Last month, PPC strategist Ana Kostic went semi-viral for a candid post explaining how following Meta's "best practice" recommendations cost her client 40% of their revenue. It hit a nerve because most advertisers already suspected it: the default Advantage+ setup, the broad audience gospel, the "just let the algorithm do its thing" advice — none of it is a strategy. It's an abdication.
At the same time, Google is quietly testing forcing searchers to sign in to see more results, Microsoft Clarity now separates branded from non-branded AI queries, and Meta's Data Manager API expansions are reshaping how audiences move between platforms. The signal is clear: targeting in 2026 is a data discipline, not a checkbox. This guide is the advanced playbook we use for both B2B and B2C clients at Omakaase — the layers, the exclusions, the creative-to-audience matching, and the mistakes to unlearn.
Why Facebook Ads Targeting Still Matters (Despite "Broad Everything")
Meta has spent three years telling advertisers that broad targeting plus Advantage+ Shopping is the future. And for some accounts — high-volume DTC ecommerce with 50+ purchases a week — it genuinely is. But that advice has been misapplied to virtually every other business type, and the results speak for themselves.
Here's the nuance the platform reps skip: Meta's algorithm optimises toward the conversion event you feed it, using the signals you send, against the audience you allow. If your pixel fires on low-intent events, if your first-party data is thin, or if your audience pool includes segments that will never buy, broad targeting simply finds the cheapest impressions — not the most valuable customers.
Advanced targeting isn't about narrowing to 5,000 people. It's about giving the algorithm better inputs, cleaner exclusions, and creative that matches audience intent. That's where B2B and B2C strategies diverge sharply.
The Advanced B2C Targeting Framework
For B2C — especially DTC ecommerce, subscription, and local service — the modern stack is layered, not siloed. We build campaigns in three concentric rings.
Ring 1: First-Party Data Retargeting
This is your revenue engine. Upload every legitimate customer touchpoint: purchasers (last 180 days), high-LTV customers, cart abandoners, email subscribers who opened in the last 30 days, and site visitors who hit key pages. With iOS signal loss, uploading hashed customer lists via the Conversions API and Data Manager is now non-negotiable. If you're still relying on the pixel alone, you're feeding the algorithm about 60% of the truth.
Segment these lists by value tier. A repeat purchaser deserves a different creative and offer than someone who bought once 90 days ago. Build a first-party data strategy before you touch a single ad.
Ring 2: Engaged Lookalikes
Skip the "1% lookalike of all purchasers" default. Instead, build lookalikes off your top 25% LTV customers or your 90-day repeat purchasers. Meta's algorithm is only as good as the seed. A lookalike of 10,000 low-margin buyers will produce more of the same.
Test 1–3% and 3–5% ranges separately. In August 2026 conditions, we're seeing 3–5% often outperform 1% because iOS attribution gaps make the tightest lookalikes over-index on retargeted users.
Ring 3: Interest + Behaviour Stacks with Smart Exclusions
This is where most advertisers stop. Don't. The rising Google Trends topics — home workout routines, learn programming, book recommendations, popular movies 2025 — are useful because they reveal behavioural clusters, not just interests. A user searching "home workout routines" in August likely intersects with fitness apparel, supplements, wellness content, and streaming subscriptions. Stack interests plus behaviours (Engaged Shoppers, mobile device users, frequent travellers) and always exclude existing customers to prevent budget cannibalisation.
The Advanced B2B Targeting Framework
B2B on Facebook is where the biggest disagreements happen. "Use LinkedIn" is the reflexive answer — and yes, our LinkedIn Ads B2B guide is the primary reference for many campaigns. But Facebook's CPMs are often 60–80% cheaper, and buying committees consume content on both platforms. Written off Meta, you're leaving pipeline on the table.
Job Title Targeting Is Gone — Use Behavioural Proxies
Meta removed most detailed employment targeting years ago. The workaround is behavioural proxying:
- Custom audiences from LinkedIn ad engagers (via UTM matching in your CRM, then uploaded)
- Website visitors segmented by URL patterns (pricing pages, case studies, docs)
- Video view audiences from thought-leadership content that only decision-makers would watch to 75%+
- Lead form completions imported via the Conversions API
Account-Based Layering
For enterprise or mid-market B2B, upload your target account list — company email domains of employees at those accounts, LinkedIn-scraped contacts (compliantly), and event attendee lists. Combine with an account-based marketing motion so ads reinforce outbound and vice versa.
