Meta lookalike audiences still work in 2026 — but they work differently, and advertisers who have not updated their approach since 2021 are leaving significant budget on the table. As part of any serious Meta Ads strategy for South Africa, lookalike targeting has shifted from a "set it and forget it" cold-prospecting tool into a precision seed-signal layer that feeds Meta's Advantage+ machine. The question worth answering is not whether to use them, but how to build the seed that makes them useful.
South Africa's Facebook audience of 27.9 million active users — 61.7% of adults — means even a tightly matched 1% lookalike generates a prospecting pool of roughly 279,000 people. That is meaningful reach for most SA campaigns, provided the signal going in is clean. Understanding how Meta audience targeting works in SA is the prerequisite; this post focuses specifically on the lookalike layer.
Quick Answer
Meta lookalike audiences remain a viable prospecting tool in 2026, but their role has changed: they now work best as a premium signal layer inside Advantage+ campaigns rather than as standalone cold-prospecting audiences. The gap between a generic lookalike and broad targeting has narrowed since iOS 14.5, so the quality of your seed audience — not the lookalike percentage — is the primary lever. High-LTV customer lists, recent purchaser segments, and Klaviyo-synced email audiences consistently carry stronger prospecting signal than large generic seed lists. Combined with a properly implemented Conversions API, they remain one of the strongest prospecting methods available to SA advertisers.
In This Guide
Running Meta Ads in South Africa and unsure whether your targeting setup is still fit for purpose? Get a Meta Ads audit from Growth Pulse Media — we'll tell you exactly what to fix.
What Are Meta Lookalike Audiences?
A Meta lookalike audience is a targeting group built by Meta's algorithm to find people who share statistically similar characteristics to an existing "seed" audience you provide. You upload a source — a customer list, a pixel event, a video engagement audience — and Meta maps that seed against its user base to find people with comparable behaviour patterns, interests, and demographics who have not yet engaged with your brand.
The Meta Marketing API specifications confirm the core mechanics: seed audiences require at least 100 matched people from a single country, though the practical threshold for reliable modelling sits at 1,000–5,000 matched records. The size of the resulting audience is controlled by a ratio from 1% to 20% of the target country's eligible user base, with 1% being the tightest match and 10%+ trading precision for reach. Membership refreshes every three days for active ad sets, and the audience takes one to six hours to fully populate after creation.
Key Takeaway
The technical minimum to create a lookalike is 100 matched people. The practical minimum to generate useful signal is closer to 1,000 — and quality of those records matters far more than sheer volume.
How Did iOS 14.5 Change Lookalike Targeting?
Apple's App Tracking Transparency framework, rolled out in April 2021, was the most disruptive event in paid social history. Between 70% and 80% of iPhone users opted out of cross-app tracking, and global opt-in rates settled around 15–25%. For Meta, this meant that pixel-based audiences — the source for most standard lookalikes — became materially smaller and less accurate overnight.
The downstream effect on lookalike performance was significant. Pixel-based lookalikes built purely from website visitor behaviour now perform only marginally better than broad targeting in many accounts, because the modelling data underneath them is incomplete. Where a 2019 lookalike was built from precise purchase signals across thousands of iOS users, the 2026 equivalent is built from a partial picture: Android users who opted in, iOS users who chose to share data, and server-side signals from accounts running CAPI.
What no longer works: Creating a lookalike from "all website visitors" using only the Meta Pixel. With 70–80% of iPhone users opted out, a pixel-only audience under-represents a large portion of your real customer base and produces a distorted lookalike model.
The March 2025 removal of detailed-targeting exclusions added another layer: advertisers who previously used exclusions to prevent lookalike campaigns from re-targeting existing customers can no longer do this natively inside Meta. CRM-level audience management — cleaning your suppression lists before they reach Meta — is now the only reliable method.
What Makes a Strong Seed Audience in 2026?
Seed quality is the single most important variable in lookalike performance. A tightly curated list of 500 verified high-value purchasers will typically carry stronger modelling signal than a list of 5,000 newsletter subscribers, because Meta's algorithm uses the seed to find patterns — and newsletter sign-ups are a weak signal compared to repeat purchase behaviour.
