Interests, Lookalikes, or Broad? The Targeting Decision Guide for DTC Operators in 2026

Nine times out of ten, when we take over an account and someone asks how we're going to target, the answer is broad. Age, gender, location, and nothing else. No interests, no lookalikes, no carefully stacked behaviour layers.

That gets a raised eyebrow surprisingly often, because a lot of founders have been taught that targeting is where the skill lives. You pick the perfect audience, Facebook finds those exact people, and the ad does its job. So "we're just going to let it run open" sounds lazy, or like we're skipping the hard part.

It isn't lazy. It's that the hard part moved. Let me explain why we default to broad, where that default is genuinely wrong, and how you can test it on your own account without taking my word for any of it.

Why broad works now: the creative does the targeting

Here's the thing most targeting debates miss. When you upload an ad, Facebook reads the ad itself and builds the audience from that, not from the box you ticked.

It pulls the image apart, reads the copy and the headline, transcribes the video, clocks whether the person on screen is smiling or sad, young or old, in a gym or a kitchen. Hundreds of thousands of signals off a single creative. Then it goes and finds the people most likely to respond to that specific ad.

There's a clean little demonstration of this I like. Run two near-identical ads in the same broad ad set, both targeting men and women. One shows a male golfer, one shows a female golfer, same copy otherwise. The male golfer ad will spend almost all its budget reaching men. The female golfer ad will spend most of its budget reaching women. Same audience setting, completely different delivery, decided entirely by the creative.

That's the whole argument in one example. The ad is the targeting. So when you bolt an interest or a lookalike on top, you're not adding precision - you're often overriding the most sophisticated targeting tool in the account with a guess about who your customer is.

This is also why I've come round to calling it "unrestricted" rather than "broad" in my own head. Broad has a stigma, like you're spraying ads at penny-pinchers and hoping. You're not. You're handing the algorithm the full market and letting it go after whoever's most likely to convert, instead of fencing it into a slice you picked.

The three methods, ranked, with the numbers that matter

Here's my honest ranking for a DTC brand in 2026, and crucially, why.

1. Broad (unrestricted)

This is the default for a reason. A few of them, actually.

• Scale. Picture your total market as a big circle. Broad lets the algorithm spend across the whole circle. An interest or lookalike is a small box inside it - you might reach 10 to 50% of your real buyers and never touch the rest. Across our portfolio, broad is consistently the only setup we can push to genuine scale; interest and lookalike campaigns tend to choke well before they get there.
• Stability. Broad accounts wobble less. Fewer random CPM spikes, fewer CPA jumps out of nowhere. You still get up and down days, but the swings are gentler because the algorithm isn't boxed into a thin, expensive pool of people.
• It frees your budget for creative. This is the one founders underrate. Audience testing is a money pit. I've seen accounts quietly spend 10% of a large annual budget chasing the perfect audience - on a A$2m-a-year account, that's A$200k that produced nothing a good creative wouldn't have produced for free. Go broad and that money goes into making better ads, which is the thing that actually moves the result.

The one real cost of broad: it puts all the pressure on your creative. If your ads don't clearly call out and appeal to your buyer, broad has nothing good to work with and it'll show. No hiding behind the audience.

2. Lookalikes

Useful in a narrow set of cases, which I'll get to, but as a default they've aged badly.

• They need data to exist at all - you want a source list of at least 500 to 1,000 customers before a lookalike is worth anything.
• They're fiddly. Purchase or page-view or engagement source, 30 or 60 or 180-day window, 1% or 3% or 5% - a lot of knobs, and most combinations are noise.
• They rarely scale. I've struggled to push pure lookalike campaigns far at all before performance falls over; a few hundred dollars a day is often the ceiling.
• A lazy lookalike can quietly poison itself. Build it off your last 30 days during a heavy discount period and you've taught Facebook to go find more discount-hunters.

Here's the kicker though: broad sort of gives you the lookalike effect for free. When you optimise for purchases, every sale teaches the algorithm what your buyer looks like, and it keeps finding more of them. You get the "people like my customers" benefit without manually building and babysitting the audience.

3. Interests

Bottom of the list for most accounts, and the gap is widening.

• You pay a premium. Everyone else is targeting "women's leggings" and "Gymshark" too, so you're bidding against a crowd for the same inventory, which pushes your CPMs up.
• Facebook is visibly walking away from them. Audience expansion, which used to be optional, is now on by default - meaning even when you hand Facebook an interest, it reserves the right to go find its own people anyway. You're not really in control of the targeting you think you set.
• To scale an interest campaign you usually have to keep stacking more interests on, which is exactly the manual upkeep broad lets you skip.

I won't pretend interests never work. They can. But as a standing strategy in 2026, you're paying more to hand the algorithm worse instructions than it would give itself.

Where I resist the dogma: two profiles that still earn a slot

Now, if you've read this far thinking "this bloke just wants me to go broad and never think again" - no. I'm wary of targeting dogma in both directions, and there are two client profiles where I'll genuinely reach for interests or lookalikes.

Profile one: the brand-new account with no signal and no creative library. A lot of people will tell you broad always works from scratch, and often it does. But there's a real case in the first few weeks, before you've got a single converting creative, where the algorithm has nothing to learn from - no pixel history, no winning ad to read. In that window, a sensible interest or a lookalike off whatever early customer list you have can act as a set of training wheels. It gives Facebook a starting hint while you build the creative that'll eventually let you take the wheels off. The plan is always to graduate to broad once you've got a couple of proven ads - but I won't pretend the cold-start moment is identical to a mature account.

Profile two: the niche product with a genuinely thin audience. If your buyer is a slim, oddly-shaped slice of the population - a B2B-ish product for, say, independent restaurant owners, or a device for a specific medical condition - broad can struggle to find the needle, because the creative-reads-the-ad magic works best when there's a decent-sized pool to sort. In a very thin market, gently pointing the algorithm with a tight interest can shorten the search. I'd still test broad alongside it, because broad with a sharp, well-targeted creative often surprises you even here. But this is the profile where I keep an open mind rather than reciting the rule.

Notice both exceptions are about a lack of signal - either you have no data yet, or your buyers are too sparse for the algorithm to find unaided. Once that constraint lifts, broad almost always wins. So the question isn't "broad or not", it's "do I have enough signal yet for broad to do its thing".

How to test this on your own account, properly

Don't take my ranking as gospel. Run the controlled version on your own numbers - it's the only opinion that counts.

Keep your current setup running as the control. Then carve out a small slice - 10% of your daily budget is plenty - and stand up a clean broad campaign next to it, loaded with your best-performing ads. Now leave it alone. The mistake is killing it after three bad days; give it a fortnight to find its feet, because a fresh campaign always looks rough before the algorithm settles.

Then compare like for like: CPA, CPM, and how far each will actually scale before the return falls over. If broad wins, feed it more budget and let it become your main campaign. If your control genuinely holds up, you've learned something real about your account rather than copying mine.

That's the spirit I'd want you to take from all of this - less "what's the right audience" and more "what does my own data say when I test it cleanly". Default to broad, stay honest about the two cases where it isn't the answer, and let the account settle the argument.

If you'd rather not run that test alone, sorting out exactly which targeting setup fits your account, your data, and your creative is a core part of what we do when we manage a brand's Meta. Either way, the call should come from your numbers - so go pull them.