How to Structure Facebook Ad Campaigns for Maximum ROAS

Most brands are leaving 3-4x ROAS on the table because they skip the campaign structure step. They just blast ads at random and wonder why they can't scale. The difference between 2x ROAS and 6x ROAS usually comes down to one thing: how you organize your campaigns, ad sets, and ads.

This post covers the exact framework we use to structure Facebook campaigns that actually scale.

Understanding Facebook's Three-Level Hierarchy

Facebook gives you three layers to work with: campaigns, ad sets, and ads. Get these wrong and your targeting is muddled. Get them right and you can scale intelligently.

The Campaign Level

The campaign is your top container. You pick one primary objective here: conversions, traffic, engagement, or leads. That objective sticks around for the life of the campaign. Everything else lives inside it - all your ad sets, all your budget controls, and your Campaign Budget Optimization (CBO) settings if you use them.

Think of this as your strategic bucket. You usually create one campaign per objective per testing period. Running a summer sale and a holiday campaign? Two separate campaigns. Even if they overlap on audience, they're different campaigns.

The Ad Set Level

This is where the actual segmentation happens. You define audience, placement, budget (if you're not on CBO), scheduling, and optimization settings. Ad sets are audience containers, and the key rule is simple: one unique audience per ad set. If you cram five different audience segments into one ad set, Facebook's algorithm gets confused and your performance tanks. It needs focus.

The Ad Level

The ad is just the creative asset: image, video, copy, headline, CTA. You can run multiple ads inside one ad set, which is where Facebook runs its own A/B tests to find the best performer. This is your testing ground for creative variations.

Campaign Budget Optimization vs. Auction-Based Optimization

The biggest structural choice you'll make is whether to use CBO or ABO. It matters.

Campaign Budget Optimization (CBO)

You set your budget at the campaign level. Facebook's algorithm then distributes that budget across all your ad sets based on real performance. Meta now prefers this approach and it usually outperforms the alternative.

CBO wins because the algorithm gets smarter with auto-reallocation, you have less manual work, and you can test more ad sets without worrying about budget fragmentation. The catch: you need enough volume (roughly 50 conversions per week across the campaign) for the algorithm to actually learn.

One downside is you lose explicit control over spend per audience. Budget can concentrate on just a few winning audiences while others stay under-explored.

Auction-Based Optimization (ABO)

You set budgets at the ad set level, so you control exactly how much goes to each audience. This gives granular control and works when you need to guarantee minimum spend on certain audiences for learning data.

The problem: it requires constant manual babysitting and you're basically telling Facebook "Spend only this much" instead of "Spend where it works." ABO underperforms CBO in most ecommerce scenarios.

Use CBO for scaling campaigns. Use ABO only when you have a specific reason (testing a particular channel, brand safety concerns) to restrict budget flow.

How Many Ad Sets and Ads to Test

This is where most people either overthink it or underthink it.

The Right Number of Ad Sets

Start a new campaign with 3-5 ad sets. Segment them by your clearest divisions: customer stage (awareness/consideration/conversion), interest cluster, or geography if you're big enough. Don't create ten ad sets with minor variations hoping one sticks. Facebook needs volume to optimize.

Once you hit 50-100 conversions, the data gets real. Look at cost per result and ROAS by ad set. Kill the consistent underperformers. Double down on the winners with manual bid adjustments or duplications with slight audience tweaks.

Creative Testing Within Ad Sets

Test 3-5 creative directions inside each ad set. This lets Facebook find its preferred creative while keeping the audience tight. Test one variable at a time: static vs. video, emotional appeal vs. benefit-driven, different demographic hooks. Not all of these at once.

At 2,000-3,000 results per campaign, your creative data becomes reliable. Take the winner, isolate it, scale it, and port it to other campaigns.

Audience Segmentation by Ad Set

Good campaign structure starts with smart audience segmentation. Each ad set needs to represent a real, coherent group.

Segmentation Strategies

Stage in journey: Separate ad sets for cold (awareness, education), warm (site visitors, engagement history), and hot (cart abandoners, past customers). Different groups need different messages.

Interest clusters: Group related interests together instead of one interest per ad set. 3-5 related interests per ad set gives the algorithm more volume to work with.

Lookalike audience tiers: If you use lookalikes, segment by similarity. A 1% lookalike belongs in its own ad set away from your 5% lookalike.

Geography: If regional differences matter to your business, segment by location. This lets you optimize messaging and creative per region.

