Why CPMs rise as you scale
Meta's auction is a competition for attention. When you increase budget, you are buying more of the same audience faster. The algorithm exhausts your best-performing segments first, then moves to progressively less responsive users at progressively higher CPMs. A CPM of 8 euros at 5,000 euros monthly spend often becomes 18 euros at 30,000 euros monthly spend targeting the same audience.
This is not a bug in the platform. It is the auction mechanics working exactly as designed. The audience that responds to your creative at low CPMs is the low-hanging fruit. Scaling means paying more to reach people who are less likely to respond.
Creative refresh rate is the real scaling variable
The most effective CPM control at scale is not audience expansion or bid adjustments. It is creative refresh rate. Meta's algorithm enters a state called creative fatigue when the same creative is shown to the same person multiple times without a conversion. The algorithm deprioritizes fatigue-heavy creatives in the auction, which increases your cost to maintain impression volume.
The threshold varies by format and audience size. For a DTC brand targeting a broad audience of 2 to 5 million users in the Netherlands and Belgium, creative fatigue typically sets in after 3 to 5 impressions per user per week. Track frequency at the creative level, not the campaign level. A campaign-level frequency of 2.5 can mask individual creatives running at 7 or 8 frequency with severely degraded performance.
Advantage Plus Shopping versus manual campaigns at scale
Meta's Advantage Plus Shopping (ASC) campaigns use full automation to select audiences, placements, and creative combinations. At lower spend levels, manual campaigns with tight audience control often outperform ASC. Above 20,000 euros monthly spend, the opposite tends to be true. ASC has more signal to work with and makes better decisions than manual targeting at that budget level.
The transition point varies by catalog size and first-party data quality. Brands with fewer than 50 SKUs and limited purchase history in Meta's pixel should stay manual longer. Brands with hundreds of SKUs and multiple years of purchase data flowing through CAPI should move to ASC sooner.
Broad targeting is not giving up control
Meta's broad targeting, zero interest or behavioral targeting with only age, gender, and geography, outperforms interest-based targeting for most DTC brands above a certain spend level. This feels counterintuitive but the math is straightforward: Meta's algorithm is better at finding your customers than you are, as long as it has strong conversion signals to learn from.
The prerequisite is clean server-side data. Broad targeting with poor conversion signal data is genuinely giving up control. Broad targeting with 90-plus percent event match rates from dual-dispatch CAPI is giving the algorithm the signal it needs to find your best customers at scale.
Measurement at scale requires different reporting
At low spend, you can read ROAS in the Meta dashboard and make reasonable decisions. At high spend, platform-reported ROAS is contaminated by creative fatigue, spillover from other channels, and attribution window choices that favor Meta. The operational metrics that matter at scale are CPM trends by creative, new customer acquisition cost versus repeat customer acquisition cost, and contribution margin per order by channel.
Report these weekly. Campaign ROAS is a lagging indicator. CPM trajectory by creative is a leading one.
---