# How Amplitude ships product improvements in 1–2 days with OpenAI

Source: [OpenAI](https://openai.com/business/customer-stories/amplitude)  
Feed7 permalink: https://feed7.dev/p/amplitude-1jwi9ut  
Published: 2026-11-12T20:00:00.000Z  
Trust: Official Source (official_source)

## Why Included

Amplitude’s Wave turns product analytics and feedback into plans for coding agents, then measures the change. Its homepage experiment reported 160% more messages sent through suggestion cards after one day.

## Source Summary

Wave combines product analytics, session replays, customer feedback, and code to identify opportunities, prepare implementation plans, and track outcomes. On Amplitude’s homepage, personalized suggestion cards increased card-driven sends by **160% after one day**. Amplitude estimates improvements can ship in **1–2 days**, down from about a month.
For a solo builder, the useful pattern is to connect observed user behavior to a small proposed change, explicit acceptance criteria, and an agent implementation, then measure the result. Amplitude assigns different models to deep investigations, frequent checks, and mockups based on quality, latency, and cost.
These are Amplitude-reported results, including a one-day experiment on its own homepage. The material does not establish that the same workflow or lift will transfer to other products.

## Practical Implication

No separate implication recorded.

## Agent-Ready Context

Wave combines product analytics, session replays, customer feedback, and code to identify opportunities, prepare implementation plans, and track outcomes. On Amplitude’s homepage, personalized suggestion cards increased card-driven sends by **160% after one day**. Amplitude estimates improvements can ship in **1–2 days**, down from about a month.
For a solo builder, the useful pattern is to connect observed user behavior to a small proposed change, explicit acceptance criteria, and an agent implementation, then measure the result. Amplitude assigns different models to deep investigations, frequent checks, and mockups based on quality, latency, and cost.
These are Amplitude-reported results, including a one-day experiment on its own homepage. The material does not establish that the same workflow or lift will transfer to other products.

## Connected Context

Feed7 judgment across 1089 accumulated Signals:

This Signal connects analytics, replays, feedback, and code in a measured product-improvement loop, adding a customer-facing outcome to candidates focused on agent traces and coding workflows. Its reported 160% lift is a one-day result on Amplitude’s homepage, so it does not establish that the method or effect transfers elsewhere.

- [Improving Agents is a Data Mining Problem — Vivek Trivedy, LangChain](https://feed7.dev/p/improving-agents-is-a-data-mining-problem-vivek-trivedy-langchain-1cgibr8) — Both use observed workflow data to guide agent improvement; Amplitude links product behavior to proposed changes and outcomes, while the talk uses production traces to find and replay agent failures.

## Context Map

- Layer: agent
- Domains: coding, data
- Topics: coding-agents, agent-evals, harness-engineering

## Uncertainty

- None recorded.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.
