Customers can judge a product, not invent it
Surveys ask people to describe the future in terms of the present. That is the narrow, defensible point inside a line that is usually quoted as a boast.
It isn’t the consumers’ job to know what they want.
Asked how much market research went into Apple’s products, Steve Jobs gave a one-word answer, “none,” and followed it with this sentence. It sounds arrogant. Read closely, it is a claim about the limits of one method, and a narrower claim than it first appears.
Market research asks people to describe their needs in terms of things that already exist. That works when the category is settled. A buyer of washing machines can say whether a quieter spin or a bigger drum matters more, and a survey will find the answer. It works badly when the product would create a category. Before 2001 almost nobody asked for a pocket device that held a thousand songs, because few people thought of a record collection as something that could travel. Before 2007 few phone owners asked for a phone with no keyboard. Asked directly, they would have described a slightly better version of what they already had.
That is the sense in which it is not the customer’s job. The customer’s job is to live with a problem and to judge whether a solution solves it. Imagining the solution belongs to whoever builds it. The line rejects asking people to design the next thing; it does not reject watching how they use the current one.
It also carries a price that admirers tend to skip. Apple could work this way because it had rare design talent, a leader with a long record of taste, and the money to absorb a miss. Misses happened. The Power Mac G4 Cube, launched in 2000 as a beautiful object nobody had requested, sold poorly and was put on ice within a year. Taste is a bet, and most people who quote the line are not placing it with Apple’s odds.
For a small company the rule applies to one stage of the work. Surveys and focus groups are a poor way to discover a new product. Watching people struggle with an existing one is a good way to find the problem, and putting a rough version in front of them is the only honest way to learn whether the answer works. People are bad at predicting what they will want and very good at showing what they actually use. A founder who stops watching for that has mistaken a method for a license.