Broad vs Interest Targeting on Meta: How to Decide
Every Meta ad set asks you the same question: who should see this ad? One camp says leave the audience wide open and let the system find buyers. The other says narrow it with interests so the budget is not wasted on the wrong people. Both camps are right some of the time. The useful question is which one your account can support right now.
Why the usual debate goes nowhere
The argument is usually framed as "broad works" versus "interests work". That skips the two things that actually decide it: how much conversion data the account produces, and how good the signals are that you feed back to Meta.
Also, the terms have shifted. Meta's current help pages describe audience inputs such as detailed targeting, custom audiences and lookalikes as suggestions: delivery starts with people who match them and may widen to others when the system expects better results. Hard limits, such as location, language and minimum age, behave differently. Check Meta's current documentation for which inputs are suggestions and which are controls in your account, because this has changed more than once.
So "interest targeting" today is closer to a hint than a fence. That changes how much it can protect you.
A four-factor decision model
Score your situation on four factors before choosing.
Conversion volume
How many purchases per week can each ad set realistically collect?
Signal quality
Are Pixel and server events complete, deduplicated and carrying correct value?
Offer and creative breadth
Does the product appeal to a wide group, or to a specific niche?
Cost of a wrong click
How expensive is it if the system explores people who will never buy?
Strong volume and clean signals favour broad. Thin data, a niche offer or a costly wrong click favour a guided start.
- Volume. Meta's learning-phase documentation refers to roughly 50 optimisation events in a week for an ad set to leave learning. Confirm the current figure in its help pages. Every extra ad set divides your events.
- Signal quality. Wide audiences depend entirely on the data you send. If purchase events are missing or doubled, broad targeting amplifies the error. Fix that first: see Meta Conversions API explained and Duplicate purchase events.
- Offer breadth. A kitchen product or a phone case has many possible buyers. A B2B tool or a specialist item does not.
- Cost of exploration. High-ticket items and low-margin products make wasted impressions expensive.
Illustrative example, not real results
Assume a store with a daily budget of 1,200 SAR and a cost per purchase of about 40 SAR. These numbers are invented to show the logic.
6 interest ad sets
Daily budget each: 200
Purchases per week each: 35 (below 50)
2 ad sets
Daily budget each: 600
Purchases per week each: 105 (above 50)
1 broad ad set
Daily budget each: 1,200
Purchases per week each: 210 (above 50)
Weekly purchases = daily budget x 7 / 40 SAR. Assumes the same cost per purchase in every structure, which real accounts will not guarantee.
The arithmetic is simple: 200 x 7 / 40 = 35, 600 x 7 / 40 = 105, 1,200 x 7 / 40 = 210. Six narrow ad sets each fall short of the 50-event reference, so each one learns slowly. Two or one ad sets clear it.
This does not prove broad is cheaper. It shows that splitting a modest budget across many audiences starves each of data, which is the real cost of over-segmenting. The cost per purchase will also differ between structures, in either direction.
When the answer changes
- Choose a guided start if the account is new or has very few purchases, the offer is niche, or the product needs a specific buyer. Use one or two audience suggestions, not six ad sets.
- Choose broad if you have steady purchase volume, clean events and creative that speaks to a wide group. Let creative do the targeting: the message decides who stops scrolling.
- Revisit after a large creative change, a new market, or a price change. Audiences that suited the old offer may not suit the new one.
Do not let a single week settle it. Performance moves with creative, season and competition as well as audience.
How to run the test
- Fix the offer, creative set and budget so only the audience differs.
- Run one broad ad set against one guided ad set, with enough budget for each to collect meaningful purchases.
- Compare cost per purchase against your break-even, not against the other ad set alone. See How to calculate your advertising break-even point.
- Decide on a pre-set window and sample, and avoid editing mid-test, which can restart learning. See The Meta Ads learning phase.
What the data cannot tell you
Meta reports results for the audience it delivered to, not the audience you suggested, so an interest label in a report does not prove the interest drove the sale. Attribution windows, view-through counting and overlap between ad sets blur the comparison. A winner in one test is evidence for your account and period, nothing more.
Next step
Decide how many ad sets your purchase volume can support, then test one variable at a time. If you want that structure reviewed against your numbers, the ads management service covers it. For campaign layout, read How to structure Meta Ads campaigns for an e-commerce store; for testing creative alongside audiences, see A simple creative testing framework.