Sell New Products to Existing Customers

A sequencing rule for launching risky products

Pedro Franceschi, co-founder and CEO of Brex, describes a go-to-market rule for shipping innovative products from inside an established company as a constraint on how many unsolved problems a launch may carry at once. He states it as a mantra: "You can sell a new product to your existing customers, you can sell an existing product to a new customer, you cannot sell a new product to a new customer."¹

The reasoning attributed to him is that any launch requires either a working product muscle, meaning you can actually build the thing, or a working go-to-market muscle, meaning you can find and convert the buyer. A brand-new product sold to a brand-new customer forces a company to solve both at the same time, which he judges too hard. The conclusion he draws is to over-index on selling the most innovative bets to the existing base first, where distribution and trust are already in place and only the product remains uncertain.

Co-design with the installed base

In practice, Franceschi describes fusing the rule with a co-design method at Brex. Every new product is built with a cohort of ten to thirty existing customers, often the most AI-forward ones, so that even an alpha version reaches real go-to-market traction rather than launching cold.¹ The cohort is deliberately wide in scale, spanning small startups up to very large public companies, because Brex holds that a customer needs one platform across its entire growth journey, in the pattern he attributes to AWS and Stripe, where the same vendor serves a company from two people to the Fortune 50. The through-line he names is finding the primitives that scale across the whole journey.

How it fits the rest of the doctrine

The pattern functions as the distribution counterpart to building new products in a separate unit: the building can be refounded away from the core, but the launch goes into the installed base, so the incumbent's real advantage in relationships, trust, and data does the work a cold go-to-market cannot. It also disciplines what gets built, since the new product has to be adjacent enough that existing customers already feel the pain.

On the product side it runs parallel to Products Ship 90% Complete, another discipline Franceschi articulates: over-invest in the small-cohort kernel before broad rollout. The existing-customer cohort is where that kernel gets battle-tested. It also complements Build the Zero-to-One Muscle Early, the rule Eric Glyman applies at Ramp to build a second product early and keep the creation muscle alive; where Ramp's rule is about when to build, Franceschi's is about where to sell. Together they describe the multi-product end state of a Compound Startup.

The tensions Franceschi leaves open

Two limits are acknowledged in the same account. The first is the problem of selling automation to the people it automates: if the existing customer is the finance person whose job the product eliminates, a company may have to route around that person to their manager. Franceschi's answer is co-design and a focus on outcomes, framing a team that went from thirty people to two as a win worth paying more for, but the internal-politics friction is described as real and unresolved.¹ The second is an adjacency limit: the rule only holds when the new product is close enough that the existing base already has the pain. Genuinely orthogonal bets still face the new-product-to-new-customer wall the rule was built to avoid.