Skip to the page

Plainly Put

One sentence about building things, taken apart

HomeProduct

Build the core idea, test it, and only then add to it

Pownce launched doing four things at once. The lesson its co-founder drew is sound, with a couple of honest exceptions.

Learn not to add too many features right away, and get the core idea built and tested.

Leah Culver

Pownce opened in 2007 as an invitation-only service where friends could send each other messages, links, files and event invitations, all from the same box. Leah Culver built much of it as a co-founder with Kevin Rose, Daniel Burka and Shawn Allen. Twitter, which launched the year before, did one thing: short public updates. Six Apart bought Pownce in December 2008 and shut it down within weeks, and Twitter kept growing.

Culver has since started Convore, a group chat service that went through Y Combinator in early 2011. Talking to Women 2.0 in July, in a short interview on her startup lessons, she named feature scope as one of the three things she had learned, alongside how emotionally draining a first startup is and how long adoption takes. The instinct she was describing is an engineer’s: when building is the easy part, building more feels like progress.

The sentence has two verbs, and both matter. “Built” means finishing the central thing to a standard where people can use it for real. “Tested” means putting it in front of users and learning whether the central thing is wanted before surrounding it with extras. Every extra feature added before that test makes the result harder to read. If people ignore a product that does four things, it is unclear which of the four they rejected; if they ignore one that does a single thing, the answer is plain.

Extra features also cost more than the time to build them. Each one has to be explained on the home page, supported when it breaks and carried through every later change to the code. A small team that ships five features has signed up to maintain five features, usually before it knows which one pays the bills.

The advice has two honest exceptions. Some products are only useful as a bundle. A marketplace needs listing, search, payment and messaging before a single trade can happen, and payroll software that handles most taxes but not all of them is useless to a business owner. In those cases the core is itself large, and cutting features means cutting the product.

The second exception is about diagnosis. Pownce’s trouble may have had less to do with its feature list than with the fact that Twitter reached the crowd first, and a network with the crowd is hard to leave. A founder reading the lesson should be careful not to blame the feature count for a problem that was really about timing and network effects.

Within those limits, the rule is one of the most useful in early product work. Decide what the one essential action is, make it work well, watch people use it, and let their behavior pick the second feature.