Intent Signals via Content Consumption
Build audiences from users who watched 50%+ of your longer thought-leadership videos or spent 3+ minutes on high-intent blog content. These are your MQL equivalents inside Meta. Retarget them with case studies, ROI calculators, and demo offers. Read our B2B lead generation guide for the funnel logic behind this.
How AI Visibility Is Reshaping Targeting Inputs
Here's a shift most advertisers haven't priced in yet: AI visibility is now a leading indicator of PPC performance. When ChatGPT, Perplexity, or Google's AI Overviews cite your brand, downstream branded search rises, and Meta's algorithm sees stronger conversion signal from users who arrive already familiar with you.
Microsoft Clarity's new branded vs. non-branded AI query split lets you actually measure this. Users referred from AI surfaces convert at 2–4x the rate of cold Facebook traffic in our client data. So your targeting inputs should now include an audience of "AI-referred visitors" — segment them in GA4, push to Meta via the Data Manager API, and build lookalikes off that seed. It's the highest-quality audience most brands aren't building yet. Our AI marketing stack for 2026 covers the plumbing.
Case Study: How a UK Skincare Brand Cut CAC by 47%
A mid-sized UK skincare brand (£4M annual revenue, DTC) came to us in Q1 2026 with a familiar problem: Advantage+ Shopping campaigns had scaled well through 2025, then plateaued. CAC had risen from £22 to £41 in six months. They'd been told by their previous agency to "trust the algorithm" and increase budget.
Our audit found three issues:
- Their purchaser lookalike was seeded off all customers, including a large cohort of discount-hunters from a viral TikTok moment.
- They had no exclusion for existing customers on prospecting campaigns — 34% of "new customer" attributed spend was actually retargeting.
- Their Conversions API was misconfigured, sending only 41% of purchase events server-side.
We rebuilt around three rings: first-party data via a fixed CAPI setup, a lookalike seeded only on 2+ purchase customers, and behavioural stacks aligned to their content pillars (clean beauty, sensitive skin). Within 60 days, CAC dropped from £41 to £22, ROAS climbed from 1.9x to 3.4x, and — critically — 60-day repeat purchase rate on new customers rose by 18%. Better targeting brought better customers, not just cheaper ones.
Common Mistakes That Quietly Kill Performance
- Trusting Advantage+ defaults without exclusions. If you're not excluding recent purchasers, employees, and low-LTV segments, you're overpaying for your own audience.
- Seeding lookalikes off all-purchasers. Quality in, quality out. Segment by LTV, frequency, or margin before you seed.
- Ignoring the Conversions API. Pixel-only tracking in 2026 loses 30–50% of events. Full CAPI (or Data Manager API) is table stakes.
- Broad targeting with weak creative. Broad only works when creative does the segmentation. If your ad speaks to everyone, it converts no one.
- Confusing audience size with audience quality. A 500,000-person audience of the right people beats 5 million of the wrong ones every time.
- Never refreshing exclusions. Customer lists, employee lists, and 180-day purchaser windows all decay. Update monthly.
- Following "best practice" blindly. The Ana Kostic story is a warning. Best practice is a starting point, not a strategy.
Putting It Together: The 2026 Targeting Stack
The advertisers winning right now are running what we call a modular stack: three to five campaigns, each with a distinct job (prospecting, warm retargeting, cart recovery, existing customer LTV expansion, lookalike scaling), each with tight exclusions to prevent overlap. They're feeding the algorithm strong first-party signal, refreshing audiences monthly, and matching creative variants to audience intent — not showing the same ad to a first-time visitor and a repeat buyer.
If you're serious about scaling Meta spend without watching CAC drift upward every quarter, the fundamentals in our Facebook Ads guide for 2026 and Meta Ads guide pair well with this playbook. And if you're weighing paid social against other channels, our breakdown of paid vs. social media marketing is worth ten minutes.
Ready to Rebuild Your Meta Targeting?
Facebook Ads targeting in 2026 rewards discipline, first-party data, and the confidence to ignore reflexive "best practice" advice when your account data says otherwise. If your CAC is drifting or your ROAS has plateaued, it's almost never a creative problem alone — it's a targeting architecture problem.
Our team audits Meta accounts against this exact framework. Build a proposal with Omakaase and we'll show you where the leaks are — and what a rebuilt targeting stack could unlock over the next 90 days.