Deividas here. Audience size on Meta is one of those topics where the conventional wisdom has flipped completely over the past three years. The old playbook was precise targeting: find the right interests, stack them carefully, keep the audience tight so you're only spending on the most relevant people. The new reality is almost the opposite. Here's why, and what audience sizing actually looks like in 2026.
Meta's algorithm has gotten significantly better at finding buyers within a broad audience pool. When you tell Meta to optimise for purchases and give it a large audience to work within, the model can identify patterns from your pixel data and purchase history that no manually built interest stack could replicate. Restricting that pool to a narrow interest-based audience essentially asks the algorithm to be smarter than it can be with less data, which is backwards.
The practical result is that most accounts see lower CPAs from broad targeting, meaning minimal or no interest restrictions, than from the tightly defined audiences that used to be standard. The creative is doing the targeting work that interest stacks used to do — a well-targeted ad finds the right people through engagement signals, not through audience parameters.
A targeting audience under 500,000 people creates problems at scale. With a small audience, Meta has limited room to optimise delivery, frequency climbs faster, CPMs rise as the system repeatedly targets the same people, and the learning phase takes longer to exit because conversion events are spread thin across fewer users. Below 200,000, most accounts see poor efficiency regardless of how relevant the audience seems on paper.
For cold prospecting specifically, audiences of 2 million or above give the algorithm enough room to learn and optimise. The sweet spot for most DTC brands in the US is somewhere between 5 million and 30 million — broad enough for the algorithm to find patterns, specific enough that you're not paying to reach audiences with genuinely no connection to your category.
There's a meaningful relationship between audience size and daily budget. A $50,000 per month budget pushing into a 200,000-person audience will saturate it quickly — frequency spikes, CPMs rise, performance deteriorates. The same budget into a 10 million-person audience has plenty of room to optimise without saturation.
A rough rule of thumb: for every $1,000 of monthly ad spend, you want at least 100,000 people in your target audience to avoid early saturation. This isn't a hard line but it's a useful starting point. When you're scaling spend and hitting diminishing returns, audience pool exhaustion is one of the first things to check before attributing the problem to creative or seasonality.
Lookalike audiences come in percentage sizes — a 1% lookalike in the US represents roughly 2.1 million people who most closely resemble your seed audience. A 5% lookalike is approximately 10.5 million. A 10% lookalike is around 21 million.
Tighter lookalikes (1 to 2%) used to outperform broader ones consistently. That gap has narrowed significantly as Meta's algorithm has improved. In many accounts today, a 5 to 10% lookalike performs similarly to or better than a 1% one, because the larger pool gives the algorithm more flexibility. Testing both is worth doing, but defaulting to 1% lookalikes as the obvious best option is a habit worth revisiting.
Advantage+ Shopping Campaigns effectively operate on the largest possible audience — Meta's entire eligible inventory. There's no targeting ceiling, which is part of why they tend to produce lower CPMs for accounts with strong purchase history. The algorithm has maximum flexibility to find buyers wherever they are across Facebook, Instagram, Reels, Stories, and Marketplace simultaneously.
The trade-off is control. If you need to exclude specific audiences, Advantage+ offers limited options. Strict exclusions — existing customers, specific locations, particular demographics that don't fit the product — are better managed in manual campaigns where targeting parameters are explicit. The audience flexibility that makes Advantage+ efficient also makes it less precise when precision matters.
Everything above applies to cold prospecting. Retargeting operates on small audiences by definition — website visitors, video viewers, add-to-cart abandoners are always going to be a fraction of your total addressable market. The rules for retargeting are different: smaller is acceptable because you're not trying to find new buyers, you're closing the ones who already showed intent.
The audience size consideration in retargeting is frequency management, not pool exhaustion. A warm audience of 40,000 people can be effectively retargeted at modest budget without saturation, as long as creative is rotated and frequency caps are watched. The problem isn't the small size — it's showing the same ad to the same 40,000 people until they start hiding it.
For how lookalike audiences work alongside size considerations, Lookalike Audiences on Meta: How They Work in 2026 covers the full strategy. For retargeting specifically, Facebook Retargeting Ads: How to Convert Warm Audiences Without Wasting Budget covers frequency management and audience segmentation. And for how Advantage+'s unlimited audience approach affects performance, Meta Advantage+: What It Is, How It Works, and When to Use It goes deeper on when to hand audience control to the algorithm entirely.
