Broad vs Detailed Targeting on Meta in 2026

There is a fight that splits every Meta advertiser into two camps. One side swears by precise interest stacks: layer three behaviors, two job titles, a lookalike, and you have a clean audience. The other side throws everything open, lets the algorithm roam, and trusts the machine. For years the precise camp won the argument by default, because controlling who saw your ad felt like the whole job. That instinct is now wrong more often than it is right. This article takes a side, explains why, and tells you the exact cases where the narrow approach still beats the broad one in 2026.

What broad and detailed targeting actually mean now

Detailed targeting is the old Ads Manager box where you type interests, behaviors and demographics. You tell Meta to show your ad to people who like CrossFit, follow vegan recipes, or work in finance. Broad targeting, in its purest form, means you hand Meta almost nothing beyond a country, an age floor and a budget, then let the system find buyers. Between those two poles sits Advantage+ Audience, Meta’s AI targeting layer that became the default for new campaigns in February 2026. It is not pure broad and it is not detailed targeting. It is a hybrid where your inputs become hints, not walls.

This distinction matters because most people argue about a version of detailed targeting that no longer exists. In the classic setup, an interest was a hard filter: select it, and Meta only showed your ad to that pool. Today, inside Advantage+ Audience, the same interest is a suggestion. Meta starts there, then expands the moment it spots better buyers elsewhere. So when someone says broad beat their interest stack, they often mean the AI version quietly outgrew the box they thought was protecting them. The debate is less broad versus detailed and more your judgment versus the model’s judgment.

Meta’s own help center spells out the split. Inside Advantage+ Audience you have two layers: controls and suggestions. Controls are hard rules: minimum age, location, excluded custom audiences, language. Suggestions are soft guides: detailed targeting, lookalikes, custom audiences as a starting point, age and gender preferences. The system respects controls absolutely and treats suggestions as a launchpad. If people you never selected start converting, Meta serves them anyway. Understanding which lever is a wall and which is a hint is the single most useful thing an advertiser can learn in 2026.

The June 2025 consolidation changed the terms of the debate

On June 23, 2025, Meta began consolidating detailed targeting in Ads Manager, then extended it to Meta Business Suite and boosted posts on August 21, 2025. The change was not cosmetic. Hyper specific interests were merged into broad buckets. Progressive House folded into House Music. Niche metal subgenres and named band fan bases collapsed into Heavy Metal. EDM fans, SUV owners and vegan food all dissolved into wider groups. If your strategy depended on slicing an audience thin, the knife you used got blunter overnight, on purpose.

Meta also killed detailed targeting exclusions. Since March 31, 2025 in Ads Manager and June 10, 2025 for boosted posts, you can no longer exclude people by interest, so no more cutting out classical music fans or family vacation seekers. Meta justified this with its own data: campaigns without exclusions showed a median cost per conversion 22.6% lower. That number is the whole argument in one figure. The exclusions advertisers fought to keep were, on average, making their results worse. Removing the option forced a behavior the data already favored.

There was a hard deadline too. Campaigns built before June 23, 2025 could keep running on the old options until January 15, 2026. After that date, any campaign still leaning on deprecated targeting simply stopped delivering. So this was not a soft nudge. Meta set a date, let legacy campaigns coast for six months, then pulled the floor out. Anyone who ignored the warning watched evergreen winners go dark in mid January. If you still have campaigns that have not been rebuilt around Advantage+ Audience, that is the first thing to check after reading this.

Why the machine usually wins now

The reason broad keeps winning is not ideology, it is plumbing. Meta’s targeting engine runs on Andromeda, a deep learning architecture rolled out in late 2024. By 2026 it weighs more than 10,000 signals per impression to decide whether to show your ad to a given person at a given moment. No human stack of three interests competes with 10,000 live signals. When you hand pick interests, you are guessing at a pattern the model can already see in the raw behavior of your past buyers. You are throwing away resolution to feel in control.

There is a second, quieter reason the machine wins: declared interests were always a weak proxy for intent. Liking a CrossFit page tells Meta you once tapped a button, not that you will buy protein powder this week. Andromeda watches what people actually do, which pages they linger on, what they bought last month, which ads they scrolled past. Behavior beats declaration every time. The interest box rewarded advertisers who guessed well at correlations; the model now measures those correlations directly across billions of events. You were never competing with the platform on equal footing, you were competing with a fraction of its data.

Apple’s App Tracking Transparency framework, shipped with iOS 14.5, made this worse for the precise camp. It cut the data pipeline Meta used to build granular profiles. Interest categories that leaned on third party behavioral data became less reliable, so the very inputs detailed targeting depended on degraded at the source. Meanwhile the Pixel and the Conversions API kept feeding the model first party conversion signals, which is exactly the fuel a broad AI campaign runs on. The platform shifted value away from declared interests and toward observed behavior, and it did so deliberately.