Device: Not always necessary, but if mobile and desktop perform wildly differently, segment them.

Prospecting vs. Retargeting Campaign Structure

This is the most important structural call you'll make: do you split prospecting and retargeting into separate campaigns or combine them?

Separate Campaigns Approach

One campaign for prospecting (new audiences), one for retargeting (people who've touched your brand). This structure clarifies measurement and lets each campaign use different optimization settings.

Prospecting needs longer conversion windows (7-28 days) and new customer optimization. Retargeting can optimize for fast conversions (1-3 days) because these people already know you. This is the default structure for mature brands.

Combined Campaign Approach

Smaller brands sometimes combine prospecting and retargeting in one campaign with CBO handling the allocation. It works if you have solid overall volume, but you lose visibility into which channel actually drives revenue.

Creative Testing Campaign Structures

Run a separate testing campaign away from your main scaling campaign.

Keep it smaller (500-2,000 per day) and prioritize learning over quick revenue. Test three completely different creative directions and let them run 1-2 weeks for statistical clarity.

Winners graduate to your scaling campaign. This keeps your money-maker focused on proven winners while you build a pipeline of new ideas on the side.

Scaling Campaign Structures

Found a winning combo of audience and creative? Scale thoughtfully.

The Duplicate and Tweak Method

Don't jack a single ad set from 1k to 10k per day overnight. Duplicate it 2-3 times and gradually increase budgets. This tells Facebook "I want to test variations at scale" and gives the algorithm room to adjust.

The Portfolio Expansion Method

Keep your winning structure intact but add adjacent ad sets. If cold audiences work, add a lookalike audience based on converters. If cart abandoner retargeting works, add a "viewed product, didn't add to cart" segment.

Budget Increase Sizing

Increase daily budget 20-50% every 3-4 days. Don't jump it all at once. Facebook needs time to gather new data at each spend level before the next jump.

Budget Allocation Across Campaigns

A typical allocation for an established ecommerce brand: 60-70% goes to top-funnel prospecting, 20-25% to mid-funnel engagement, 10-15% to bottom-funnel conversion (cart abandoners, past visitors). If you're brand new, flip this. You need to build audiences first.

Running multiple product categories? Allocate budget based on sales contribution plus an extra 20% on underperformers for testing.

When to Consolidate vs. Segment

Should you combine ad sets or keep them split?

Consolidate when:

Multiple small ad sets each have under 10 conversions per week (consolidate to hit the learning threshold faster)
You're running nearly identical audiences (consolidate lookalikes that are only 1% apart)
Your test is complete and you want to merge learnings into scaled performance

Segment when:

An ad set passes 50+ conversions per week and you notice demographic subgroups performing differently
You're testing an entirely new audience (keep it separate from proven audiences)
You're testing new creative (maintain audience separation so creative performance is clear)

Measuring Which Structure Works Best

How do you know your structure is actually working?

Track these in ORCA, which gives you cross-platform analytics:

Cost per result by ad set: Look for 10-15% efficiency differences across segments. If you're not seeing meaningful variation, your segmentation is too granular.

Creative performance by ad set: If the same creative performs completely differently in different ad sets, your audience segmentation is actually working.

Campaign-level ROAS: Track ROAS by campaign type (prospecting vs. retargeting) to verify the split matches your business.

Learning efficiency: In CBO campaigns, monitor how quickly Facebook reaches the learning phase. Slower than expected? You probably have too many ad sets.

Putting It All Together

A solid structure for a DTC ecommerce brand looks like this:

Campaign 1: Prospecting (CBO, 5-7 ad sets)
- Ad Set 1: Cold audience, interest targeting
- Ad Set 2: Lookalike audience, 1%
- Ad Set 3: Lookalike audience, 2-3%
- Ad Set 4: Engagement retargeting
- Ad Set 5: Education-focused content (soft launch)
Campaign 2: Conversion Retargeting (CBO, 3-4 ad sets)
- Ad Set 1: Cart abandoners, 1-3 days
- Ad Set 2: Cart abandoners, 4-7 days
- Ad Set 3: Viewed product, didn't purchase

Each ad set runs 3-4 creative variations. Budget flows to what works. You keep the structure consistent month to month, which makes comparison easy.

This approach gives Facebook's algorithm the focused audiences it needs to optimize, lets you measure audience effectiveness, and keeps clear visibility into which segments actually drive profit. Structure isn't exciting, but it's what separates 2x ROAS from 6x.