In descending order of signal strength, the most effective seed sources for SA advertisers in 2026 are:
| Seed Source | Signal Quality | Practical Notes for SA |
|---|---|---|
| High-LTV customers (top 20% by lifetime spend) | Strongest | Export from Shopify or your CRM; include email + phone for higher match rate |
| Repeat purchasers (2+ orders) | Strong | Filters out one-off buyers; ideal for fashion, beauty, supplements |
| Recent purchasers (last 30–90 days) | Strong | Use a 30-day window for fast-moving products; 90-day for higher-consideration items |
| Value-based lookalike (weighted by spend) | Strong | Upload customer list with a "value" column; Meta weights higher spenders more heavily |
| Video viewers (50%+ completion) | Moderate | Useful for brand awareness; lower intent signal than purchaser data |
| Email subscribers only (no purchase) | Weakest | Better than nothing, but consider filtering to engaged-only segments |
For South African ecommerce stores using Klaviyo, the most reliable workflow is a daily Klaviyo-to-Meta sync of purchase-based segments, keeping suppression lists (existing customers) updated at CRM level before they hit Meta. This replaces the targeting exclusions Meta removed in March 2025.
POPIA note for SA advertisers: Customer lists uploaded to Meta as seed audiences must be data you collected with a valid lawful basis under POPIA. Consent is one basis — but so are contractual necessity and legitimate interests. Whichever basis applies to your customer relationship, ensure you have it documented before uploading to Meta's Custom Audiences tool.
September 2025 audience restriction — critical for SA finserv and healthcare advertisers: From 2 September 2025, Meta's platform blocks Custom Audiences and lookalike audiences that suggest specific health conditions or financial status. South African advertisers in financial services, medical aid, insurance, and healthcare must audit their seed audiences to ensure event data does not imply these categories. Meta returns API error code 471 for flagged audiences, preventing campaign delivery entirely. If you operate in these verticals, review your seed composition with your Meta Ads manager before building lookalikes.
Which Lookalike Percentage Should SA Advertisers Use?
The percentage controls how closely the resulting audience matches your seed: 1% is the tightest match, 10% trades precision for reach. With South Africa's Facebook base at 27.9 million users, these translate into very different prospecting pools.
| Lookalike % | Estimated SA Audience Size | Best For | Verdict |
|---|---|---|---|
| 1% | ~279,000 | High-quality seeds; value-based; tight budgets | Start here |
| 2–3% | ~558,000–837,000 | Practical sweet spot for most SA campaigns | Scale once 1% is profitable |
| 5% | ~1.4 million | Broader prospecting with scale budget | Test after 2–3% is proven |
| 10%+ | ~2.8 million+ | Large awareness budgets; looser matching | Rarely matches Advantage+ broad performance at this size |
A practical starting framework is to test 1%, 3%, and 5% in separate ad sets using your best seed (purchasers from the last 180 days), evaluate cost per purchase at equal spend, and then consolidate budget behind the winner. Note: 180 days is Meta's full conversion-history modelling window — a different concept from the 30–90-day recency ranges in the seed table above. Those govern which recent buyers you select when building a recency-focused seed; this governs how far back Meta looks internally when modelling your audience. These are practitioner guidelines, not Meta-published benchmarks — your results will vary based on creative, offer, and campaign objective.
Key Takeaway
With only 27.9 million SA Facebook users, 10% lookalikes eat most of the country's eligible audience. For most SA advertisers, 1–3% provides a meaningful match quality advantage and enough reach for a healthy campaign volume.
How Meta Lookalike Audiences Fit Inside Advantage+
The most important structural shift in 2026 is how Meta Advantage+ Audience targeting has repositioned lookalikes. Under Advantage+, any audience you add — including lookalikes — becomes a suggestion to the algorithm rather than a hard constraint. Meta will start delivery within your specified audience, then expand beyond it when the algorithm detects stronger performance signals elsewhere.
