I’ve been playing around with different targeting setups lately, and something that keeps popping up in chats with other folks running igaming ppc is lookalike audiences. Every time I heard the term, I thought it sounded either too simple to matter or too unpredictable to trust. But I kept seeing people mention small wins here and there, so I finally sat down and gave it some attention. What surprised me most is how different the results can be depending on the quality of the seed audience and how patient you are with the learning phase.

The funny part is that I used to avoid anything “automated” because I always felt like I could manually target players better than any platform algorithm. I’d build my lists, tweak placements, adjust bids, and do that whole daily dance. But as igaming ppc keeps getting more competitive, I found myself hitting a ceiling. Some campaigns did pretty well, but others kept plateauing no matter how many little adjustments I made. That’s when I started to wonder if maybe I was leaving performance on the table by not at least testing lookalikes.

The first time I tried them, I’ll be honest, the results weren’t great. I fed the platform a seed audience that I thought would work, but it turned out the data wasn’t strong enough. Too mixed, too broad, and honestly just too small. I also made the classic mistake of expecting fast results instead of giving the audience time to settle. After a few days of wobbly performance, I shut it down and told myself lookalikes were overhyped.

A few weeks later, someone on another forum said the exact same thing I experienced, and they pointed out that the problem wasn’t the lookalike feature itself but the seed list. That made me rethink things. So I went back into my data and pulled a much cleaner list of players who had actually completed deposits over the past few months rather than general signups. This time the results behaved very differently. The cost per conversion came down slowly but steadily, and the quality of players coming through looked stronger.

What I noticed is that lookalikes react best when the seed audience is clear and intentional. If the platform can’t tell what “type” of player you want, it just spreads too wide. But when the list is focused, things tighten up and start to make sense. I also tested multiple percentage ranges instead of sticking to just 1 percent. The narrow audiences sometimes delivered higher quality, while the wider ones gave cheaper traffic. I still switch between them depending on the campaign goal.

I also stopped expecting lookalikes to magically fix bad ads. The creative still matters a lot. If the ads aren’t speaking to the right mindset, even the best targeting won’t save the campaign. In my case, the ads that leaned into simple starter offers performed better with lookalikes than anything too flashy or complicated. It just aligned better with what new audiences respond to.

Another thing that helped was comparing the performance of lookalikes with other pricing structures. There’s a post I came across that explains the differences in a simple way, and it gave me a better understanding of when certain bidding types make more sense. If anyone’s curious, here’s the link I found helpful: CPC vs CPM vs CPA for iGaming advertisers

I’m not saying lookalikes are a magic switch. They still need testing, and some niches inside igaming respond better than others. Sports bettors behave differently from casino players, and both behave differently from poker players. Sometimes the audience expansion works beautifully, and sometimes it pulls in people who look good on paper but don’t convert deeply. I’ve had mixed batches, and I think that’s just part of the territory.

But after giving it a fair shot, I can say they’re worth testing instead of assuming they won’t work. They helped me push past a few stagnant campaigns and opened up traffic I wouldn’t have reached manually. If you’re stuck in that spot where the same targeting groups keep recycling the same users, lookalikes can give your campaigns a bit of breathing room.

If you do try them, my casual advice is this. Make the seed list as clean as you can, give the learning phase enough time, test different ranges, and don’t treat them as a standalone solution. They work best as part of the mix, not the whole strategy.