Meta Detailed Targeting Removed 2026

On 15 January 2026 Meta stopped delivering ads that rely on deprecated detailed targeting options across Facebook and Instagram. That is not a UI change or a tooltip warning. Existing ad sets built on retired interests are now serving to a much smaller pool, and in some accounts, not serving at all. If your Meta CPA jumped in the second half of January and you couldn't figure out why — this is the reason. This piece is the playbook we run on every coaching account under management.

What Meta actually changed on 15 January 2026

This is the final enforcement step of Project Andromeda, the algorithm rebuild Meta fully rolled out by October 2025. Andromeda replaced the audience-based delivery model with a creative-based one. The algorithm now reads the ad creative — visual format, hook, tone, subject matter — and predicts the right audience from behavioural signals, rather than delivering to the audience the advertiser defined.

The consolidation is granular. Meta did not delete interest targeting outright. What it did was collapse thousands of narrow interests into fewer broad clusters:

• Individual musicians and TV shows became broader genres (rock music, reality TV).
• Specific food items (sushi, vegan desserts) rolled up into food and drink.
• Author-name interests, niche podcast interests, and low-audience-size book interests were retired.
• Under Advantage+ Audience, any detailed targeting you enter is treated as a suggestion, not a fence — Meta can and will spend outside it if it predicts better performance.

In practice, the interests coaches used to stack (self-help authors, specific business books, niche podcast hosts, competing coach names) are the exact category that got retired. That is why coaching accounts are seeing this harder than DTC accounts running on Advantage+ Shopping.

Why this hits coaches harder than e-commerce

E-commerce brands under $10M in revenue have been running Advantage+ Shopping Campaigns with broad targeting since 2023. Coaches largely have not. The typical coaching ad account we inherit looks like this:

• One traffic or engagement campaign, one ad set, 8-14 stacked interests (Alex Hormozi, Russell Brunson, GoHighLevel, Kajabi, Skool, and the coach's own competitors).
• One video ad, one image ad, one carousel, all rotating.
• No first-party pixel data older than 60 days (because the offer changed).
• No CAPI, or a broken CAPI that dedupes wrong.

Every single element of that setup is the wrong side of Andromeda. Narrow interests are the retired category. Signal is the compensation. Coaches typically have neither.

The broad targeting playbook — coaching-specific ad structure

The playbook has three layers. They are ordered by impact — do not skip layer 1 to obsess over layer 2. Every account we run under The Community Flywheel™ ships all three.

Layer 1 — Audience settings that survive Andromeda

Stop stacking interests. The audience settings that consistently spend well post-Andromeda across coaching accounts:

• Country: single country per ad set (US, UK, CA, AU). Do not stack them — CPMs and buyer economics differ too much.
• Age: 25-55 as a starting default for high-ticket coaching, 22-45 for cohort-based courses. Wider if the offer is universal (fitness, sleep, general business).
• Gender: All, unless the offer is gendered.
• Detailed targeting: leave empty. If you must add one interest for board approval, add exactly one broad cluster (Small Business Owner, Entrepreneurship) and turn Advantage+ Audience ON so Meta can spend outside it.
• Placements: Advantage+ Placements. Do not manually exclude anything.

Layer 2 — Creative concepts that self-select your buyer

Under Andromeda the creative is your targeting. If the hook, on-screen text, and pain point in the first three seconds do not name your buyer, Meta cannot classify them for you. For a coaching offer this means running three problem-first concepts per ad set, refreshed weekly:

• Concept A — Pain callout: 15-30 second UGC video where the coach names the exact operator pain in the first three seconds ("You've spent $2,000 on Skool ads with no members").
• Concept B — Contrarian POV: 30-45 second talking head calling out the widely-accepted-but-wrong tactic in the niche.
• Concept C — Proof reel: 45-60 second sequence of member wins, screen-recorded, with dates and dollar amounts on screen. No music beds louder than the voiceover.

Never launch fewer than three concepts. Andromeda needs comparative signal to allocate delivery. One creative gives it nothing to choose against and delivery gets stuck. For the weekly refresh cadence and testing structure that keeps this system fed, see <a href='/blog/meta-ads-creative-testing-2026'>Meta ads creative testing in 2026</a>.

Layer 3 — First-party signal (CAPI + audiences)

First-party data is the input Andromeda scores your account on. Coaches without CAPI are being under-delivered because the algorithm has no signal to weigh against. The minimum acceptable stack:

• CAPI firing on Lead, Complete Registration, Schedule (or Book), and Purchase — with event dedup between browser pixel and CAPI.
• Weekly export of your buyer list into a Custom Audience (email + phone hashed).
• A value-based Lookalike (1%) seeded off buyers, weighted by revenue.
• A 30-day landing page visitor custom audience — used only for suppression on cold prospecting, not as a target.

Budget allocation model — 30 / 50 / 20

For accounts spending $3,000 to $30,000 per month on Meta, this is the split we ship as the default:

• 30% — Broad prospecting (single-interest or no-interest, Advantage+ Audience on, three problem-first concepts).
• 50% — Value-based Lookalike prospecting (1% off buyer list, weighted by revenue, same three creatives).
• 20% — First-party retargeting (30-day landing page visitors, 14-day opt-in-no-book, buyer list suppression on).

This is not a fixed ratio for life. It is the split that lets Andromeda see enough signal in the first 14 days without starving retargeting. Once your buyer list crosses about 300 hashed rows, we shift to 20 / 60 / 20 (heavier Lookalike weight). If your funnel has a paid challenge front-end, the 30% broad line moves to a challenge landing page rather than a discovery-call page — the challenge is the top of your Community Flywheel™ and pixel data compounds fastest there.

A week-1 launch checklist for coach advertisers

1. Audit every existing ad set. Any set with more than one interest stacked — pause it.
2. Rebuild your top-performing offer into three problem-first creative concepts (Pain, POV, Proof).
3. Wire CAPI on all four events. Verify dedup in Events Manager (dedup rate should sit above 90%).
4. Export your buyer list (email + phone) and build a value-based Lookalike 1%.
5. Launch three ad sets on the 30 / 50 / 20 split with a $50-$100 daily test budget per set.
6. Do not touch the campaign for 7 days. Andromeda's learning phase requires uninterrupted delivery.
7. On day 7, kill the bottom concept per ad set. Replace only that concept. Never rebuild the whole ad set.

What to stop doing today

• Interest stacking. Six interests per ad set is not a strategy — it is legacy 2019 muscle memory.
• Duplicating winners into new ad sets to force fresh learning. Andromeda accounts audience overlap across ad sets and penalises delivery.
• Manual placement exclusions. If Reels gave you a bad three-day window, that is creative fatigue, not a Reels problem.
• Rebuilding CBOs weekly. Every rebuild resets the learning Andromeda needed to score the account.
• Sending cold ad traffic to a signup page you don't own. See <a href='/blog/meta-ads-skool-why-fail'>why direct-to-signup Meta ads fail</a> for the pixel argument — Andromeda makes that penalty worse, not better.

The uncomfortable truth Andromeda forces on coaches: the era where a niche interest stack could compensate for weak creative and no first-party data is over. What replaces it is what the Acquisition Genesis Playbook has argued for since day one — an offer clear enough that the creative can carry the targeting, a landing page you control, and a pixel that actually fires.