How big should a new product’s launch be?

The question behind the question

When a company creates a new category, it must decide how much to promise and to whom. A broad launch captures attention, signals ambition and may pre-empt rivals. A narrow launch tests whether anyone wants the thing. The record shows a striking difference between the two, and it is one of the clearest patterns on the site: the pattern a small working thing beats a grand design.

When starting narrow worked

M-Pesa, launched by Safaricom in Kenya in 2007, began as a way for customers of a microfinance institution to repay loans. Users quickly found their own use, sending money to family, and the service grew into a national payment system. Facebook required real names and college email addresses and opened one campus at a time, so that each campus reached critical mass before the next opened. Workday built human resources software for large enterprises as an online service, a narrow, demanding market where its approach was clearly better, before widening into finance. Amazon’s fulfilment service of 2006 started by storing and shipping goods for independent sellers and grew into the backbone of its marketplace. OpenAI released ChatGPT in 2022 as a free “research preview” of an existing model, with modest expectations, and let its users show what it was for. Tesla’s founders wrote in 2006 that they would start with an expensive, low-volume sports car and use the money to fund cheaper, higher-volume cars.

When launching big failed

Google Wave was unveiled in 2009 as a new category meant to unify email, chat and documents, in an invitation-only preview that left early users unsure what it was for; Google stopped developing it as a standalone product about fifteen months later. Motorola developed Iridium, a global satellite telephone network, for business travellers and others outside terrestrial coverage, a vision that cost billions before launch; by the time service began in 1998, mobile networks had spread, and the handsets cost about $3,000. Apple launched the Newton in 1993 as a new category of pen-based personal assistant with handwriting recognition that did not work well enough. IBM turned Watson’s quiz-show victory into a business unit aimed at cancer care and other high-stakes fields, promising more than the technology could deliver.

These failures share a structure: the launch was sized to the vision. The product promised a whole new way of working to everyone, before evidence showed that any particular group wanted it enough to change.

A diagnosis to start from

The diagnosis starts with who has the problem badly enough to use an imperfect first version. If nobody can be named, the launch is a bet on a vision rather than a response to demand. The second question is how much the launch commits the company irreversibly: a satellite constellation must be built before the first customer, while a web service can start small. The more irreversible the investment, the more evidence it needs first, the idea explored in Irreversibility and the Value of Waiting.

The guiding policy is to launch at the scale of evidence, choose a first group connected enough that the product can spread within it, and widen as users show what they value. The coherent actions include a deliberately narrow first market, a product that does one job well, and a plan for learning that can change the product quickly.

The strongest counter-case

The strongest objection is that some categories cannot start small. Networks need many users to be valuable; infrastructure needs scale before it works at all; and a narrow launch can let a better-funded rival take the broad market. The iPhone launched to everyone at once. GPS needed a constellation of satellites before anyone could use it. And starting narrow can trap a product in its first niche.

The objection identifies real exceptions, but they share a feature: the launch was big, but the product’s job was narrow and clear. The iPhone combined a phone, an iPod and an internet device that already existed in people’s pockets; it asked users to do familiar things better, not to adopt a new way of working.

What the wider research says

Geoffrey Moore’s Crossing the Chasm (1991) argued that new technology products should win a narrow beachhead of pragmatic customers before expanding, because early enthusiasts and mainstream buyers want different things. Clayton Christensen showed that disruptive products start in small markets that incumbents ignore. Eric Ries’s The Lean Startup (2011) turned the idea into a method of building, measuring and learning in small steps. Kahneman and Lovallo’s account of bold forecasts explains why large organisations launch big anyway: planners anchor on the vision and neglect the base rate of new products, most of which fail.