This is not a downgrade — it is an upgrade, if you provide good signal. A value-based lookalike fed into an Advantage+ campaign gives the algorithm a head-start with a relevant starting population while still allowing it to discover segments you would not have targeted manually. A 2025 vendor study by Coinis — an ad-tech company with a commercial interest in automated campaign adoption, cited across industry blogs — found that Advantage+ automated campaigns delivered a 22% higher ROAS compared to equivalent manual setups. Treat this as directional rather than independently verified; SA results will vary by account, vertical, and campaign structure.
What works: Upload a high-LTV purchaser list as a Custom Audience, create a 1–3% lookalike from it, then add that lookalike as an "audience suggestion" inside an Advantage+ campaign. The algorithm uses your lookalike as a warm starting signal, then expands where performance improves — combining your data quality with Meta's reach.
Advantage+ shopping campaigns grew 70% year-over-year in Q4 2024, according to practitioner-cited industry reporting, signalling where Meta is investing its delivery infrastructure. Lookalikes fed into Advantage+ campaigns are where experienced advertisers are finding consistent prospecting results — not lookalikes run as standalone manual-targeting campaigns.
Why the Conversions API Is Now Non-Negotiable for SA Advertisers
A lookalike audience is only as good as the signal used to build it, and that signal begins with attribution. The Meta Pixel and Conversions API (CAPI) are complementary infrastructure: the Pixel fires from the browser; CAPI fires server-side, bypassing ad blockers, Safari's ITP, and iOS privacy restrictions.
Running only a Pixel means a meaningful portion of your purchase events are invisible to Meta. Industry data from 2026 puts the combined Pixel-plus-CAPI recovery at 20–30% more attributed conversions versus Pixel-only, with one illustrative example showing an account moving from 68 attributed purchases (Pixel-only) to 91 attributed purchases (Pixel + CAPI) from the same 100 actual orders.
For lookalike creation, this matters directly: every purchase event that goes unattributed is a buyer who does not appear in your pixel-based seed audience. A seed built from complete server-side data produces a more accurate lookalike model than one built from partial browser data alone. CAPI is not a campaign optimisation option — it is the foundation that makes lookalike audiences worth running at all in a post-iOS 14.5 environment.
Not sure if your Meta Pixel and CAPI are configured correctly? A misconfigured setup silently degrades every lookalike you run. Speak to Growth Pulse Media about a technical audit before your next campaign launches.
Why South African Businesses Choose Growth Pulse Media for Meta Ads
Dirk built and scaled a South African ecommerce operation before founding Growth Pulse Media — which means when we talk about audience signal quality, we are talking about lessons paid for from actual ad budgets, not borrowed frameworks from global case studies. We have configured lookalike workflows across Shopify, WooCommerce, and CRM-synced stores in the SA market, integrated Klaviyo list syncs with CAPI pipelines, and seen first-hand how a single seed audience quality improvement changes downstream campaign economics.
We run a limited client roster — each client gets senior attention, not a junior account manager following a templated playbook. If your Meta Ads account is spending real Rand and you are not certain whether your lookalike setup reflects the Advantage+ era, our Meta Ads management service starts with a full technical and structural audit. We do not pitch strategy before we understand your existing data quality — because all the targeting sophistication in the world produces noise if the seed is broken.
Who Should NOT Use Meta Lookalike Audiences as Their Primary Strategy
New accounts with fewer than 100 purchases: Meta's minimum technical threshold is 100 matched people, but the modelling quality below 1,000 is unreliable. If your account is new or your purchase volume is low, broad targeting with Advantage+ will typically serve you better than a lookalike built on thin data. Build the transaction history first.
Advertisers without CAPI implemented: A lookalike built entirely from Pixel data in 2026 is built on an incomplete picture of your iOS buyers — the majority of whom opted out of cross-app tracking. Running lookalike campaigns without CAPI means you are seeding Meta's algorithm with a distorted sample of your actual customer base. Fix the infrastructure before investing in audience sophistication.
Local service businesses with hyper-local geographic constraints: If your service area is a single suburb or small town, a 1% South Africa-wide lookalike may cover the right profile but entirely the wrong geography. For plumbers in Fourways or dentists in Stellenbosch, geographic targeting layered with interest signals often works better than a lookalike built from a national customer base.
