A native campaign running identical creative across a supply pool produces wildly different results by site. The variation is larger than targeting or format can explain, and it comes from context.
Mindset on arrival differs by publisher
Someone reading a news report is in a different state from someone reading a recipe or a product review, and that state persists into what they do next.
A reader who arrived to solve a problem is receptive to a solution. A reader following a news story is not shopping.
Native ads are served after the content, so they inherit that state rather than creating it.
Adjacency shapes interpretation
The surrounding article sets the frame within which a headline is read. The same words carry different implications after a personal finance piece and after entertainment coverage.
Where the adjacency aligns, the ad reads as a continuation and performs above its average.
Where it clashes, the ad reads as an intrusion, and the click rate falls even though nothing about the creative changed.
Audience composition is not the same as audience definition
Two sites nominally reaching the same demographic will differ in income, intent and familiarity with the category in ways demographics do not capture.
Targeting parameters describe who a person is. Site choice captures what they were doing, which predicts response better in most categories.
This is why a campaign optimised purely on audience attributes usually underperforms one that also curates supply.
Traffic quality varies underneath everything
Some inventory is reached by readers who arrived deliberately, and some by readers routed through paid acquisition or endless slideshow pagination.
Those visitors behave differently on arrival and convert at a fraction of the rate, regardless of how well the ad is matched.
The difference is invisible in click reporting and obvious in post-click behaviour, which is why site-level performance has to be judged after the click.
What to do with the variation
Site-level reporting is available in every native platform and is the highest-value report in the account, because the spread between best and worst sites is usually enormous.
Concentrating spend on the sites that convert, rather than the sites that click, changes campaign economics more than any creative iteration.
The list needs revisiting regularly, since supply composition shifts as publishers change their own monetisation and traffic sources.