Lookalikes vs. Broad: Findings from a Facebook Marketing Agency

Spend enough time in a facebook ads account and the Lookalike vs. Broad debate stops being a thought exercise. It becomes a budget line item with consequences. Our agency has run this test many times, across ecommerce brands, subscription products, app installs, and lead gen. The answer is not a one size fits all. It starts with the data you can feed the system, the way you handle creative, and the patience you have for the learning phase. What follows are field notes from a facebook marketing agency that has scaled and broken plenty of campaigns along the way.
Quick definitions that matter in practice
Lookalikes build an audience based on a seed. That seed might be purchasers, leads, high value customers, or predictive signals like Value-Based Lookalikes sourced from purchase value. The size of a lookalike is a sliding percentage of a location's population. A 1% lookalike in the United States is roughly 2.5 to 3 million people. A 5% is five times that. You can stack multiple lookalikes, or keep them separated for control.

Broad means little to no targeting beyond age, gender, and location. In modern accounts, Broad often uses Advantage+ Audiences, which gives the delivery system wide latitude to find conversions based on your pixel and account history. There are knobs to turn, but the best results usually come from trusting the system, not boxing it in with interests.

Both approaches still rely on creative, bidding, and clean signal quality. The targeting is only an amplifier.
Why this question still matters
Meta has leaned into automation. Advantage+ Shopping Campaigns, audience expansion by default, simplified objectives, and recommendations to avoid heavy targeting filters. Many advertisers read that as a full endorsement of Broad. Yet we still see lookalikes win in certain conditions, often by meaningful margins. If you are a performance ads agency accountable https://andyuqnk195.lucialpiazzale.com/seasonal-campaigns-a-facebook-marketing-agency-strategy-1 https://andyuqnk195.lucialpiazzale.com/seasonal-campaigns-a-facebook-marketing-agency-strategy-1 for hard numbers, you need to know when to go Broad, when to anchor on lookalikes, how to avoid overfitting, and what to watch during scaling.
How the platform changed what works
Two shifts define the current environment. First, privacy changes reduced stable identifiers. That made old school interest stacking brittle. Second, Meta’s modeling improved, especially when you give it high intent conversion signals and enough volume. Broad targeting benefits the most from strong modeling. Lookalikes depend on seed quality, so bad seeds hurt more than they used to.

We also notice that the learning phase is stricter on noisy conversions. Optimizing for top of funnel events like ViewContent or ATC often yields cheap but empty traffic. Broad with a purchase objective can look worse in the first three days, then surpass lookalikes after the algorithm locks onto reliable purchasers. This is where many teams blink too soon and misread the race.
When lookalikes beat broad
We see lookalikes outperform Broad when the seed combines high intent and differentiation. A good example is a subscription coffee brand that tagged first to third month retained customers as the seed. Their 1% and 2% lookalikes beat Broad by 18 to 32 percent in blended CAC over six weeks, with more stable CPA during spend increases. Broad found buyers, but too many were one and done. The retained cohort lookalikes tilted acquisition toward stickier customers.

Value-based scenarios behave similarly. A DTC jewelry brand uploaded 90 days of purchase value and built 1 to 3% VBLALs. Those audiences produced a 22 percent higher AOV than Broad at similar CPA, lifting MER at scale. The seed introduced a tilt toward higher order values that Broad only discovered later with more budget.

Smaller countries or niche categories also tend to favor lookalikes at first. In markets where total reach is limited, Broad can burn on low probability impressions before it triangulates on converters. A 1% lookalike in Sweden gave us faster time to first purchase and healthier CPC on a boutique skincare client. After about 20 days and stronger signal density, Broad caught up, but the early cash flow from lookalikes mattered.

Seed size is a common culprit. With fewer than 500 to 1,000 high quality seed events in a 30 to 90 day window, we still see lookalikes outpoint Broad if the seed is specific. Add only purchasers, not ATCs or email signups. If you have 150 to 300 purchases a month but strong creative and clean pixel events, 1% lookalikes often give a more predictable CPA floor for scaling to the first 1,000 daily spend.
When broad is the better bet
Broad shines when the account has healthy recent conversion volume and your offer appeals to a wide swath of users. A apparel marketplace with thousands of SKU options and daily purchases across price points is a classic fit. In those accounts, Broad paired with Advantage+ Shopping unlocks lower CPM and faster learning. We have seen 10 to 25 percent cheaper CPA than 1% lookalikes after two weeks, provided the creative rotates aggressively and the catalog feed is clean.