The numbers back the shift. Meta reports a 14.8% lower cost per result on awareness campaigns and 7.2% lower on sales campaigns when advertisers use Advantage+ audience features. On product catalog sales, Advantage+ Audience showed a 13% lower median cost, a 7% lower median cost per website conversion, and a 28% lower average cost per click, lead or landing page view. In its Q1 2025 earnings call, Meta said advertisers see 4.52 dollars in revenue per dollar spent with Advantage+ campaigns, roughly 22% higher than manually managed ones. These are Meta’s own figures, so read them as a vendor would present them, but the direction is consistent across independent tests too.

Independent tests tell the same story without the marketing gloss. Top Growth Marketing ran Advantage+ Shopping against manual campaigns during Black Friday 2024 and saw 3.14 ROAS versus 2.70, about a 16% lift. The analytics firm Lebesgue reported broad targeting carrying a 113% ROAS index against 76% for lookalikes, with comparable click through rates near 1%. Different methodologies, same conclusion: when conversion volume is healthy, handing the audience decision to the model beats hand built audiences in the median case. The argument for detailed targeting has to live in the exceptions, not the average.

Real cases where broad paid off

Specifics beat slogans, so here are documented outcomes. Ray Ban improved its Advantage+ sales campaign by adding value optimization and saw a 9% rise in ROAS alongside a 32% jump in average order value, the kind of result that comes from letting the system chase high value buyers instead of a fixed interest list. FULLBEAUTY Brands, a fashion retailer, adopted AI generated creative variations through Advantage+ Shopping in 2025 and reported a 45% jump in ROAS, a 22% rise in conversion rate and a 36% lift in click through rate. Neither outcome came from a clever interest stack. They came from broad reach plus strong creative plus value signals.

The pattern repeats at smaller scale. A jewelry brand running Meta hit a 39.51% increase in ROAS within 30 days, a 33% lift in store conversion rate and an 18% rise in ecommerce revenue month over month, after consolidating spend into broad campaigns with better creative. A high ticket home improvement retailer moved ROAS from 1.18 to 6.47 and cleared more than 700,000 dollars in purchase value once the account stopped fragmenting budget across narrow ad sets and let the algorithm pool its learning. The common thread is consolidation: fewer, broader, better fed campaigns out earning many small precise ones.

It helps to picture the failure mode these brands escaped. Imagine an account with fifteen ad sets, each carrying its own narrow interest stack, each splitting a 50 dollar daily budget into slivers. No single ad set ever clears the volume the model needs to exit the learning phase, so every one of them stays jittery and expensive. Consolidate those fifteen into two broad campaigns and the same budget suddenly feeds enough conversions to stabilize. The brands above did not find better audiences, they stopped starving the algorithm. Fragmentation, not the wrong interest, is what quietly drains most underperforming accounts.

Notice what these cases share. None of them is a story about finding a magic interest. They are stories about giving one campaign enough volume to learn, pairing it with creative that does the persuasion work, and feeding clean conversion data. That is the real shift. In the detailed targeting era, the lever you pulled hardest was the audience. In the broad era, the audience is mostly automatic, and your leverage moves to creative, offer and signal quality. Advertisers who keep obsessing over the audience box are tuning the part of the machine that matters least.

This reframes what creative is for. In a broad campaign, your creative is the targeting. A video that opens on a runner lacing trail shoes will self select runners far more reliably than any interest checkbox, because the people who stop scrolling tell Meta who they are. Strong hooks, clear product shots and a specific offer do double duty: they persuade, and they signal audience fit back to the model. So the energy you used to pour into building three layered ad sets now goes into shipping more creative variations. The advertisers winning in 2026 test angles and hooks, not interests.

Two myths the broad camp gets wrong too

Going broad is not the same as going lazy, and the popular advice often blurs that. Myth one: smaller audiences are always safer because you waste less. The data says the opposite at scale. The exclusions advertisers used to feel safe cost them 22.6% in median cost per conversion. A tiny audience does not protect your budget, it starves the algorithm of the volume it needs to find your real buyers. Below roughly 50 conversions per week, the model cannot learn cleanly, learning phases drag on, and results swing wildly. Small can be the riskier choice, not the safe one.

Myth two: broad means you give Meta nothing. That is the version of broad that genuinely backfires. With no Pixel, no Conversions API, weak creative and no value signals, broad targeting just spends fast against a vague objective and burns budget. Broad targeting fails when the signal is poor, not because the idea is wrong. The model needs a clear optimization event, clean conversion data and creative that filters the audience for you. Broad without strong inputs is not trusting the machine, it is asking a powerful engine to drive blind. The camp that screams just go broad rarely says this part out loud.

When detailed targeting still wins

Here is where this article refuses the lazy take. Broad does not win everywhere, and pretending it does is its own cliche. Detailed targeting still earns its place in a precise set of situations: genuinely small, well defined audiences that broad exploration would dilute rather than discover. Think ultra niche B2B, like CFOs at SaaS companies with 50 to 200 employees using a specific tool. The total addressable audience is tiny. Letting the AI expand from that seed mostly means spending on the wrong people. A focused job title and company size filter, layered with a custom audience, can still beat broad here.

