Meta Advantage+ audience: keep the suggested interests or go fully broad?

A lot of prospecting setups still start with a tight interest stack, and then people wonder why CPMs look fine but sales don't scale. My view is that creative matters more than detailed targeting now. With Advantage+ audience, the suggested interests are a nudge, not a wall, so adding them mostly just narrows the pool Meta can learn from. Say a dental clinic runs three hooks aimed at different pain points. Going broad with only age and location set lets the hooks find the right people on their own.

The pushback is that broad needs enough conversions to learn from, so on a small budget interests can still be a reasonable starting point. Where do you draw that line, and how much conversion volume do you need before you trust broad?

Agreed on creative. Broad is only as good as the event you give it, though. Advantage+ audience will find whoever is most likely to trigger that event, so if you optimize for landing page views or cheap form opens, you'll get plenty of curious clicks from people who never buy.

Before you judge broad, check that the pixel and CAPI send the same purchase or lead event with a matching event_id, and look at the match quality score in Events Manager. If deduplication is broken, Meta sees inflated conversion counts, so the volume threshold looks met when it isn't. A common rule of thumb is around 50 real optimization events a week per ad set after dedup. Below that, interests as a seed are a fair compromise, but measure success on the downstream event, not the one Meta is optimizing.

Agreed, though I'd check the 50-a-week number against the store before trusting it. Pull last week's Meta purchase count and put it next to the orders in Shopify or your cart platform for the same dates. Privacy changes and ad blockers usually mean Meta reports a bit under the store, not over. If Meta shows noticeably more, dedup is broken, and no amount of broad targeting fixes that. Say a shop does 300 orders a week but Ads Manager shows 450 purchases. The broad ad set is learning mostly from phantom conversions.

Same check for local services, but the source of truth is different. Meta counts a lead when someone submits a form or taps a call, while the number that actually matters is booked jobs. Say a plumbing company gets 60 leads a week in Ads Manager and the office books 18 jobs. That ratio can be perfectly normal. If it drifts over a few weeks, though, the ad set is getting good at finding people who fill in forms, not people who need a plumber.

The attribution window matters too. If Ads Manager is on 7-day click plus 1-day view and your call tracking only counts same-day calls, the two numbers will never line up. Pick one window, compare like with like, and only then decide whether broad is working.

Agreed, and the usual fix for that drift is to feed the booked job back to Meta instead of tuning the audience. When the office marks a lead as booked in the CRM, you can upload it as an offline conversion matched on the lead's ID. Lead ads expose that ID, so the match is more reliable than phone number matching. Once the ad set optimizes toward booked jobs, broad has something much better to learn from than form fills.

Depends on your sales cycle. If a booked job or meeting usually lands two weeks after the form fill, that signal reaches Meta too late to steer the ad set, so you may need to send back an earlier stage instead, like a qualified call.

Agreed on the lag. I'd also check that the stage you send back actually shows up in Events Manager's offline events with a decent match rate, because a stage Meta can't match teaches it nothing.

Split the match rate by lead source before you trust it. Lead ID matching only works for leads that came through a Meta lead form. Leads from a website form or a call come in without that ID, so they fall back to hashed email or phone plus the click ID, and those match at a different rate. If you upload one blended file and look at one average, a weak source can hide behind a strong one.

I'd export the CRM booked-job rows by original source, upload each source separately, and check the match rate for each. A form lead at 90% match and a call lead at 40% tells you the call tracking needs fixing before you touch the audience. Also make sure your uploads land inside the attribution window you chose, or the offline events will look like they aren't attributed at all.

Agreed on splitting by source, and weight the match rate by booked-job value too. A 40% match on call leads that are 5% of your jobs is a small problem. The same rate on half your revenue is a big one. For a client report, the weighted number is the one to show, since a blended 70% can look healthy while the source that matters most quietly drops out. Which source do you think is losing the most matched revenue right now?