Concept

Zero-Support-Tickets

conceptmetricsproduct-managementcustomer-successstripe

Zero-Support-Tickets

Zero-Support-Tickets is Jeff Weinstein‘s north-star product metric for Stripe Atlas: the percentage of customers who complete the entire flow — from first page load, through the wait on the government and IRS, plus a two-week buffer — without filing a single support ticket. It is a metric designed by working backwards from “why would someone recommend this to a friend?”, on the premise that a customer who needed help is unlikely to become a promoter.

How it is constructed

Three design choices distinguish it from an ordinary support metric:

  • It is binary per customer, not an average. Each customer is a yes (zero tickets) or a no (any tickets). Weinstein rejects the average explicitly: dropping the per-customer average from 0.3 to 0.2 tickets does not necessarily make anyone more likely to recommend Atlas, whereas crossing to zero might.
  • It measures the customer’s outcome, not the support queue. The number improves only by removing the reasons to contact support — redesigning every point where customers got stuck (a stuck DocuSign, a missing co-founder address, opaque 83(b) instructions) and baking the fix into the product — not by optimising or deflecting tickets.
  • It is a metric you would be proud to leak. Weinstein’s test: if you accidentally showed customers your dashboard, they would be ecstatic to learn this is what you were measuring all along. (Stripe routinely screenshots and publishes Atlas metrics.)

The result

Atlas started at 15% of founders completing with zero tickets and reached 85% over roughly 18 months. Plotted on the same timeline, market share traced the same shape. The metric also reorganised the team: engineers were each handed a ticket reason, told to reply to the support email to learn the need, and asked to spec and build the fix — turning engineers into PMs and producing 83(b) automation, an in-house signing service, and “turning everything into a click.”

Guarding against perverse incentives

A single optimised metric invites gaming — Lenny’s counter-example is an Airbnb team that cut “contact support” by making support harder to reach. Weinstein’s guard: keep the one overarching metric, but drive it through specific tactical KRs owned by engineers (e.g. getting risk-review decisions to ~100% within an hour; reducing overturned rejections), so the team chooses which tactics it will use rather than letting a perverse one “casually exist.” Zero-Support-Tickets pairs naturally with his users-having-a-bad-day metric (emit a log line whenever a user hits a problem, then stack the reasons) as a more granular companion.

Why it matters

The metric operationalises Compounding: a customer who finishes without friction recommends Atlas to a friend, and word-of-mouth is the go-to-market (you cannot easily target someone “about to start a company” with ads). It is a concrete answer to the common objection that craft and customer experience have no direct line to growth — here the experience metric is the growth metric, and the curves match.