One of the most common questions we get at RocketShip HQ from app marketers is whether they should still use interest targeting on Meta or just go broad. Having managed over $100M in mobile ad spend across hundreds of app campaigns, we've watched this question evolve dramatically since Apple's ATT rollout. The short answer: broad targeting wins more often than not in a post-ATT world, but interest targeting still has specific, valuable use cases that smart marketers shouldn't ignore entirely.
Page Contents
- Does broad targeting actually outperform interest targeting on Meta for mobile app campaigns?
- Why does Meta's algorithm perform better with broad targeting after iOS 14.5?
- When does interest targeting still make sense on Meta?
- How should I structure a test between broad and interest targeting on Meta?
- Does broad targeting work differently across Meta's placements like Reels vs. Feed?
- What about Advantage+ app campaigns vs. manual broad targeting?
- How much budget do I need before broad targeting becomes effective?
- Should I use lookalike audiences instead of broad or interest targeting?
- Related Reading
Does broad targeting actually outperform interest targeting on Meta for mobile app campaigns?
In our experience running campaigns across dozens of B2C apps, broad targeting outperforms interest targeting roughly 70-80% of the time in post-ATT environments. The performance gap has widened since iOS 14.5 because Meta's algorithm lost much of the granular signal that made interest targeting precise. Broad targeting gives the algorithm more room to explore and find high-value users the interest graph might miss.
The reason is structural. Post-ATT, Meta lost access to significant purchase and conversion data, which means the behavioral signals that once powered interest segments are less reliable. When you constrain the algorithm to a smaller, less accurate interest pool, you're compounding data loss with artificial restrictions. Broad targeting sidesteps this problem by letting Meta's machine learning optimize across the full available inventory.
- Broad campaigns typically see 15-30% lower CPAs compared to interest-targeted campaigns for app install objectives
- Interest targeting restricts Meta's auction liquidity, often increasing CPMs by 10-25%
- Broad targeting performs best when paired with strong creative that effectively self-selects the right audience
- The performance gap is most pronounced in English-speaking markets where ATT opt-in rates are lowest (around 25-35%)
Why does Meta's algorithm perform better with broad targeting after iOS 14.5?
Meta's algorithm uses a Bayesian Bandits (explore-exploit) framework to allocate spend. With broad targeting, the algorithm has a much larger pool to explore, which means it can find pockets of high-performing users faster and then exploit those signals at scale. Interest targeting constrains this exploration phase artificially.
How Meta's Spend Allocation Actually Works
As explained in detail in this breakdown of Meta's Bayesian Bandits approach, the algorithm constantly balances exploring new audiences with exploiting proven ones. When you use broad targeting, the algorithm has more "space" to explore. It can test impressions across a wider variety of user profiles, gather conversion signals, and then concentrate spend on the profiles converting best. Interest targeting front-loads a human assumption about who will convert, and post-ATT, those assumptions are often wrong.
The Signal Loss Problem
Before ATT, interest targeting worked because Meta had deep behavioral data: purchase history, app usage, cross-site browsing. With roughly 75% of iOS users opting out of tracking, those interest segments are built on increasingly stale or incomplete data. Broad targeting lets the algorithm rely on real-time conversion signals from your own campaign rather than degraded behavioral profiles.
When does interest targeting still make sense on Meta?
Interest targeting still has value in three specific scenarios: when you're launching a brand-new app with zero pixel or SDK data, when you're targeting a genuinely niche audience that broad can't identify from conversion signals alone, and when you're using it as a creative testing mechanism to validate messaging angles before scaling broad.
At RocketShip HQ, we've seen interest targeting outperform broad in highly niche verticals, for example, a meditation app targeting people interested in specific spiritual practices, or a fintech app going after cryptocurrency traders. In these cases, the niche interest signal is strong enough to overcome the data degradation. But for mainstream consumer apps (fitness, shopping, entertainment, dating), broad almost always wins.
- New app launches with fewer than 50 conversions per week: interest targeting can help bootstrap initial signal
- Niche audiences under 5M total addressable users where broad targeting may waste too much spend on exploration
- Creative concept testing: use interest targeting to test whether specific messaging resonates with a defined persona before going broad
- Android-only campaigns where signal loss is less severe and interest segments remain more accurate
How should I structure a test between broad and interest targeting on Meta?
Use a core/test ad set structure. Allocate 85-90% of your budget to your proven approach (usually broad, if you already have data) and dedicate 10-15% to testing the alternative. Run the test for at least 7 days with identical creative across both ad sets to isolate the targeting variable.
This follows the core/test framework we recommend for all Facebook creative testing. The key principle: never risk your core performance to test a hypothesis. For a thorough walkthrough of campaign structure, see our guide on how to run Meta ads for mobile apps.
Step-by-Step Test Setup
Create two ad sets with identical creative (use 3-5 of your top-performing ads). Set one to broad targeting (age and geo only) and one to your best interest stack. Keep daily budgets proportional. Critically, do not change budgets by more than 10% per day once the test is running, as larger changes reset Meta's learning phase. Evaluate on cost-per-conversion events that matter downstream (purchases or subscriptions, not just installs) and use a 24-48 hour data lag buffer before drawing conclusions.
