The settled mistakes playbook

• A rule is compressed experience. Ogilvy answers the objection that rules make dull work by pointing to the sonnet and the sonata. Both are strict forms that produced great art.
• Knowing what fails is a fence. Much of marketing still needs exploring. Some of it is already known not to work, and knowing which leaves more room for the exploring.
• The name changes. Rules became frameworks, then playbooks, and now skills in AI tools. What matters is whether evidence sits behind the name.

Takeaway: Learn what is already known first, so your tests go to questions that are still open.

• Ogilvy's list of what to test: the promise, the media, headlines and illustrations, ad size, frequency, spend and commercials.
• Most teams change five things at once. With pressure to do more with less, results get blended, and nobody can say which change moved them.
• Test markets put failure in a small place. Holdout groups and small submarkets are the modern version, and few teams use them.

Takeaway: Change one variable per test, and run it where a failure is cheap.

• Spend was the wrong variable. Ogilvy's statistician treated spend as the only factor, and missed that effective advertising at one-tenth the budget can outsell ineffective advertising.
• Some activity does harm. Repeated cold calls to the same accounts left a bad impression that lasted, and those prospects told others. Distasteful ads and negative talk about competitors work the same way.
• Incentives steer the advice. An agency paid on media spend leans toward advising more spend. Inside a company, attribution models do the same: people orient to whatever earns the credit.

Takeaway: Before adding volume, ask what the next unit is worth, and what the people who recommend it are paid to do.

• Ogilvy tested promises directly. He put different promises on the package, asked buyers to rank cards, and built a series of ads around one promise each, then counted orders for each.
• A feature describes the product. A promise describes what the buyer gets, and buyers choose between promises.
• Show the result and let buyers draw the conclusion. Apple's Shot on iPhone campaign showed photos and left the specifications for later.

Takeaway: Write several candidate promises before you write the copy, and put at least two in front of buyers.

• Believability. An oddly specific claim reads as true. "The best" reads as noise.
• Differentiation. If every competitor says "the best," the one that says how stands out. Ogilvy advertised facts that Shell and KLM shared with their rivals but that no rival mentioned.
• Expertise. In crowded markets, buyers want to know whether you can help them later. Specific detail shows you did the research.

Takeaway: Replace one superlative on your homepage with a fact only you can state.

• You see the campaign far more than your audience does. After three weeks you are tired of it, and most of your audience has not seen it yet.
• Keep winners in their own campaign. Pull the top-performing ads from each campaign into an evergreen one, so they keep running and stack up.
• Keep the next campaign in the ice box. When one is approved, start the next and hold it in reserve, so a flop does not cost a quarter.
• Brand lowers the cost of everything else. Price-off deals and lead generation aimed at this quarter's number fill the top of the funnel at the expense of later years. When I called an account, the reaction depended on whether they knew the company, not on the time of day or the script.

Takeaway: Run winners longer than feels comfortable, and keep the next campaign built.

• Nobody believes "number one in our category." Logo reels add some validation. Case studies add more when they name the company, the job title and the outcome.
• Buyers arrive asking "Are you for me?" They mean their role, their industry and their problem. Answer all three near the top of the page.
• People buy from people. Ogilvy advised showing the client's face and not the factory. Today that means a visible point of view and a named person behind the company.

Takeaway: Put a named customer result and a person's face where a visitor first lands.

• Judge the layout where it will run. Ogilvy's version was to see it pasted into the publication. Ours is to see it in the feed or in search results.
• One ad on every platform ignores the audience. People scroll LinkedIn, Reddit and YouTube in different frames of mind, so a result on one does not carry to the others.
• Look like what people came to read. Ogilvy found ads styled as editorial pages drew about 50% more readers. Factual case studies written as native posts and video testimonials are the current form.
• Give video a shape. In Gallup's data, commercials that set up a problem, showed the product solving it, then proved it sold to four times as many people as ones that only preached.

Takeaway: Before you publish, look at the asset where it will appear.

• Awards say little about results. Ogilvy noted that juries rarely see sales results, so they go by taste. Even inside a company it can be hard to learn whether a campaign worked.
• Data can be cover. Reports get used to support a decision already made. Judgment still has to decide what the data cannot, such as what image to build.
• Experts can say why. Ask why copy fails in reverse type. A real answer involves legibility and accessibility, and someone repeating a rule they heard cannot give it.
• Take rules with a grain of salt, and keep the reason. Ogilvy found photographs outsell drawings. With AI images everywhere, a drawing can now signal a human behind the work. The surface rule aged, and the psychology under it still holds.

Takeaway: Treat a rule as a hypothesis with a reason attached, and keep the reason when the rule ages.