Visitors Followed the Crowd—even When the Crowd Was Stealing.

At Arizona’s Petrified Forest National Park, 7.92% of marked wood disappeared after visitors saw a warning that many others had taken pieces; only 1.67% disappeared under a direct message of disapproval. The warning meant to deter theft had made theft look normal. For marketers, that changes a basic creative decision: if a behavior is harmful, do not advertise its popularity while asking people to stop.

The message advertised the behavior

The 2006 Social Influence study by Robert Cialdini and colleagues was conducted on real park paths, not in a survey. Across five weekends, 2,655 visitors encountered experimental signs. Researchers placed 20 marked pieces of petrified wood along each path during two-hour blocks, then counted how many disappeared.

One sign stressed a descriptive norm: many past visitors had removed wood. Another stressed an injunctive norm: taking wood was disapproved. The strongly worded descriptive message produced 7.92% theft, versus 1.67% under the strongly worded injunctive message. The U.S. Forest Service archive summarizes the result plainly: focusing people on what others did increased theft; focusing them on disapproval suppressed it.

The message did not endorse stealing. It did something subtler. By making the misconduct vivid and common, it supplied behavioral evidence: people like you do this here. The moral instruction and the social signal pointed in opposite directions, and the social signal won.

“Is” and “ought” pull different levers

A descriptive norm answers, “What do people actually do?” An injunctive norm answers, “What do people approve?” They can support each other, but they are not interchangeable.

Descriptive norms are efficient. In an unfamiliar setting, watching others can reduce uncertainty about what works and what belongs. That is why bestseller shelves, subscriber counts, queues, review volume and “most popular” labels can work—when the evidence is truthful and relevant.

The danger appears when the same mechanism is attached to a problem. “Thousands of customers abandon setup,” “too many people ignore renewal reminders,” or “returns are at a record high” may dramatize urgency internally. In customer-facing copy, each can imply that abandoning, ignoring or returning is the normal response.

A 2019 meta-analysis by Vladimir Melnyk and colleagues covered 297 studies. Wageningen University’s research record reports that descriptive norms affected behavior more directly, while injunctive norms were more closely related to intentions; specificity, source and audience characteristics moderated the results. Social proof is therefore not magic wording. It is evidence people use to infer the safe, common or sensible move.

Alarm is not proof

Marketers often publicize the size of a problem because large numbers feel persuasive. But problem magnitude and desired behavior are different facts. A security campaign can truthfully say that millions reuse weak passwords and still make weak passwords sound ordinary. A retention email can say “customers keep forgetting” and accidentally grant permission to forget.

The correction is not to hide evidence or invent a happier majority. It is to select the norm that supports the decision. If verified, show how many customers complete setup, choose the durable plan, renew on time or return packaging. If the preferred behavior is not yet common, use a clear injunctive cue, explain the benefit or show progress without pretending a majority exists.

Limits matter. The park experiment tested a particular behavior in a particular place, and later syntheses show that source, relevance and context change the size of norm effects. A global statistic may be useless to a local buyer; a peer-group norm can be more credible. Test observed behavior—completion, renewal, complaint, return—not just whether people say they liked the message.

Make the preferred behavior easy to see

For a small business, this may mean replacing “we get many questions about booking” with a truthful cue such as “most clients choose a 30-minute first call,” if the data supports it. A midsize company should audit onboarding, support and cancellation copy for sentences that make failure sound routine. A large organization should segment norm claims so the comparison group is recognizable rather than statistically impressive but socially distant.

OrionPilot’s published methodology connects business context and strategy with creative execution, connected data, analytics and continuous learning. That matters here because a norm claim should remain attached to the evidence that justified it, then be judged by the behavior it produces across channels—not by how dramatic the sentence sounds in isolation.

The transferable decision is simple but demanding: before publishing any popularity claim, ask which behavior it makes normal. If the answer is the behavior you want to reduce, the message is advertising the wrong crowd.