What Metrics to Compare
Don't just compare CPI. Look at cost per trial start, cost per subscription, and Day 7 ROAS. We've seen cases where interest targeting had a lower CPI but significantly worse downstream metrics because the "interested" users were lower-intent browsers. Broad targeting often delivers users who are further along in their decision journey because Meta's algorithm optimizes for actual conversion behavior, not profile attributes.
Does broad targeting work differently across Meta's placements like Reels vs. Feed?
Yes, and this is an underappreciated nuance. Even within broad targeting, your creative choices heavily influence which placements Meta allocates spend to, which in turn determines the demographic and behavioral profile of users you reach. Broad targeting is never truly "random" because creative acts as a targeting lever.
Meta is not a single monolithic channel. It comprises distinct placements (Facebook Feed, Instagram Feed, Facebook Reels, Instagram Reels, Stories, Audience Network) that each attract different user demographics. Facebook Feeds tend to skew older (35+), while Instagram Reels skew significantly younger (18-34). When you run broad targeting with a UGC-style vertical video, Meta will naturally allocate more spend to Reels placements, effectively targeting a younger audience without any interest segments.
- Vertical 9:16 video creative pushes spend toward Reels and Stories placements (younger demographics)
- Static images and link-post formats tend to concentrate spend in Feed placements (older demographics)
- Monitoring your placement breakdown is essential even with broad targeting to understand who you're actually reaching
- At RocketShip HQ, we use creative format as a deliberate targeting strategy within broad campaigns
What about Advantage+ app campaigns vs. manual broad targeting?
Advantage+ campaigns take broad targeting even further by automating audience selection, placements, and budget allocation entirely. In our testing, Advantage+ performs comparably to or slightly better than manual broad targeting for mature campaigns with strong creative libraries and sufficient conversion volume (50+ events per week per ad set).
However, Advantage+ gives you less visibility and control. You can't see placement breakdowns as granularly, and you can't test specific creative-audience combinations. For a deeper comparison, we've written about what Advantage+ is and when to use it for app campaigns. Our recommendation: use Advantage+ as your scaling vehicle once you've validated creative in a more controlled manual broad setup.
How much budget do I need before broad targeting becomes effective?
Broad targeting generally needs a minimum of 30-50 conversion events per week per ad set to exit the learning phase and optimize effectively. For most app campaigns optimizing on installs, that means a minimum daily budget of roughly $100-200 per ad set, depending on your category's CPI.
If you're optimizing for deeper funnel events like purchases or subscriptions, you may need $500-1,000+ daily per ad set to generate enough signal. This is where many smaller advertisers struggle with broad targeting. They don't have enough budget to generate the conversion volume Meta needs. In those cases, optimizing for a higher-funnel event (like app installs or trial starts) with broad targeting often outperforms optimizing for purchases with interest targeting, because the algorithm gets more data to work with.
- Under $50/day per ad set: interest targeting or lookalikes may still outperform broad
- $100-500/day: broad targeting typically starts winning, especially for install optimization
- $500+/day: broad targeting almost always outperforms, and Advantage+ becomes viable
- The threshold is about conversion volume, not raw spend. Lower CPI categories can go broad on smaller budgets
Should I use lookalike audiences instead of broad or interest targeting?
Lookalike audiences sit between interest and broad targeting, and their effectiveness has degraded post-ATT but not as severely as interest targeting. In our data, 5-10% lookalikes based on purchasers perform within 5-10% of broad targeting, but rarely beat it outright. Narrow 1% lookalikes almost always underperform broad now.
The degradation of lookalikes is directly tied to the same signal loss affecting interest targeting. The seed audiences are less accurate because Meta can't fully track user behavior across apps and sites. That said, lookalikes can be useful as a middle step for advertisers who aren't yet comfortable going fully broad, or as a way to nudge the algorithm's exploration in a specific direction during the early days of a campaign.
- 1-2% lookalikes: often too narrow, underperform broad by 15-25% on CPA
- 5-10% lookalikes: closer to broad performance, sometimes useful as a stepping stone
- Value-based lookalikes (based on purchase value) still show marginal benefit in some categories
- For most mature advertisers, the recommendation is: skip lookalikes and go broad with strong creative differentiation
The data is clear: for the majority of mobile app campaigns on Meta, broad targeting outperforms interest targeting in a post-ATT world. The algorithm is smarter than our manual audience assumptions, provided it has enough conversion data and strong creative to work with. At RocketShip HQ, we've shifted the vast majority of our clients to broad targeting with creative-led audience strategy, and the results have consistently validated that approach. The real targeting lever in 2024 and beyond is your creative, not your audience settings.
Looking to scale your mobile app growth with performance creative that delivers results? Talk to RocketShip HQ to learn how our frameworks can work for your app.
Not ready yet? Get strategies and tips from the leading edge of mobile growth in a generative AI world: subscribe to our newsletter.
