How to Target a Specific Audience on Facebook Ads (Without the Black Box)

If you have run Meta budgets for more than a couple of years, you already know what happened: the targeting panel got quietly emptied. Interest and behaviour categories were culled in waves, employer and job-title options went, sensitive-category options went, and Advantage+ Shopping arrived with the audience controls switched off by design. The official answer to almost every targeting question is now the same three words: go broad, test creative.

That is a reasonable answer if you are a supermarket. It is a poor answer if your addressable market is a defined slice of the UK, because it converts your audience strategy into a creative treadmill. Creative is the last input most advertisers still control, so creative is where all the effort goes - ten new hooks a week, a rolling test calendar, an agency retained mostly to feed the machine - while the question that actually decides efficiency, who is this reaching, gets answered by a black box. For the full timeline of what was removed and when, see what Meta removing detailed targeting actually means.

This is the practical version: what you can still specify natively, what you genuinely cannot, and how to put the input layer back so the creative you are already paying for lands on households you chose.

What you can still target natively in 2026

Strip away the interface and there are four real levers left inside Ads Manager:

• Location - country, region, city, postcode, or a radius pin. Still the most robust control you have, and still blunt: a postcode is not a household.
• Age, gender, language - reliable, low-resolution, and rarely the variable that separates a buyer from a browser.
• Custom Audiences - lists you upload or generate from the pixel, plus the Lookalikes you seed from them. This is the only lever where you define the people.
• Detailed targeting - still present, materially thinner than it was, de-prioritised by delivery, expanded automatically whenever Meta thinks it can find cheaper results, and unavailable inside Advantage+ Shopping.

Read that list again and notice the asymmetry. Three of the four levers describe where and roughly who. Only one lets you name the actual people, and it depends entirely on data you bring with you.

Why broad quietly taxes the smaller advertiser

Broad targeting is not a trick. On a large, generic addressable market with a clean signal and heavy creative volume, it genuinely outperforms hand-built interest stacks - the model finds pockets you would never have written into a segment.

The problem is arithmetic. Suppose your product realistically suits four million UK households out of roughly thirty million. Broad delivery starts from the whole pool and learns its way inward using whatever conversion signal it can gather. Every impression served before it narrows is spend against households that were never candidates. With a large budget you buy your way through that learning phase quickly and it amortises. On five or fifteen thousand a month, you are funding an education you never graduate from - and every time you refresh creative or the auction shifts, some of that learning gets re-paid.

The fix is not to fight the algorithm. Meta's delivery is genuinely very good at picking the likeliest converter from the pool it is given. The lever worth having back is the pool.

The specific asks Meta can no longer answer

The detail behind each row is below. The mechanics of how delivery interprets your inputs are covered in how Meta ad targeting actually works.

Business owners, job titles, employers

Gone as a native option. Employer and job-title targeting were removed, and the interest proxies that survive ("small business", "entrepreneurship") select for people who read about business, not people who run one. The route that works: flag owner-operators in your own data - signup field, order type, VAT number, plan tier - and push that as a Custom Audience seed.

Members of a specific Facebook group

Not possible, and it never was. Group membership is not an ad-targetable object. Neither is "people who follow this competitor page". If a tool claims otherwise, it is scraping, and it will get your account actioned.

Specific named individuals

Only through a Custom Audience you own, and only in aggregate. You upload identifiers, Meta matches them, and audiences below the minimum size will not deliver. There is no "show this ad to this person" control, by design.

Ethnicity, religion, health, sexuality, politics

Removed as sensitive categories, and they are not coming back. Do not look for proxies - that is exactly the behaviour the policy exists to stop, and Meta's classifiers are looking for it.

A specific area, town or postcode

Partly answerable. Location targeting handles it at radius or postcode level, but a UK postcode can span a new-build estate and a row of bedsits. Where geography matters because of what the geography implies - affluence, tenure, property type - specify those directly and let location be a filter rather than the whole brief.

High income or high-value households

No native income targeting in the UK. This is the single most requested attribute and the widest gap in the panel. It is answerable, but only with data you bring: income band, purchasing power band, or property value band resolved at the household level and uploaded as a seed.

