200,000 taste tests, one mistake

In 1985, The Coca-Cola Company ran 200,000 taste tests, hired scientists with clipboards, and concluded with statistical confidence that people preferred the new formula.

Then they launched it, and the American public responded as if Coca-Cola had personally driven to each of their childhood homes and urinated on the porch.

There were protests. Petitions.

A man in Los Angeles spent $1,000 stockpiling original Coke before it even had a name to stockpile under. A senator from Arkansas gave a speech on the Senate floor about a soft drink.

Within 79 days, Coca-Cola reversed course and rereleased the original as Coca-Cola Classic. They put "Classic" on it. As if the thing that existed for ninety years had simply leveled up.

𝗧𝗵𝗲 𝗮𝗰𝘁𝘂𝗮𝗹 𝗹𝗲𝘀𝘀𝗼𝗻: stated preference is what you say when asked where to eat. Revealed preference is where you actually go.

200,000 tests can't close that gap.

𝗧𝗵𝗶𝘀 𝗶𝘀 𝗮 𝗽𝗼𝘀𝘁 𝗮𝗯𝗼𝘂𝘁 𝘁𝗵𝗲 𝗰𝗼𝗺𝗽𝗮𝗻𝘆 𝘁𝗵𝗮𝘁 𝗽𝗿𝗼𝗯𝗮𝗯𝗹𝘆 𝘄𝗶𝗹𝗹.

The agent is coming. Actually coming. You give it one thing: a sentence. "Find me something like Patagonia but cheaper." That sentence is where all the commercial action has migrated.

The company that shapes what you say before you say it owns something that does not have a name yet. 𝗡𝗼𝗯𝗼𝗱𝘆 𝗵𝗮𝘀 𝗯𝘂𝗶𝗹𝘁 𝗶𝘁.

Apple holds your payment credentials, health data, location history, and the 47 apps you open daily. The problem: "we protect your data" and "we built the world's most valuable preference platform" cannot occupy the same org chart. 𝗕𝗲𝘀𝘁-𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻𝗲𝗱 𝗰𝗮𝗻𝗱𝗶𝗱𝗮𝘁𝗲. Most likely to sit on it for six years and release it as an Apple One subscription tier.

Spotify solved the hardest problem in market research by refusing to do market research. Then once a year, show people what you saw and make it shareable. 600 million accounts. I say this as someone who shared their Wrapped card last December: what we handed over was a taste map. What we played at 2am on a Wednesday. What we played seventeen times without adding to a playlist because it would have made it official, and we were not ready. 𝗔 𝗯𝗿𝗶𝗱𝗴𝗲 𝗻𝗼𝗯𝗼𝗱𝘆 𝗵𝗮𝘀 𝗯𝘂𝗶𝗹𝘁.

Stripe we covered yesterday. 𝗧𝗵𝗲 𝗱𝗼𝘄𝗻𝘀𝘁𝗿𝗲𝗮𝗺 𝗵𝗮𝗹𝗳 𝗼𝗳 𝘁𝗵𝗲 𝗺𝗮𝗽 𝗻𝗼𝗯𝗼𝗱𝘆 𝗵𝗮𝘀 𝗯𝘂𝗶𝗹𝘁.
Three companies. Does that close a gap that has existed since 1985?

No. Obviously no. Come on.

Amazon built the recommendation engine in 2001. It is a golden retriever. The preference platform is the person behind you who noticed which direction you looked before you threw, and has quietly rearranged the entire yard. It precedes your preference.

The company that builds this will not run 200,000 taste tests.

They will not ask. They will know.

Before you've played that song seventeen times at 2am and decided you're not ready for it to be official.

We will name them something that sounds like a project management tool.

We almost always do that with the important ones.