Tiny budgets are the second case. Campaigns spending under 20 to 30 dollars a day rarely generate enough conversions to feed the broad model, so a focused starting audience often does better than open exploration on a shoestring. New accounts with no Pixel history sit in the same boat: with no conversion data to learn from, a sensible interest seed gives Meta a direction it otherwise lacks. The broad engine is hungry. If you cannot feed it volume, a narrower, more relevant pool can squeeze more out of the few impressions you can afford. This is not nostalgia, it is matching the tool to the constraint.

Tight local service areas are the third case, with a caveat. A plumber serving three postal codes does not want the AI exploring a whole region in search of cheaper clicks that never convert into a callout. Strong location controls plus a focused audience used to be the only safe play. The caveat: Advantage+ now handles location constraints far better than it did a year ago, because location is a hard control, not a soft suggestion. So even here, the move in 2026 is often broad inside a tight geofence rather than pure detailed targeting. The geography stays locked, the audience stays open.

A pottery studio in Lyon offering weekend classes is a clean example. Its market is a 20 kilometer radius and people who might book a creative weekend. Pure broad across France would waste most of the budget. The 2026 answer is a locked radius as a control, a broad audience inside it, and a custom audience of past bookers as a suggestion to steer the model. That is neither old school detailed targeting nor naive broad. It is the hybrid Meta actually wants you to run, and it respects the genuine constraint of a tiny local market without throwing away the engine’s intelligence.

There is a fourth, easily missed case: brand new product categories with no behavioral history. If you launch something the platform has never seen people buy, like a genuinely novel gadget, the model has no past purchases to pattern match against, so its early broad exploration is closer to a guess. Seeding it with a defensible interest, your closest adjacent category, gives the algorithm a credible starting hypothesis while it gathers real conversion data. Once enough purchases stack up, you loosen the seed and let broad take over. Detailed targeting here is training wheels, not the destination, and you should plan to remove them.

The 2026 verdict and how to run it

The honest verdict is not broad wins, full stop. It is this: broad through Advantage+ Audience is the correct default for the large majority of accounts in 2026, and detailed targeting is now a specialist tool for small, well defined, low volume or hyper local situations. The June 2025 consolidation did not start this trend, it ratified it. Meta removed the granular knobs because its own data showed they hurt more than they helped at scale. Fighting that with ever more elaborate interest stacks is rebuilding a road the platform already demolished on purpose.

Be skeptical of where the numbers come from, though. Most of the headline lifts in this article are Meta’s own figures or vendor case studies, and both have an obvious interest in making automation look good. A 22% revenue advantage announced on an earnings call is a marketing claim wearing a finance suit. That does not make it false, but it means you should treat it as a hypothesis to test on your own account, not a law. The right posture is neither cynicism nor faith. Run a clean holdout, compare broad against your best detailed setup on your own data, and let your account, not a press release, settle the argument.

In practice, most mature accounts in 2026 run a portfolio rather than a religion. A common split puts 70 to 80% of budget on broad Advantage+ campaigns, 10 to 20% on retargeting warm audiences, and 5 to 10% on lookalike or interest seeded tests. The big broad block does the scaling. Retargeting catches the people already close to buying. The small test block is where detailed targeting still earns a seat, as a controlled experiment rather than the backbone. This structure treats broad as the engine and detailed targeting as a scalpel you reach for deliberately.

One practical warning before you flip everything to broad: do not judge it in week one. A fresh broad campaign needs roughly 50 conversions a week to exit the learning phase, and until it does, the early numbers lie. Advertisers panic at a bad day three, switch back to a narrow audience, reset the learning, and conclude broad does not work, when they simply never let it learn. Give a broad campaign a full learning window with a stable budget and untouched settings before you read its verdict. Patience is part of the strategy, not a soft skill bolted on.

So pick your side with eyes open. If you run a typical ecommerce or lead gen account with real conversion volume, default to broad, invest your energy in creative and signal quality, and stop manicuring interests. If you sit in a genuinely narrow niche, on a tiny budget, with a cold account or a tight service radius, detailed targeting still earns its keep, ideally as a hard locked control wrapped around an otherwise open audience. The mistake is not choosing broad or detailed. The mistake is choosing one out of habit instead of matching it to your volume, your data and your market.

Sources

Meta Business Help Center, Updates to Detailed Targeting and About Advantage+ Audience (audience controls vs suggestions). Meta for Business, Advantage+ Audience product page. Meta Q1 2025 earnings call (4.52 dollars revenue per dollar, Advantage+ performance). Social Media Today, Meta Is Consolidating More of Its Detailed Ad Targeting Options. Jon Loomer Digital, Detailed Targeting Announcement and A Guide to Meta Ads Targeting in 2026. Lebesgue, Broad or Lookalike Audience, Targeting in 2025. Top Growth Marketing Black Friday 2024 Advantage+ test. Shoelace jewelry brand case study (39.51% ROAS). Ray Ban Meta Advantage+ case study. FULLBEAUTY Brands Advantage+ Shopping case study. UM.marketing home improvement retailer case study. Conversios and Adligator 2026 targeting guides.