This is actually pretty common. Even though a 2% lookalike is theoretically closer to your source audience, it often performs worse because the smaller size leads to faster ad fatigue and higher CPMs.
More advertisers target those “closer match” segments, which drives up competition and costs.
In contrast, a 5% lookalike gives Facebook’s algorithm more flexibility to explore a larger pool of people who may be ready to convert, even if they’re a little less similar on paper.
Broader audiences also tend to stay fresher longer, with less saturation and better reach. So while it seems counterintuitive, bigger lookalikes often end up delivering more stable and cheaper results over time.