Broad also does better when creative drives the segmentation heavy lifting. Hooks, UGC angles, and product education will isolate the right people even in a wide audience. If your creative library is thin and repetitive, Broad often looks wasteful. With a steady stream of fresh assets, Broad becomes a flexible canvas. We measured this on a home fitness brand. When we ran two new concept families per week, Broad stabilized. When we paused ideation for three weeks, CPA drifted up 40 percent, and lookalikes temporarily won again.

Another Broad advantage shows up at higher budgets. Once you push past 3 to 5 times your daily CPA target in spend, narrow audiences can saturate quickly. Frequency climbs, CPC rises. Broad has more breathing room, so the cost curve is flatter. A shoe brand with a 45 dollar CPA target could spend 12 to 20 thousand a day on Broad with a steady 1.1 to 1.3 frequency per 7 days. Their 1% lookalike ad set hit the same CPA at 3 to 5 thousand a day, then climbed fast.
Broad vs. lookalike in one page
To keep the comparison sharp, here is a compact cheat sheet we use in our fb ads agency when planning a new account.
Choose lookalikes if your seed is high intent and distinct, especially value based or retained customers, and you have at least 500 to 1,000 seed events in the past 30 to 90 days. Choose Broad if your account already logs steady purchases every day, your product has wide appeal, and you can ship new creative weekly. Favor lookalikes in smaller markets or when budgets are modest, to reduce early waste and stabilize CPA fast. Favor Broad when scaling past 3 to 5 times daily CPA target in budget, to avoid frequency spikes and audience saturation. Use both in parallel when testing new geos, new price points, or new creatives, then reallocate once 7 to 14 days of stable data accumulates. The seed: what separates good from junk
A lookalike inherits the character of its seed. That line sounds obvious, but in practice we see messy seeds all the time. A beauty brand tried to build a lookalike off “7 day purchasers,” but more than half the conversions were false positives from a misfiring integration. No wonder their 1% lookalike did worse than Broad by 60 percent.

The best seeds share three traits. First, clear intent. Purchase events tracked via server side API with order value and product IDs, or leads scored by qualification, not just form fills. Second, recency. A 30 to 90 day window reflects current creative and offers. Third, representativeness of the goal. If you want subscribers, seed on active subscribers, not one time buyers.

Value based lookalikes deserve their own note. They work when your value data is real and not overly skewed by a few whales. For small catalogs with lumpy revenue, consider trimming the top 1 to 5 percent of outliers from the seed upload to reduce noise.
Creative is the real targeting
Neither audience type saves bad creative. We have ad sets where the best UGC video drives 80 percent of conversions regardless of audience. That is not an accident. Creative is how the algorithm learns. It is the language you speak to the feed.

What helps most in both Broad and lookalike campaigns:
A rotating cadence of new concepts, not just variations. New aspect ratios, fresh hooks, and different angles. Small trim edits do not count as new concepts.
That is one list. Keep count. There is still room for one more.

Product education over pure sizzle matters more in Broad because you are meeting colder prospects more often. The first line must flag the problem and the role of the product, not just a discount. In lookalikes, you can push price or urgency a bit harder because the users already resemble buyers.

Catalog feeds anchor Broad performance in ecommerce. Verify that your top sellers have robust product images and accurate availability. When we fixed broken fields and pruned 35 percent of dead SKUs in a home decor shop, Broad catalog campaigns picked up 17 percent ROAS without changing audiences or bids.
Budget, pacing, and the learning phase
The platform needs signal density. A good rule of thumb is to fund an ad set to generate at least 25 to 50 target conversions per week. If your CPA target is 50 dollars, you need 1,250 to 2,500 dollars a week per ad set. If the budget cannot clear that threshold across multiple ad sets, consolidate. A single Broad ad set might learn better than three lookalike splits that each starve.

