A picture of a product sits beside a pair of running shoes. There is no athlete in the frame and the caption makes no promise about athletic performance. If you were counting references to exercise in cannabis advertising, would this one qualify? If you were counting explicit health claims, would you give the same answer? Such small decisions are the raw material of advertising research.
First define what an ad can be
An advertisement in a newspaper has edges, a publication date and often a named advertiser. A social-media post can be edited, reposted, answered or embedded in a longer thread. Researchers need rules for where one observation ends. Counting every appearance measures repetition. Counting each unique creative once measures variety. Either may be useful, but they will give different answers when one campaign runs the same image twenty times.
A content analysis of posts from five prominent cannabis companies examined 3,755 original posts from 2022. Alongside routine company news, the authors identified posts presenting products as answers to conditions including chronic pain. That is evidence about what the companies said. It supplies no clinical test of the claim. The choice of companies, platform and period establishes which material can enter the study. It does not establish what a typical person saw while scrolling. Someone who follows none of the companies may never encounter a post, while an especially active follower may see the same campaign repeatedly.
Give the second reader a rule
A codebook turns a broad question into observable features. In a four-state analysis of marketing requirements, researchers coded company posts against a set of specific criteria. The wording of each criterion determined what the coders could count. Researchers might separately record the presence of a price, an image of a product, a direct statement about health and an image suggesting a lifestyle. Those features can coexist. If a codebook forces a post into just one category, the study should explain how the reader is supposed to resolve a tie.
Now hand the running-shoe image to two coders working independently. If they disagree, the important question is whether a rule was unclear or the image genuinely ambiguous. Revising the rule on a practice sample can improve consistency. On the actual study sample, reporting agreement helps readers judge how dependable the category proved to be. A precise percentage built from a vague category is still a vague result.
The publication changes the view
A study of cannabis advertisements in free print publications in western Washington examined 305 ads in two publications in 2017. That sample captures a quite different slice of commercial communication. Print ads compete for a fixed area on a page. A social post may direct viewers to a profile or rely on a sequence of earlier posts for context. A difference between the two samples could reflect the medium, the place, the companies present or the years covered.
In particular, the number of advertisements is a different denominator from the number of firms. If one company produces half the posts, its vocabulary can dominate the results. An average across posts answers what a randomly selected post in the sample contained. An average across companies gives small and large advertisers more similar weight. Asking both questions can reveal whether an apparent industry pattern belongs chiefly to one prolific advertiser.
The audience lives outside the dataset
A content analysis can show that particular words or images appeared. Measuring how readers interpreted them requires readers. Measuring a change in behavior requires a design that can connect exposure with subsequent behavior while considering competing explanations. A review of research on cannabis advertising provides context for the different kinds of evidence in this area.
Return to the shoes. A content analyst can document their presence, the product beside them and the absence of an explicit performance claim. A reader might take the arrangement as an endorsement of exercise, a reference to a brand partnership or simply a visually effective arrangement of objects. Those interpretations are testable. The crucial editorial decision is knowing when to close the spreadsheet of advertisements and open a study of the people looking at them.