LaL worked, broad got 0 purchases - what next?

buggedout

First two days of a 1% Lookalike campaign based on my Instagram followers were solid - purchases came in and CPA looked healthy. Then day 3, CPA spiked hard. My guess: the audience pool is just too small (only around 160 followers).

So I switched to broad targeting. It's been running for 1 day and got 0 purchases so far.

I'm trying to understand a few things:

Does this mean the algorithm lost the signal when I turned off the Lookalike?
Is one day simply too early to judge broad targeting?
Should I keep spending on broad or revert to the Lookalike?
Would you duplicate the winning creatives into a broader audience gradually instead of flipping straight to broad?

Since the Lookalike campaign was converting, I assume the issue isn't the offer itself but rather audience quality or size.

Any thoughts from those who've been through this?

backlinker

One day is far too short a window to call a winner or loser. Split your budget evenly between the lookalike and broad campaigns, then keep a close eye on where the spend is actually landing - placement-level data is key. Let both run for at least three days with the exact same creatives before you even think about tinkering with audiences, placements, or ad format. The learning phase needs to settle first, otherwise, you're just chasing noise. After day three, compare results and make surgical changes based on real performance, not gut feel.

neonlights

here's the fix: run an A/B test with minimum 2x target CPA budget on two campaigns - broad and LAL - using identical creatives. Meta's built-in testing tool prevents audience overlap so you get a clean, statistically significant result. let it run at least 14 days before you judge anything. Broad can be a slow burner depending on the ad assets, sometimes it needs more runway to find the right buying cohort. Don't kill it early if the numbers look scary.

pixelperfect

honestly, I feel your frustration - it's hell when a lookalike audience crushes it and then broad just tanks. I've been in that exact spot with my own clients.

What worked for me was switching back to the lookalike but changing the source to website visitors or past buyers. Gives the algorithm a much bigger data pool to work with. And here's the bit most people skip: let it run for a full three days without touching it. that's usually enough time for the pixel to recalibrate properly.

It's not a magic fix, but it stopped the bleeding for me. Hope that helps a bit.

trafficcop

Oh honey, a 1% Lookalike built from 160 Instagram followers? That's not a seed, that's a single grain of rice. Of course it worked for two days - Meta found the three obsessed fans and patted itself on the back. Then reality set in the moment that pocket dried up. This isn't a "broad is broken" situation, it's a "your signal is so fragile it could shatter from a light breeze" situation.

When you jump from that tiny high-intent hole straight into broad, the algorithm needs a minute to figure out what the hell happened. First 24 hours usually feel like a ghost town - CPAs climb, purchases vanish, and everyone panics and restarts campaigns like rookies. I've seen this exact rodeo with a small ecom brand that had a 1% LaL from about 220 followers. Day one: $14 CPAs, feeling like geniuses. Day four: $60+ CPAs, crying into their spreadsheets. They finally switched to broad, had no sales for two days despite a CTR around 3.1%, and nearly threw in the towel.

Instead of nuking everything, they left the best creatives running, let broad stabilise, and slowly built proper custom audiences from real site engagement and purchase data. Three weeks later: CPA settled at £22-28, purchases jumped from 1-2 a day to 8-11 consistently. Broad ended up outperforming the original Lookalike because Meta finally had enough real conversion signal to optimise instead of chasing a handful of Instagram ghosts.

So tell me: is your broad campaign still serving healthy CTR and Add to Carts but just not sealing the deal? Or did the whole funnel collapse the second you left the Lookalike safety blanket?

dailygrind

Honestly, 160 followers is a tiny seed audience for a 1% LaL, so it's highly likely Meta found a small pocket of easy early conversions and exhausted them quickly. that doesn't mean the offer is dead. one day on broad is nowhere near enough to call it - especially if the pixel or account hasn't built up much purchase history. broad generally needs more time and signal density before it reliably finds the right buyers.

What I wouldn't do is completely scrap the winning LaL angle. I'd personally keep the same creatives, let broad run a bit longer, and layer in slightly larger seed audiences - website visitors, engagers, purchasers if you've got enough data - instead of relying solely on that tiny IG follower pool. The fact that something converted initially is a solid sign, now it's about finding scalable signal quality, not panicking after one quiet day.

datanerd

160 followers on a source list is basically noise for any lookalike algorithm. You need at least 500-1,000 quality conversions (not just followers) before a LAL has enough signal to differentiate your audience from the general population. that small seed set just forces the model to guess - you're better off burning that budget on broad or interest-based targeting and letting the pixel accumulate real data.

broad needs time, sure, but it also needs enough spend to exit the learning phase. For a conversion objective, I'd give it at least 50-100 purchase events before even glancing at the metrics. if you're evaluating broad after a couple of days with a handful of conversions, you're evaluating noise, not performance. give it a proper test budget - think 3x your target CPA - and let it run for a week minimum. Then you can compare apples to apples.

buggedout

thanks - so just to confirm: you're running LAL and broad as two separate ad groups inside the same campaign?

It's a logical setup for isolating performance. if LAL converts but broad gets zero purchases, I'd check a few things first:

Audience overlap - broad might be hitting the same pool as LAL if your pixel hasn't matured
Bid strategy - are both using the same? If broad is on lowest cost with no purchase signals, it'll spin its wheels
Match type in broad (phrase, exact, actual broad) - actual broad can waste budget without conversion history

What's the daily spend split between the two? If broad has less than 50-100 events per week, it's likely still in learning. wouldn't kill it yet, but I'd cap the budget until it accumulates data

backlinker

Keep two separate campaigns - one broad, one LAL. Mixing audiences dilutes performance data, isolating them lets you see exactly where the spend is wasted.

trafficcop

Ah, the classic "broad campaign goes rogue then suddenly remembers its manners" story. had a client go through the exact same circus - CPA hit $180, we made a few changes, and within a week and a half it started behaving like a civilised campaign again.

The magic is usually in what you're flogging and how you're showing it off. So spill: what's the product and what do your creatives look like? could be the same fix applies, assuming your client isn't selling something truly cursed.