Patience is contextual. We give Broad more time to settle than lookalikes, because it starts wide. A 7 day window is the minimum for meaningful evaluation, ideally 10 to 14 days if the budget allows. Pull decisions earlier only if you see catastrophic metrics like CPM three times your norm or no add to carts after a few thousand impressions.

CBO versus ABO plays differently here. CBO with Broad can over allocate to click bait creative. If you use CBO, cap bad actors with minimums or use ad level cost controls to nudge distribution. ABO makes it easier to keep cleaner apples to apples tests between lookalikes and Broad, at least during the learning phase.
Geography and catalog depth
In large markets with deep catalogs, Broad becomes a natural fit. The United States, Canada, the United Kingdom, and Australia with SKU depth above 200 tend to reward Broad. In smaller markets or verticals with considered purchases, lookalikes help focus initial spend. Germany and the Nordics have given us repeated lookalike wins for high AOV goods, particularly when the brand story requires more education.

Cross border buyers also respond differently. If you run multi country ads with different currencies, separate ad sets per country with their own lookalikes often outperform a single Broad audience that lumps everyone together. Currency mismatch in creative suppresses conversion rates more than most teams estimate.
Edge cases and how we handle them
Lead generation. Lookalikes built on raw leads frequently underperform Broad on actual pipeline. The better play is to build a seed of qualified leads, demo completes, or opportunities, even if it is smaller. While waiting for volume, run Broad with a lead form that weeds out casual interest. Form friction is a feature.

Apps. For app installs, Broad usually wins once the SDK event stream is clean and you optimize for downstream events like purchase or level complete. Lookalikes help early if sampling is tiny, then Broad takes over as cohorts stabilize.

High AOV and low frequency purchases. Luxury, furniture, B2B software, and similar categories often do better with lookalikes up front. Include post purchase, multi touch creative that addresses objections. Broader audiences come later once you have a narrative that can cold start strangers.

Regulated categories. Alcohol, supplements, and financial offers can trigger stricter delivery. We have seen lookalikes moderate CPM volatility there, although approvals and compliant creative matter far more than audience type.

Nonprofits. Donor lookalikes built from recurring givers or higher lifetime contributions tend to outperform Broad on donor quality. However, Broad can find more one time donors inexpensively during giving season. Plan for both, just with different creative.
Measurement that keeps you honest
Attribution drift can mislead. If you only look at platform reported ROAS, Broad will sometimes look like a hero because it touches so much reach. We pair platform numbers with blended metrics and, when budgets justify it, geo holdouts or media mix modeling. For small to midsize advertisers, a simple leading indicator is new customer revenue per day relative to spend, checked against a baseline week. Track repeat purchase rate by audience source if you can.

Guard against creative confounds. Run the same top four to six ads in both Broad and lookalike tests, with consistent budgets, placements, and conversion objectives. If Broad gets the edgy UGC and lookalikes get polished product demos, your test is already spoiled.
A simple testing playbook that scales
Use this sequence when the account can support it, and adjust only to maintain statistical sanity.
Phase 1, two weeks: Run ABO with two ad sets, one Broad via Advantage+ Audience, one 1% lookalike built on 30 to 90 day purchasers or a clean value based seed. Same creative pack in both, at least six distinct concepts. Fund each to achieve 25 to 50 purchases per week if possible. Phase 2, weeks three to four: Add a 2 to 3% lookalike ad set if the 1% holds up, and add a second Broad ad set with new creative concepts. Keep exclusions minimal. Monitor CPA, AOV, and 7 day new customer revenue by ad set. Phase 3, month two: Consolidate to the winners. If Broad wins, switch to CBO with a guardrail on spend per ad set. If lookalikes win, split seeds by value bands or retention. Scale budgets 15 to 30 percent every 2 to 3 days if CPA is within 10 to 20 percent of target. Creative cadence: Ship at least two new concept families per week. Kill underperformers quickly, but retain a few evergreen anchors for stability. Measurement: Check blended CAC or MER weekly. If platform CPA diverges from blended by more than 25 percent, pause changes and audit tracking, discounting, and promo overlap. Common mistakes that waste money
Stacking too many lookalikes into one ad set in hopes of scale. You lose the ability to see which seed drives performance, and the delivery system does not magically average them. Better to test a few precise lookalikes and only combine once you know their behavior.

