If the first version doesn't embarrass you, it shipped too late

Reid Hoffman has given founders this advice for years, and he has the record to back it. LinkedIn, which he co-founded after his time as an executive at PayPal, went live in May 2003 as little more than a place to post a work history and invite colleagues. Much of what people now use it for came later, shaped by what early members actually did on the site. The exact occasion on which he first put the idea into this sentence is not documented; it has become one of the most repeated lines in Silicon Valley.

The argument underneath is about learning. Until real users touch a product, every decision about it is a guess. Each extra week of polish before launch is a week of guesses stacked on guesses, and some of the polish will go into features nobody wants. Launching early swaps guesses for evidence. The embarrassment is simply what a team feels when it looks at the gap between the product it imagines and the one it could ship now, and Hoffman’s point is that feeling this gap is normal. A team that feels no embarrassment has probably closed the gap by waiting, and paid for it in time and in lessons not learned.

Gmail is a good example of the practice. Google released it in 2004 with a beta label and left the label on for more than five years while millions of people used it daily. The early product lacked features competitors had, and the beta label gave Google room to learn which ones mattered.

The sentence is easy to misread, and the misreading is expensive. The embarrassment it describes should come from scope: missing features, rough design, a narrow first audience. It should not come from quality in the sense of lost data or failed payments. A thin product that works earns patience. A broad product that breaks earns distrust, and distrust is much harder to undo than a missing feature. Hoffman’s own background shows why: a payments company cannot ship a version that loses money and learn from the complaints.

Some fields do not allow the rule at all. Medical devices, aircraft parts and anything handling other people’s money or safety have to be right the first time they are used, and the early learning happens in testing rather than in public.

The rule also depends on who is watching. Cuil launched in July 2008 with heavy press attention and a claim to rival Google, and its thin results became the story. The first reviews stayed attached to the name long after. An embarrassing first version works when it goes to a small group who expect roughness and will say what they need. It works badly as a public event.

Read that way, the line is sound advice: ship something small that works to the people most likely to forgive it, and let them tell you what to build next.