---
title: "Audiences and targeting in a privacy-first era"
description: "How targeting has changed A few years ago, advertisers built narrow audiences stacked with interests and demographics. Privacy changes (such as Apple's…"
url: https://optimizeall.com/learn/paid-social-advertising/audiences-and-targeting
updated: 2026-10-05
---

Paid Social Advertising · Campaign structure, objectives, audiences and automation · lesson 2 of 17 · 14 min

# Audiences and targeting in a privacy-first era

## How targeting has changed

A few years ago, advertisers built narrow audiences stacked with interests and demographics. Privacy changes (such as Apple's App Tracking Transparency), regulation and improved machine learning have shifted best practice. Today, platforms often find buyers better with **broad targeting plus strong creative and good conversion data** than with narrow manual audiences. Your creative increasingly *is* your targeting: the message attracts the right people and the algorithm learns from who responds.

That does not make audiences irrelevant. You still need to understand the main types and when to use them.

## Main audience types

1. **Broad / Advantage+ audience:** only location, language and basic age limits, letting the algorithm decide. With Meta's Advantage+ audience, any age, interest or custom-audience inputs you add act as **suggestions** the system may go beyond; only a few settings (such as location, minimum age and certain exclusions) remain hard controls. Works best with enough conversion data and budget.
2. **Interest and behaviour targeting:** people the platform associates with topics (for example "skincare" or "small business"). Useful as a guide or for small budgets, but less precise than it looks.
3. **Custom audiences (first-party):** people who already interacted with you – website visitors, video viewers, page or profile engagers, lead-form openers, app users, or customer lists uploaded with a lawful basis.
4. **Lookalike / similar audiences:** new people who resemble a source audience such as purchasers. Their value has declined somewhat as broad targeting has improved, but they remain useful for some accounts.
5. **B2B targeting (LinkedIn):** job title, function, seniority, company, company size and industry. Powerful for B2B but costs per click are generally higher than on consumer platforms.
6. **Exclusions:** remove recent purchasers from acquisition ads, or existing leads from lead campaigns, to avoid wasted spend. Custom-audience exclusions still work on most platforms, but Meta has removed detailed-targeting (interest) exclusions. In Advantage+ sales campaigns, you can instead define **existing customers** with custom audiences and cap the share of budget spent on them.

## Special and restricted categories

Ads related to **housing, employment, credit or financial products**, and **social issues, elections or politics** fall into special categories on Meta and similar policies on other platforms. Targeting options are restricted (for example age, gender and postcode targeting may be limited), and political ads usually require authorisation and "paid for by" disclaimers. Some platforms restrict or do not allow political ads at all – for example, Meta and Google stopped serving political, electoral and social-issue ads in the EU from October 2025 in response to the EU's political-advertising regulation (check current status). Declare the category honestly; failing to do so leads to rejections and account restrictions.

Health and wellness advertisers may also face limits on the data and events they can use for targeting and optimisation in some markets. Always check current policy for your industry.

## Location and language nuance

- **Gulf markets:** audiences are highly multicultural. A UAE campaign may need Arabic and English versions, and sometimes Urdu, Hindi, Tagalog or Malayalam for specific communities.
- **Pakistan:** Urdu, English and Roman Urdu creatives can perform differently; test them.
- **UK and US:** regional targeting (for example by city) can help local businesses, but avoid audiences so small that delivery stalls.
- Target by **location** carefully: "people living in" versus "people recently in" a place matters for tourism and events.

## Audience sizing and overlap

- Very small audiences drive up costs and frequency quickly.
- Overlapping ad sets compete against each other in the same auction. Consolidate where overlap is high.
- Retargeting pools shrink if you stop generating new awareness, so balance prospecting and retargeting spend.

## Worked example: a UK B2B SaaS company

A Manchester-based invoicing app for freelancers:

- **LinkedIn:** targets job functions and seniority associated with self-employed consultants and small agency owners, with a Lead Gen Form offering a free invoice template.
- **Meta:** broad audience in the UK aged 21+ with creative that calls out freelancers explicitly ("Freelancer? Stop chasing invoices"), letting the creative filter the audience.
- **Retargeting:** website visitors in the last 30 days who did not sign up, with a testimonial video.
- **Exclusions:** existing customers uploaded from the CRM (covered by the company's privacy notice) are excluded from acquisition campaigns.

## Common mistakes

- Stacking many interests into tiny audiences.
- Ignoring special ad category rules.
- Uploading customer lists without an appropriate lawful basis or notice.
- Forgetting to exclude recent customers.

## Hands-on: an audience testing matrix

Plan audiences as a small matrix rather than dozens of ad sets. Each row is a hypothesis you can test with the same creative:

| Test | Audience A | Audience B | Constant | Primary metric | Min. run |
|---|---|---|---|---|---|
| Broad vs guided | Advantage+ audience, no inputs | Advantage+ with purchaser custom audience as suggestion | Same 6 ads, same budget | Cost per purchase | 7 days / 50 purchases total |
| Language | Arabic creatives, UAE | English creatives, UAE | Offer, budget | Cost per purchase | 7 days |
| B2B seniority (LinkedIn) | Manager | Director+ | Lead form, ad | Cost per SQL (from CRM) | 3–4 weeks |

Use the platform's A/B test tool for clean splits where possible.

