Marketing is full of rules that are presented as if they work everywhere: keep headlines short, post at the right time, use a strong CTA, follow the format that’s already performing well. These recommendations can be useful, but the problem starts when a best practice becomes a substitute for thinking.

What works for one brand can easily fail for another. A B2B company with a six-month sales cycle has very different content needs from a consumer app that people can buy in seconds. An established brand can communicate differently from a company that is still building trust. Even two businesses targeting the same audience can get very different results from the same tactic.

This is where marketing teams often go wrong. They see a competitor’s successful campaign, notice a popular format, or read that a certain tactic “converts better” and immediately try to reproduce it. But copying the tactic isn’t the same as understanding the strategy behind it. A campaign may have succeeded because of its audience, offer, timing, brand reputation, or something that had nothing to do with the tactic everyone noticed.

Best practices are still useful as starting points. If short-form video is working across your industry, test it. If successful companies are using a particular landing page structure, experiment with it. But then look at your own results: which message gets people to act, where do they drop off, what objections keep appearing, and which ideas consistently perform better?

Over time, your own data should become a better source of guidance than a generic list of marketing rules.

AI makes this even more relevant. It can instantly generate “best” hooks, CTAs, content formats, and strategies, but the fastest answer isn’t necessarily the right one for your audience. If everyone follows the same recommendations, marketing becomes more predictable—and much harder to differentiate.

The better approach is simple: borrow ideas, not conclusions. Understand why a tactic works, test it in your own context, and keep it only if the results support it.