In 1924 Kimberly-Clark put a new product on sale. It was made from Cellucotton, a creped wood-pulp material the company had produced during the First World War and had too much of afterward. Ironed flat and soft, it became Kleenex.
The company knew who it was for. Women were wearing more makeup, and Kleenex was positioned as a clean replacement for the “cold cream towel” in the bathroom. Its first advertised function, according to the brand’s history page, was a “marvelous new way to remove cold cream.” The first ad, in 1925, promised “the new secret of keeping a pretty skin as used by famous movie stars.” From 1927, stars such as Jean Harlow and Helen Hayes lent their names.
Then, in 1930, by the company’s account, Kimberly-Clark was swamped by letters from customers. The letters backed the tissue’s use as a handkerchief.
A question with an obvious answer
In 1954, Peter Drucker turned that kind of surprise into a question. In The Practice of Management, he devoted a chapter to a question he said is rarely asked clearly, because “the answer seems so obvious that it is seldom given”: what is our business?
The obvious answer comes from the factory. Kimberly-Clark made facial tissue for removing cold cream. Drucker’s answer comes from the other side of the counter. “What is our business is not determined by the producer but by the consumer,” he wrote. It is defined “by the want the consumer satisfies when he buys a product or a service.” And: “What the customer thinks he is buying, what he considers ‘value,’ is decisive.”
Your business is what the customer thinks they are buying, not what you make.
Here is the mechanism. A maker describes a product by what it is: the material, the size, the use it was designed for. A customer describes it by what it does for them on an ordinary day. Both descriptions are true. The second is usually the one that decides whether it gets bought again.
Someone inside already knew
Read the Kimberly-Clark history closely and the turn is less sudden than it looks. Ernst Mahler, the company’s head of research, had hay fever and had been using the tissues as handkerchiefs. He persuaded the advertising group to pitch them, too, as a way to avoid spreading germs, and in 1927, the brand page says, the company proposed “absorbent kerchiefs.” The handkerchief idea was in the building. It was one message among several, under the movie stars.
What the letters added was weight. After more research, the company switched its advertising to “the handkerchief you can throw away,” aimed at “men, women, young and old alike.” Sales doubled in the first year of the new campaign, according to the company. That is the company’s own history, written well after the fact, and it doesn’t say how many letters arrived or what the additional research found.
The question has a limit, too. Asking customers works for a product they already have in their hands. It works worse for one they have never seen: one survey firm, by the account of Haloid’s research chief, told the company it could expect to place no more than five thousand of its plain-paper copiers in American offices. That is a different job, creating a customer.
The dashboard that became a spreadsheet
Imagine you run an internal dashboard team. Your dashboard was built for regional managers to track sales against target, and the usage numbers are healthy. Then someone shows you the export logs. Most visits end in a download to a spreadsheet, where analysts rebuild the one table they need and paste it into a weekly email.
Your roadmap has six new charts on it. Your customers are buying one table.
Letters nobody writes
Kimberly-Clark’s customers wrote letters. Most people using an AI assistant don’t, and privacy concerns and practical problems, Anthropic’s researchers wrote, make their conversations hard to study. In December 2024 they described Clio, a system that uses AI models to analyze conversations and surface patterns, without human reviewers reading the raw chats.
Run on a random sample of one million conversations from Claude.ai’s Free and Pro plans, Clio found coding at the top, with web and mobile app development over 10 percent of conversations, and writing, research and education each at 6 to 10 percent. It also surfaced smaller clusters, among them interpreting dreams and acting as a Dungeon Master for tabletop role-playing games. The authors say Clio is meant to find “unknown unknowns.”
One finding is the Kleenex story in miniature. Anthropic’s safety classifiers, the paper reports, often flagged conversations about Dungeons & Dragons creatures’ combat stats as harmful. The system had a picture of what those users were doing. The users were playing a game.
The categories Clio reports are produced by a model, and business accounts were left out, so the shares describe consumer use only. Before you trust anyone’s usage dashboard, yours included, read twenty raw requests yourself and sort them by hand. Then compare your piles with the dashboard’s labels.
Where they disagree, you are holding Kimberly-Clark’s letters.