience is small, and it is the one most likely to make things quietly worse.

A retention window looks like a size setting. It is a freshness setting. Going from 30 days to 180 days does make the audience bigger — by adding exactly the people with the oldest and weakest signal. Someone who visited your release page last week is a materially different prospect from someone who visited six months ago, and widening the window does not give you more of the first kind. It gives you more of the second kind, mixed in with the first, addressed by the same ad.

There is a second cost that is easy to miss. A longer window pulls in more history, and history is where the contamination lives — bought promotion, an inflated period, a campaign you no longer trust. Widening the window to fix a size problem can reintroduce a quality problem you already solved. That interaction is covered in bot traffic in retargeting audiences.

Widen the window, or widen the source?

When an audience is too small to run against, you have two moves. They are not equivalent and the order is not obvious, so here it is plainly.

Widen the source first. That means adding more kinds of warm signal rather than more history: video viewers alongside site visitors, engagers on the platform itself, an email list you already own. Everyone you add this way is a fresh signal from someone who did something recently. You are making the audience bigger without making it older.

Widen the window second, and treat it as a decision with a cost rather than a free size increase. Do it when your source is genuinely maxed out, when the older traffic came from a period you trust, and preferably as a separate ad set so you can see whether the old cohort performs at all instead of averaging it into the new one.

Or do neither, and spend less. A small audience with a small daily budget is a perfectly reasonable campaign. Most of the damage attributed to small audiences is really damage from a normal-sized budget aimed at an unusually short list. Reducing spend is a legitimate answer to “my audience is small”, and it is the one nobody offers because it does not sound like a tactic.

Lookalikes are not a size fix

The standard advice when a warm audience is small is to build a lookalike. It is worth being precise about what that does, because it is not what people think.

A lookalike is not a larger version of your warm audience. It is a model built from your warm audience and projected onto people who have never heard of you. It is a cold audience with a better starting guess. Treating it as warm — same creative, same ask, same expectations — is one of the most common ways an indie campaign wastes money.

Two consequences follow. First, a lookalike needs a source audience, so it cannot rescue you from having no warm audience at all; it can only extend one you already have. Second, it inherits everything wrong with that source. Contaminate the seed and you do not get a slightly worse audience — you get a large, confident, precisely wrong one. Meta's own guidance on lookalikes emphasises the quality of the source over its size, and that is the useful part: a short list of people who actually bought something is a better seed than a long list of everyone who ever loaded a page.

We are not going to quote you a lookalike source-audience minimum. Meta may publish requirements for lookalike creation, and unlike the custom-audience minimum those may well be real — but we did not re-verify them on the day this page was written, and this page in particular has no business repeating a number it did not check. Meta's own lookalike documentation is the only authority worth trusting on it.

What to do if your audience is genuinely small

In rough order of how much they help:

• Lower the daily budget before you widen anything. Most small-audience problems are budget-density problems.
• Add sources, not history. Video viewers, platform engagers, your email list. Fresh signal beats old signal.
• Run fewer ad sets. Splitting a small pool across two ad sets halves the signal each one gets. See exclusions and audience overlap.
• Fix the events before blaming the size. An audience that is small because events are not being recorded properly is a tracking problem wearing an audience problem's clothes.
• Let it fill. A retargeting audience is a rolling window fed by traffic. If the traffic is real, waiting is a strategy.

A note on numbers

This page corrects a number and does not replace it with one of ours. That is on purpose. Substituting git-royal's invented minimum for Meta's imaginary one would be worse, not better.

Where this page states a fact about Meta's behaviour, it is linked to Meta's own pages with an accessed date, or it is described as undocumented. Where we could not verify something on the day of writing, we said so — see how we checked this.

Common questions

Where this fits

This is part of our guide to retargeting for musicians. If your audience is small because events are going missing rather than because your traffic is, server-side tracking is the fix, and the Conversions API setup takes a few minutes. If it is small because you have been splitting it across ad sets, start with exclusions and overlap.