It is not possible to isolate variables when studying a large population being exposed to public media.  Therefore, although a correlation between exposure to the campaign and change in behaviour can be observed, cause and effect cannot be determined.

Levels of exposure to the media can only be measured through self-reported data.  It is unlikely that the average person can accurately determine the number of times that they have been exposed, either directly or indirectly, to the campaign.

Sampling is often problematic, so the results may not reflect the diversity of the population.

Even if a public health campaign is judged to be effective, often the findings are not transferable to other populations.

As noted above, it is difficult to measure the effectiveness over a large period of time to determine if the results are simply short-term change or a durable, long-lasting change in behaviour.

Simply measuring health outcomes through hospital or work related data is not enough to determine that the campaign itself made a difference.

Even if a campaign fails, it is difficult to know whether it failed because of the actual campaign or the way that it was delivered.

Much of the data obtained is self-reported, leading to potential demand characteristics influencing the final outcome.

Who is doing the research is very important. Often those that design the program play an integral role in its evaluation, potentially leading to a lack of objectivity and researcher bias.

Strong evaluation requires triangulation.  This is time-consuming and expensive.