## Hands-on: preparing a customer list lawfully

Before uploading a customer list for custom or lookalike audiences:

1. **Check the lawful basis and your privacy notice** – it should say you may use contact details for advertising on social platforms, and in the EU and UK you may need consent.
2. **Remove** people who opted out of marketing and anyone under the platform's minimum age.
3. **Normalise** data: lowercase and trim emails; phone numbers in international format with country code (for example `+923001234567`, `+971501234567`).
4. **Hash** with SHA-256 if you prepare files yourself (Meta and other platforms also hash automatically in their upload tools). A quick Python example:

```python
import csv, hashlib

def norm_hash(value: str) -> str:
    return hashlib.sha256(value.strip().lower().encode("utf-8")).hexdigest()

with open("customers.csv", newline="", encoding="utf-8") as f, \
     open("customers_hashed.csv", "w", newline="", encoding="utf-8") as out:
    writer = csv.writer(out)
    writer.writerow(["email_sha256"])
    for row in csv.DictReader(f):
        if row.get("marketing_opt_out") == "yes":
            continue  # respect opt-outs
        email = row.get("email", "")
        if "@" in email:
            writer.writerow([norm_hash(email)])
```

5. **Refresh** lists regularly (or connect your CRM or e-commerce platform) so exclusions stay current.

## Worked example 2: a multilingual UAE campaign

A Dubai supermarket delivery app wants new customers across communities. Instead of 12 language-and-interest ad sets, it runs **one Advantage+ campaign** with ads in Arabic, English, Urdu, Hindi and Tagalog, each video naming the community's favourite products and festivals. The creative does the targeting: the algorithm learns who responds to which ad. A separate exclusion keeps existing customers in a small retention campaign. Reporting by ad (language) shows which communities are cheapest to acquire, which then guides creator partnerships.

## How to measure success

- Prospecting spend reaches genuinely **new** customers (check the share of new vs existing customers in store data or Meta's new/existing customer reporting where available).
- Frequency in prospecting stays moderate; retargeting audiences keep being refilled by prospecting.
- No policy flags for special ad categories; customer lists refreshed on schedule.

## Video lecture: Audiences in a privacy-first era: when your creative does the targeting

Lecture coming soon · 11 chapters · about 8 minutes. Read the full transcript below.

1. Audiences in 2026
2. What changed
3. Six audience types
4. Exclusions and special categories
5. Declare honestly, localise wisely
6. Example 1: Manchester invoicing app
7. Example 2: Dubai grocery app (illustrative)
8. Customer lists, lawfully
9. Audience testing matrix
10. Measure success
11. Recap and try this now

## Lecture transcript

### Audiences in 2026

A few years ago, running social ads felt like stacking Lego. Women, twenty-five to thirty-four, interested in yoga, and organic food, and travel, living in three postcodes. The more bricks you stacked, the cleverer you felt. Today, that approach often loses to a campaign with almost no targeting at all. Why? In this lecture you'll learn how targeting has changed, the six audience types and when each still matters, the special categories you must declare, how to prepare a customer list lawfully, and how to test audiences with a simple matrix instead of dozens of tiny ad sets.

### What changed

Here's what changed. Apple's App Tracking Transparency, privacy laws and browser limits reduced the data platforms can see. At the same time, their machine learning got much better at finding buyers from signals like who watches, who clicks and who purchases. So today, broad targeting plus strong creative plus good conversion data often beats narrow manual audiences. Here's the key idea: your creative increasingly is your targeting. A video that says, freelancer, stop chasing invoices, attracts freelancers, and the algorithm learns from whoever responds. The message filters the audience. That doesn't make audiences irrelevant. You just use them differently.

### Six audience types

Let's go through the six audience types. One, broad, or Advantage plus audience on Meta: just location, language and basic age. Any extra inputs you add act as suggestions, not fences. Two, interests and behaviours: useful as a guide or for small budgets, but less precise than they look. Three, custom audiences: people who already interacted with you, like site visitors, video viewers, engagers and uploaded customer lists. Four, lookalikes: new people similar to a source like purchasers. Five, B2B targeting on LinkedIn: job title, function, seniority, company and industry. And six, exclusions: keep recent buyers out of acquisition ads.

### Exclusions and special categories

A couple of updates on exclusions. Custom-audience exclusions, like excluding your purchasers, still work on most platforms. But Meta has removed interest-based exclusions. And in Advantage plus sales campaigns, instead of excluding customers entirely, you can define your existing customers with custom audiences and cap how much of the budget reaches them. That's handy because some existing customers will buy again anyway, and you don't want to pay to reach them. Now, special ad categories. Housing, employment, credit and financial products, and social issues, elections or politics have restricted targeting, and political ads need authorisation and disclaimers.

