Every performance marketer sizing a first CTV test eventually asks how much it should cost, and the honest answer changes completely depending on where the campaign runs. A single global CPM number, used as the benchmark for every market, is often what gets a promising test judged unfairly. Regional variance in CTV pricing is planning information, not noise to average out, and it belongs in the conversation before a campaign starts rather than in the postmortem after it.
This piece breaks down how CTV pricing differs market by market, using the regional CPM benchmarks from Singular & YouAppi’s 2026 CTV Growth Playbook. It covers what drives that variance, how a first test typically gets structured across a 90-day window, and the mistakes that most often break a regional read.
The starting point is a simple table, although reading it correctly takes more care than the raw numbers suggest. Each region carries a different story about market maturity, inventory quality, and what a team should reasonably expect to learn from the spend.
YouAppi's benchmark data, contributed to the 2026 CTV Growth Playbook built together with Singular, lays out five of them:
Read as a set, the five benchmarks suggest a pattern worth noting. Markets with more mature programmatic supply, stronger broadcaster inventory, and more developed measurement infrastructure tend to command a higher CPM. The playbook's own regional notes point in that direction:
A Japan test at $20 to $25 CPM is not a worse deal than an APAC test at $3 to $7 CPM. It is a different strategic bet, and a lower CPM does not automatically mean a better campaign: success is measured by quality, efficiency, and incremental value, not by cost alone.
Regional CPM sets the price of a single impression. It says nothing about whether a test has run long enough or wide enough to produce a reliable read.
Practitioners cited in the 2026 CTV Growth Playbook point to twenty five thousand dollars per month as the level where a test typically starts to accumulate the kind of learning a smaller test does not. Below that level, results tend to read as noise rather than a clear signal, since delivery has not reached enough volume to separate real performance from statistical variance.
The complete phase-by-phase spending framework in the playbook lays out how test delivery is typically structured across a full ninety-day window, from broad learning to a causal read on incremental ROAS.
Spend is not typically flat across a test window, since a level budget applied evenly across ninety days tends to waste the same learning phase twice.
Days 1 to 30, launch and learn.
Targeting stays broad while the model works across several genuinely different creative directions. Attribution settings need to be active before delivery starts, and the metrics that matter during this phase are limited to delivery health: completion rate, frequency, and early CPI, with first read-outs typically landing around day three to five.
Days 31 to 60, optimize on signal.
Differences between audiences and creative begin to separate, and typical practice shifts delivery toward what is working rather than what launched first. Frequency caps tighten as fatigue appears, and spend on clear winners tends to scale gradually, at roughly twenty to thirty percent per week, rather than all at once.
Days 61 to 90, prove and decide.
A geo or holdout incrementality test runs with a pre-set duration. CTV typically needs four to six weeks of active flight plus a two to four week post-treatment window, since a meaningful share of conversions land after exposure ends.
fail to produce a clear answer if the campaign gets read the wrong way. Three mistakes account for most of it.
The first is judging a test against the wrong mental model, since measuring CTV like a last-click mobile channel undervalues it and measuring it like traditional TV overvalues it. The second is moving too fast, pausing or adjusting a campaign before attribution cohorts have had time to complete. The third is comparing CTV partners without standardizing attribution windows and MMP configurations, which ends up comparing different measurement setups rather than different performance.
Read against the wrong model, window, or setup, even a well-delivered CTV campaign can look like it failed.
CTV pricing is not a single number. It moves with market maturity, inventory quality, and measurement infrastructure, and reading a benchmark outside that context tends to produce the wrong read on whether a channel is working.
A few questions help translate a regional benchmark into context:
The 2026 CTV Growth Playbook, built together with Singular, covers the complete regional benchmark set, the full ninety-day framework, and the attribution setup that determines whether any of these numbers get read correctly. Download it here.