Homeowners versus renters

Not natively, and it is the difference between a wasted impression and a sale for anything fitted, installed or structural. Tenure is a household fact, not a browsing behaviour, so it has to come from a household dataset.

Recent movers, new parents, downsizers

Meta has thin, lagging signal on life-stage moments, which is unfortunate because these are the highest-spend windows in a household's year. Movers in particular: see recent movers, the highest-intent audience you are not targeting.

Pet owners, car owners, families with teenagers

Interest proxies exist and are weak - "loves dogs" is not "owns a dog and buys 12kg of food a month". Ownership and household composition are attributes, not affinities, and they behave completely differently in a model.

Getting the input layer back: build the audience outside Meta

Everything above converges on one conclusion. If Custom Audiences are the only lever where you define the people, then your audience strategy is now a data problem that happens before Ads Manager opens.

Which makes seed quality the whole game. A good seed is specific enough that the definition would survive being read aloud to your CFO, large enough to clear Meta's minimums and give delivery room to work, current rather than a 2023 export, and paired with a suppression list so you are not re-buying customers you already own. A bad seed is "all purchasers, all time" - a list whose only shared characteristic is that they once bought something, which is precisely the audience the algorithm was already going to find. Depth on the mechanics: Facebook Custom Audiences, and on why a named segment beats a modelled one, household audiences versus lookalikes.

What multi-dimensional household data lets you specify

Outra is often shelved as "the property and movers people". Those are two of the sharpest dimensions we hold, but the dataset is considerably wider than that, and most of the interesting briefs cut across several dimensions at once:

• Life stage - employment stage, age band, family stage (pre-school through older teenagers), move stage
• Household - occupancy status, property type, bedrooms, property size, property value band, garden, garage, parking, area type, region
• Purchasing power - purchasing power band, household income band, credit score band, ultra-high wealth
• Lifestyle - pet owners, car owners, multi-car households

All of it hangs off one identifier: the household. Not a cookie, not an email, not a device graph that decays the moment someone changes phone.

What that looks like as a brief:

• Garden furniture - owner-occupied, homes with gardens, high purchasing power, suburban and rural. Three dimensions, and the second one is the whole business.
• Premium pet food - pet owners, established professionals, higher purchasing power. No property attribute anywhere in the definition.
• A financial product - household income band plus credit score band plus age band. Again, zero property signal - the dimensions that matter are affluence and life stage.
• Home improvement - just moved, detached owners, mid-to-upper property value band. Here property and move stage are the brief, which is why they get the billing they do.
• A family SUV accessory - multi-car households, families with secondary-school teens, garage or driveway present.

Each of those is a sentence a marketer can defend, and each one resolves to a list of UK households that can be pushed into Meta as a Custom Audience.

Suppression is targeting too

Half the value of controlling the input is controlling who you exclude. Email-based suppression misses the same household under a different address, a partner's inbox, or a new email after a phone upgrade - so a meaningful share of "prospecting" budget goes on people already in your CRM. Suppress at the household level and prospecting budget stays prospecting. It is the least glamorous change on this page and frequently the one that moves CPA first.

How to actually run it

1. Write the audience brief in one sentence. If it needs a paragraph, it is two audiences.
2. Resolve your customer file to households and look at what the buyers actually share. This routinely contradicts the persona deck.
3. Select prospect households against those dimensions, screened against your existing customers.
4. Push the segment into Meta as a Custom Audience, with the customer file loaded as an exclusion.
5. Hold the audience fixed while you test creative. A creative test against a shifting broad pool measures nothing - see how to test paid social creative properly.
6. Read results per segment, not account-wide. Segment-level CPA is what tells you whether the brief was right; the blended number tells you almost nothing.

Where this does not help

Three honest caveats. Match rates are never one hundred percent - you are matching households to Meta accounts, and some fraction will not connect, so build seeds with headroom. Very tight definitions can fall below Meta's minimum audience size, and an audience that will not deliver is worse than a broader one that does; widen a dimension rather than abandoning the brief. And if your product genuinely suits most UK adults and you have the budget and creative volume to feed broad delivery, broad plus strong creative may already be your best answer - in that case the value here is suppression and measurement, not selection.

Everywhere else, the position is simple: Meta removed the input, not the need for it. Bring the input yourself and creative goes back to being creative rather than the only lever you have left.