### Declare honestly, localise wisely

Some platforms go further. Meta and Google stopped serving political, electoral and social issue ads in the EU from October twenty twenty-five, in response to the EU's political advertising rules. So always check current status for the market you're in. Declare the category honestly. If you don't, you'll face rejections and account restrictions. Health and wellness advertisers should also check current rules, because some platforms limit the data and events they can use in certain markets. And think about language and location. A UAE campaign may need Arabic and English, and sometimes Urdu, Hindi or Tagalog. In Pakistan, Urdu, English and Roman Urdu can perform very differently.

### Example 1: Manchester invoicing app

Let's do a simple worked example. A Manchester invoicing app for freelancers. On LinkedIn, it targets job functions and seniority linked to self-employed consultants and small agency owners, with a Lead Gen Form offering a free invoice template. On Meta, it goes broad, UK, aged twenty-one plus, with creative that calls out freelancers directly: freelancer, stop chasing invoices. Retargeting shows a testimonial video to site visitors from the last thirty days who didn't sign up. And existing customers, uploaded from the CRM and covered by the privacy notice, are excluded from acquisition. Four audience types, each with a clear job.

### Example 2: Dubai grocery app (illustrative)

Now a realistic scenario. A Dubai grocery delivery app wants new customers across many communities. The old plan was twelve ad sets: every language crossed with every interest. Each one got a tiny budget and none left the learning phase. The new plan is one Advantage plus campaign with ads in Arabic, English, Urdu, Hindi and Tagalog, each naming that community's favourite products and festivals. The creative does the targeting. The algorithm learns who responds to which ad. A separate small campaign handles existing customers. And reporting by ad shows which communities are cheapest to acquire, which then guides which creators they hire.

### Customer lists, lawfully

If you use customer lists, prepare them lawfully. Check your lawful basis and privacy notice cover advertising on social platforms. In the EU and UK, you may need consent. Remove people who opted out and anyone under the platform's minimum age. Normalise the data: lowercase, trimmed emails, and phone numbers in international format with the country code. Platforms hash the data with an algorithm called SHA two fifty-six when you upload through their tools, and you can hash it yourself first. There's a short Python script in the lesson text that does exactly that and skips opted-out customers. And refresh lists regularly so exclusions stay current.

### Audience testing matrix

How do you test audiences without creating chaos? Use a matrix. Each row is one hypothesis, with everything else held constant. Broad with no inputs versus broad with a purchaser audience as a suggestion, same six ads, same budget, judged on cost per purchase over seven days. Arabic versus English creatives in the UAE, same offer. On LinkedIn, managers versus directors, judged on cost per sales-qualified lead from the CRM over three to four weeks. Use the platform's A/B test tool for clean splits. Common mistakes: stacking interests into tiny audiences, ignoring special categories, uploading lists without a lawful basis, and forgetting to exclude recent customers.

### Measure success

How will you know your audiences are working? Prospecting should reach genuinely new customers, so check the share of new versus existing buyers in your store data, or in Meta's new and existing customer reporting where it's available. Prospecting frequency should stay moderate. Retargeting pools should keep being refilled by prospecting, because if you stop generating new awareness, retargeting audiences shrink and get expensive. And you should see zero policy flags for special categories, with customer lists refreshed on schedule. If prospecting is quietly spending on existing customers, your costs look fine but your growth stalls.

### Recap and try this now

Let's recap. Targeting has shifted from narrow manual audiences to broad delivery guided by creative and conversion data. Know the six audience types and use each for its job. Declare special categories honestly, localise language, and prepare customer lists lawfully with opt-outs removed and data normalised and hashed. Test audiences with a simple matrix instead of dozens of ad sets. Here's your try this now. For a business you know, write one broad prospecting audience, one custom retargeting audience and one exclusion. Then add one audience test to the matrix, with its constant, metric and minimum run, and note whether any special category applies.

## Key takeaways

- Broad targeting with strong creative and good conversion data often beats narrow manual audiences today.
- Know the main types: broad, interests, custom, lookalike, B2B firmographic and exclusions.
- Special ad categories (housing, employment, credit, politics) restrict targeting and must be declared.
- Consider language and location nuances, avoid tiny or overlapping audiences and use lists lawfully.

## Try it

For a business you know, write down one broad prospecting audience, one custom retargeting audience and one exclusion. Note whether any special ad category applies.

- [Previous: Campaign structure and objectives across platforms](https://optimizeall.com/learn/paid-social-advertising/campaign-structure-and-objectives)
- [Next: Steering automation: Advantage+, Smart+, Accelerate and Performance+](https://optimizeall.com/learn/paid-social-advertising/platform-automation-advantage-smart)
- [All lessons of Paid Social Advertising](https://optimizeall.com/learn/paid-social-advertising)