Over filtering Broad. Slapping on interests, behaviors, or narrow age bands can strangle Broad before it breathes. The point of Broad is to let the system explore. If you do not trust that, stick with lookalikes or fix your creative.

Optimizing for soft conversions. Broad suffers the most when the goal signal is cheap and noisy. Link clicks and view content events teach the wrong lesson. Use purchases or at least add payment info or subscribe events in subscription funnels.

Underfunding tests. If you spread 2,000 dollars across six ad sets for a week, you have not tested anything, you have sprinkled. Consolidate, learn, then expand.

Ignoring frequency and overlap. As budgets grow, your lookalikes and Broad will start hitting the same people. That is fine until frequency climbs and creative fatigues. Rotate hooks, refresh thumbnails and first lines, and cull stale ad sets even if they were winners last month.
How online ads agency teams can operationalize this
Agency workflows benefit from predictable decision gates. In our facebook ad services practice, we keep a standing weekly review where each account presents a one page dashboard: spend, CPA, AOV, contribution margin, and a simple Broad vs. lookalike status line. That line might read Broad +18 percent CPA improvement week over week, lookalike VBLAL holding AOV +22 percent, next step: expand Broad creative pack B. This cadence prevents pet theories from lingering past their usefulness.

We also write down the seasonality context. Holiday CPMs can rise 30 to 60 percent. Product release cycles, paydays, or gift giving windows shift purchase intent. In those swings, lookalikes sometimes hold their edge because they target people closer to your established buyers. After the seasonal surge, Broad often reclaims the low CPA ground as auctions normalize.

Coordination with other channels influences which audience type wins. If search captures a chunk of branded demand and email drives returning buyers, Broad social media ads may look worse at first touch but win on incrementality. Conversely, a heavy influencer push primes pools that make lookalikes shine for a few weeks. Build your plans assuming cross channel echoes.
Practical ranges from recent accounts
To calibrate expectations, here are grounded ranges we have seen in the last year across a mix of ecommerce and subscription advertisers spending 50 thousand to 400 thousand a month on facebook advertising:
In mature accounts with 50 to 200 daily purchases, Broad CPA tends to beat 1% lookalikes by 10 to 25 percent after two weeks, assuming healthy creative rotation and clean conversion objectives. In newer accounts with 10 to 40 daily purchases, 1% and 2% lookalikes often win by 10 to 30 percent on CPA during the first 30 days. Broad catches up or surpasses as volume grows and the creative library expands. Value based lookalikes can lift AOV by 10 to 30 percent versus Broad at parity CPA when the brand has clear price tiering and accurate order value tracking. For high AOV goods above 200 dollars, lookalikes commonly hold a CPA advantage for longer, especially in smaller markets, while Broad delivers higher top line scale at slightly worse efficiency.
These are not promises. They are steady patterns that help set the test design and the patience level.
What a facebook advertising agency should recommend now
Start with both. If you have the budget to fund two or three ad sets to statistical relevance, launch one Broad through Advantage+ Audience and one lookalike built on clean purchase or value data. Use the same creative pack. Watch not just CPA but also AOV and early retention proxies. Over the first 7 to 14 days, resist the urge to make five changes a day. Let the system learn, then act decisively.

If you run a social media marketing agency for smaller brands with modest spend, lean on lookalikes first to stabilize cash flow, and add Broad as your creative and signal quality improve. If you are a performance ads agency for scaled ecommerce, give Broad room to breathe, especially inside Advantage+ Shopping, but keep value based lookalikes in rotation to pull the customer mix toward profitability.

Above all, treat audiences as levers, not identities. Broad and lookalikes are vehicles. The engine is your creative, the fuel is clean conversion data, and the driver is your process discipline. Agencies that remember that order tend to win more reliably, regardless of which audience type holds the lead in a given month.