Broad targeting and Advantage+ audience usually beat narrow interests under Andromeda, but not always. Here is how targeting still works in 2026, when broad fails, and how to use targeting as a guardrail instead of a precision tool.
"Go broad" has become the default advice for Meta ads. Like most default advice, it is right most of the time and wrong some of the time. This article explains how targeting actually works under Meta Andromeda, why broad usually wins, when it does not, and how to test it in your account.
How targeting works under Andromeda
Meta's October 2026 engineering post was explicit: index matching enforces each advertiser's targeting rules before ads are scored. So your targeting defines the pool of people your ad is eligible for. Andromeda then decides, for each person in that pool, whether your ad makes the shortlist.
That gives targeting a new role. It is no longer how you find buyers; it is how you rule people out. Every restriction you add shrinks the pool Andromeda can search. If the restriction removes people who would never buy, that is good. If it removes people who would buy but do not match your guess, you have paid to miss them.
Your targeting options in 2026
- Broad: location, age and gender only, with no detailed targeting.
- Advantage+ audience: Meta's AI targeting. You can add audience suggestions, such as interests or custom audiences, which Meta uses as a starting point but can go beyond.
- Original audience options: classic detailed targeting, lookalikes and custom audiences used as hard limits.
- Audience controls: hard limits that still apply with Advantage+ audience, such as location, minimum age and exclusions.
Why broad usually wins
- A bigger pool for retrieval. Andromeda can consider your ad for more people, so it finds buyers your interest list would never include.
- Interests are rough signals. Someone interested in "marketing" might be a student; your buyer might have no relevant interest listed at all.
- Creative already targets. A video about payroll problems for finance leads will mostly reach finance leads, because they are the ones who respond.
- More data per ad set. Broad ad sets usually get more conversions, which helps them exit learning and optimise faster.
When broad does not work
Your conversion signal is weak
If you optimise for page views, link clicks or cheap form fills, broad targeting finds people who do exactly that, often low-quality ones. Fix the signal first: optimise for purchases, qualified leads or activated trials, using the Conversions API.
Your creative is vague
Broad works because creative tells Meta who it is for. A generic ad that could be for anyone gets shown to anyone. Make each ad clearly about a specific problem for a specific person. See creative diversification.
Your account has little history
New accounts or pixels with few conversions give Meta less to learn from. Here, audience suggestions in Advantage+ audience, or a lookalike of your customer list, can speed up the start.
Your audience is truly tiny
If your buyers are a few thousand people at named companies, Meta's broad approach may never find them efficiently. Account-based approaches with uploaded lists, or other channels, may fit better. Our account-based marketing glossary entry and LinkedIn marketing guide cover the alternatives.
Quick quiz
Broad or narrow?
Four scenarios. Pick the better starting point.
1. An online store with thousands of purchases a month and varied creative.
2. A new account with a fresh pixel and almost no conversions yet.
3. Under Andromeda, are your targeting rules still enforced?
4. Broad targeting brings lots of cheap but useless leads. What should you fix first?
0 of 4 answered
How to test broad against your current targeting
- Keep everything else equal: same creative, same optimisation event, same budget.
- Use a clean split for big decisions. Meta's A/B test tool in Ads Manager splits people so no one sees both versions.
- Run long enough: at least two weeks, and until each side has a meaningful number of conversions.
- Judge on real outcomes: cost per purchase or per qualified lead, checked against your CRM, not cost per click.
- Look at who converted: for B2B, check job titles and company fit of the leads each side produced.
Exclusions and guardrails worth keeping
- Existing customers, when your offer is only for new ones.
- Countries you cannot serve.
- Age limits required by law or your product.
- Recent converters, to avoid paying to reach people who just bought.
- Employees and job seekers, by excluding your own page engagers where it makes sense.
Frequently asked questions
It works, but usually not as well as broad or Advantage+ audience when you have good creative and a strong conversion signal. Test both in your account.
Meta's AI targeting option. You can give it suggestions, like interests or custom audiences, and it may go beyond them if it finds better results. Hard controls like location and minimum age still apply.
Not dead, but less necessary. A lookalike of your best customers can still help new accounts or serve as a suggestion inside Advantage+ audience.
Often yes, if they optimise for a deep event like qualified leads. See Meta ads for B2B SaaS.
