Company

What a million apps taught us

Nest has now been used to build a million apps. Here is what we learned about the apps that succeed, how people actually build and what we got wrong.

Nadia Mensah & Clara Novak

A building facade covered in a mosaic of small colored tiles against a blue sky
Photo: Jan van der Wolf / Pexels

Last week, someone in Porto published the millionth app built on Nest. It was a sign-up sheet for a youth football club. That felt about right, and it says a lot about what people actually build.

A million apps is enough to see patterns that are hard to spot one project at a time. Clara and I spent the last few weeks going through anonymized, aggregate usage data with the team. We never look inside people’s projects for this kind of work, and prompts and code are never used to train models. These are the lessons that stood out.

Most successful apps are small

When we started Nest, we assumed the apps that stuck around would be the ambitious ones. The opposite is true.

Among apps still in active use six months after they were published, the median has 7 pages and 5 database tables. More than half have a single kind of user. They do one job well: take bookings, track equipment, collect applications, run a rota.

The large, sprawling projects are real, and some of them are impressive. But they are much more likely to be abandoned before they ever go live. Small scope is not a limitation people put up with. It is the most reliable predictor that an app will be used.

People build in many small messages

The second surprise was how people iterate. We expected a few long, detailed prompts per project. What we see instead is many short ones.

  • The median published app took 38 messages to build.
  • The median message is 19 words long.
  • Projects where the average message was under 30 words were 1.6 times more likely to be published than projects built from long prompts.

Short messages are easier to review, easier to roll back and easier for Nest to get right. We have since redesigned the composer to encourage this, and we now suggest the same rhythm to everyone: a clear start, then small steps.

A plan prevents rework

Earlier this year we started testing a step where Nest writes a short plan before making a larger change: the tables it will add, the pages it will touch, the decisions it needs from you.

The effect was larger than we expected. For changes that touched three or more files, showing a plan first reduced follow-up messages that undid or corrected the change by 41%. About one plan in five was edited by the person before Nest started, usually to reuse something that already existed or to cut scope.

A plan costs a few seconds to read. Rework costs far more. This is now central to how we are designing the agent work we will share later this year.

Internal tools are the quiet majority

Public launches get the attention, but they are not most of what people build. Roughly 58% of active apps are internal tools: used by a team inside a company, behind a login, never seen by the public.

What they replace

When people describe what the app replaced, the answers are remarkably consistent:

  1. A spreadsheet that had grown too large to trust.
  2. An email or chat thread used as a queue.
  3. A paper form.
  4. An expensive tool where the team used a small fraction of the features.

These apps rarely make headlines, but they tend to last. Internal tools have the highest six-month retention of any category we track. If you have one of those spreadsheets, our page on internal tools is a good place to start.

What we got wrong

Not everything we believed survived the data.

  • We overbuilt the template gallery. Most people start from a blank prompt, and the templates that do get used are the simplest ones. We have trimmed the gallery to the templates people actually choose.
  • We underestimated export. Few people export their code, but knowing they can is one of the most common reasons teams give for choosing Nest. It stays a first-class feature.
  • We assumed founders would outgrow us. Many do hire engineers, but they mostly keep building on Nest, with developers working through GitHub sync alongside them.

The lesson running through all of this is simple. People succeed with Nest when they build something small, iterate in small steps and agree on a plan before big changes. Our job is to make each of those easier.

Thank you to everyone who built one of the first million. We are looking forward to seeing what the next million look like, and we suspect many of them will be sign-up sheets for football clubs.

#company#insights
Portrait of Nadia Mensah

Written by

Nadia Mensah

Co-founder and CEO

Portrait of Clara Novak

Co-author

Clara Novak

Head of Product

A team cheering around laptops at a meeting table in a bright office

Company

Our $48M Series B

Nest has raised a $48M Series B led by Harbourline Ventures. Here is where we are, what the money is for and what stays the same.

Streaks of purple and green light curving through a dark frame

Product

Introducing Agent mode

Nest can now plan a change, make it across your whole project, test it in a real browser and fix what breaks before handing the app back to you.

Type to search every post.