Businesses with non-buyer seed lists only: If your CRM contains only newsletter subscribers with no purchase data attached, your lookalike seed is carrying weak signal. Meta will find people who look like your subscribers — not people who look like your buyers. Segment rigorously before uploading, or accept that the resulting audience will behave like a cold interest-targeting list, not a high-intent prospecting audience.
Know which category you're in — but not sure what to do next? Book a Meta Ads assessment with Growth Pulse Media and we will map the right audience architecture for your account size and data quality.
Frequently Asked Questions: Meta Lookalike Audiences
Do Meta lookalike audiences still work in 2026?
Yes — but with an important qualification. Meta lookalike audiences work best when seeded with high-quality first-party data (purchaser lists, high-LTV segments) and integrated into Advantage+ campaigns as audience suggestions rather than run as standalone manual-targeting campaigns. The gap between a generic pixel-based lookalike and broad Advantage+ targeting has narrowed since iOS 14.5, but a value-based lookalike built from your top customers still tends to carry stronger performance than broad targeting because it incorporates profit signal the algorithm cannot independently generate.
What is the minimum seed audience size for Meta lookalike audiences?
Meta's hard minimum is 100 matched people from a single country. In practice, reliable modelling starts at around 1,000 matched records, with 1,000–5,000 being the widely cited practitioner guideline for consistent results. Quality always outranks volume: a list of 500 verified purchasers will produce a stronger lookalike than a list of 5,000 unqualified contacts.
How does the Advantage+ Audience feature affect lookalike targeting?
Under Advantage+ Audience, your lookalike becomes a starting suggestion rather than a hard boundary. Meta's algorithm begins delivery within your lookalike population, then expands beyond it when it finds better-performing segments elsewhere in the eligible audience. This means your lookalike still contributes signal value — it just does not constrain delivery the way a manual audience used to. Running a quality lookalike inside Advantage+ tends to deliver stronger results than both a standalone lookalike and a fully broad Advantage+ campaign with no audience seed.
Which lookalike percentage should I use for South Africa?
For most South African advertisers, 1–3% is the practical range. South Africa's addressable Facebook audience sits at 27.9 million users, which means a 1% lookalike produces a prospecting pool of roughly 279,000 people — sufficient reach for most campaign objectives without sacrificing match quality. Scale to 3–5% once the 1% is profitable and you need broader volume. Lookalikes above 5% in the SA market start to approach the characteristics of broad targeting at meaningful budget levels.
Why does the Conversions API matter for lookalike audiences specifically?
The Conversions API supplements browser-based Pixel tracking with server-side data, recovering purchase events that iOS privacy settings and ad blockers prevent the Pixel from firing. Industry data shows Pixel-plus-CAPI setups attribute 20–30% more conversions than Pixel-only configurations. For lookalike creation, this matters because every unattributed purchase is a buyer missing from your seed — and a seed built on incomplete data produces a less accurate lookalike model. CAPI is the infrastructure that keeps your seed list complete.
Can I still use customer email lists as seed audiences for Meta lookalike audiences?
Yes. Customer email lists — especially when exported with phone numbers and addresses for higher match rates — remain one of the strongest seed sources available in 2026. Sync them via Klaviyo or direct CRM upload, and segment by purchase behaviour rather than uploading your full database. A purchaser segment of 1,000 people will typically carry stronger modelling signal than a large general subscriber list — Meta finds people who look like your buyers, not people who look like your sign-ups. Ensure your data collection is compliant with POPIA's lawful basis requirements before uploading to Meta's Custom Audiences tool.
Run Better Meta Ads in South Africa
Growth Pulse Media manages Meta Ads for South African businesses across Shopify, WooCommerce, and service industries. We build audience architecture from scratch — Custom Audiences, value-based lookalikes, CAPI configuration, and Advantage+ campaign structure — using your first-party data, not borrowed global templates. Klaviyo integrations, POPIA-compliant list management, and senior-level account attention on every engagement.
Request a Meta Ads audit — no obligation. We'll get back to you within 24 hours.
