Interest targeting on Meta is functionally dead for scaled DTC accounts in 2026 because Andromeda's broad-targeting signal already incorporates everything the detailed-targeting layer used to add. But interests still beat broad in three specific account types: niche B2B products below five hundred qualified customers in the lookalike seed, local-service businesses with a geographic constraint tighter than a metro area, and categories where the buyer self-identifies strongly with a specific interest cluster (gaming hardware, home brewing, specialty cycling). The blanket advice to abandon interests misses these cases and costs real performance in accounts that still need them.
Interest targeting is dead for scaled DTC but alive in three specific account types: niche B2B with small seed audiences, hyper-local service businesses, and strong-identity categories. The 'broad wins everything' rule is right for roughly 80 percent of accounts and actively wrong for the other 20. Classify the account against the exceptions before switching.
Why is interest targeting dead for scaled DTC in 2026?
Interest targeting is dead for scaled DTC because Andromeda — Meta's current ranking and delivery system — already incorporates the behavioural, demographic, and page-engagement signals that detailed-targeting categories used to represent. Adding an interest layer on top of a broad audience in a scaled DTC account now restricts the delivery system from exploring segments that the interest filter excludes but which Andromeda would otherwise identify as high-potential buyers. In accounts above roughly $5k per month in spend with addressable audiences over a hundred thousand people, broad targeting consistently outperforms interest-stacked ad sets on both CPA and incremental ROAS. This is one of the reasons broad targeting dominates for scaled DTC accounts in 2026 and the default guidance has shifted so sharply.
Where did the broad-wins-everything rule actually come from?
The rule emerged from large account tests across US and EU DTC between 2022 and 2024 as iOS signal loss forced Meta to rebuild its optimisation surface. Broad audiences gave Meta more exploration room to compensate for degraded tracking data. By 2025, major agencies and performance marketing creators had generalised the finding into a universal recommendation. The finding is accurate for the account class the tests ran on — scaled consumer brands with large addressable audiences and healthy conversion signal. The generalisation to every Meta account is where the error entered. Smaller accounts, niche categories, and local businesses were not represented in the test base and the broad-wins rule does not transfer cleanly to them.
When does interest targeting still beat broad on Meta?
Interest targeting still beats broad in three specific account types. First, niche B2B products with fewer than 500 qualified customers in the lookalike seed — the seed is too small for Meta to build a reliable lookalike, broad does not have enough category signal to find the right accounts, and interest clusters (specific job titles, software categories, industry publications) narrow the pool to something Meta can optimise against. Second, hyper-local service businesses with a geographic constraint tighter than a metro area — the geo filter already narrows the audience to the point where additional broad reach wastes impressions on people outside the service zone. Third, strong-identity categories where the buyer self-identifies with the interest (gaming hardware, home brewing, specialty cycling, specific hobby crafts) — the interest filter captures a behavioural segment broad cannot infer from browsing alone.
Why do niche B2B accounts need interests when broad fails?
Niche B2B accounts need interests because the broad targeting pool contains too many consumer-grade buyers for Meta's optimisation to find the handful of qualified B2B accounts inside it. A product selling to pharmaceutical manufacturing operations has maybe two thousand qualifying companies globally and perhaps ten thousand decision-makers inside them. Broad targeting against this audience drowns the signal in consumer noise. An interest layer narrowing to "pharmaceutical industry," "manufacturing operations," or specific industry publications gives Meta a smaller but more concentrated pool to optimise against. Meta's B2B CPM at around $4 versus LinkedIn's $25-60 only becomes a real cost advantage when the targeting narrows enough to deliver qualified impressions, and interests are often the only lever that produces that narrowing without destroying scale.
How do you tell whether your account is a broad-wins or interest-still-wins case?
1) Count your addressable audience under broad targeting with only the required filters (geo, age, language). If below 100,000, interests may still win. 2) Measure the ratio of qualified-to-unqualified buyers in your addressable audience. Niche B2B typically runs 1:100 or worse; interests needed. 3) Check whether your product category has a strong self-identification pattern (hobby, profession, lifestyle). If buyers describe themselves by the category, interests capture signal broad misses. 4) Measure the geographic concentration of past buyers. Sub-metro concentration above 70 percent is a hyper-local signal. 5) Run a four-week paired test with one broad ad set and one interest-stacked ad set at matched budget. If interests do not beat broad on blended CPA within four weeks, abandon them permanently.
What is the practical rule for 2026?
Default to broad for scaled DTC accounts above $5k monthly campaign spend with addressable audiences over 100,000. Default to interests for niche B2B with small seeds, hyper-local service businesses, and strong-identity hobby or profession categories. Run the four-week paired test when the account is ambiguous. Rerun quarterly as the account scales, because an interest-wins account at $3k monthly spend often crosses into broad-wins territory by $8k. The rule is spend-tier and category dependent, not philosophical.
Conclusion
Interest targeting is not universally dead, it is category-dead. For scaled DTC accounts above a hundred thousand monthly addressable buyers, broad wins and interests cost signal. For niche B2B, hyper-local, and strong-identity categories, interests still provide the specificity Andromeda cannot infer from broad behaviour. The practical next step is to classify the account against the three exception types before accepting the default advice to abandon detailed targeting entirely. In Advize's experience, roughly two in ten accounts fall into one of the exceptions and are being under-served by the broad-wins-everything rule.