Two languages, one channel
Connected television is sold with the language of broadcast — reach, share, premium content, co-viewing — and measured with the language of digital: impressions, completion rates, unique devices. The two do not reconcile, and the gap between them is where planning errors live.
A completed impression on a shared living-room screen is not one exposure. A unique device is not a unique person. And a completion rate near 100% tells you the format is non-skippable, not that anyone watched.
The co-viewing problem
Co-viewing is the single largest source of error. Depending on daypart and content genre, the ratio of viewers to devices in South African households varies enough that ignoring it can understate delivered reach by a meaningful margin in prime family viewing and overstate frequency badly in solo viewing occasions.
Most CTV reporting simply does not model this. Applying a flat co-viewing factor is better than nothing but still crude. We plan with genre and daypart-specific factors and treat the output as a range rather than a point estimate.
A workable framework
Plan CTV as broadcast, verify it as digital, and reconcile the two explicitly rather than pretending one of them is the truth. In practice that means three things.
Set reach and frequency goals in broadcast terms against a person-level universe. Buy and verify at the impression level with independent verification on viewability and fraud. Then reconcile with a co-viewing adjustment you can defend, and report the range, not a false-precision single number.
Finally, insist on deduplicated reach across CTV and broadcast. Planning them in separate silos is how advertisers end up paying premium CPMs for frequency against people the broadcast schedule already reached six times.