Amplitude: Analytics for Understanding User Behavior

The mental model

Think of Amplitude as a DVR for user actions, not just a traffic counter for pages. While traditional web analytics focuses on page views and sessions, Amplitude focuses on what users do inside your product. The fundamental unit is an 'event'—a specific action like 'Song Played' or 'Item Added to Cart'—which provides a much richer view of user behavior.

How it works

Engineers instrument an application using an Amplitude SDK. They call a tracking function to send named events, often with properties that add context (e.g., event 'Song Played' with property 'genre': 'Pop'). This event data streams into Amplitude. From there, product managers, analysts, and marketers use a web-based interface to analyze the data. They can build charts, funnels, and dashboards to answer questions without writing SQL. A core part of a good setup is a 'tracking plan', a central schema that defines all valid events and properties to prevent messy, inconsistent data.

When to use it

Use Amplitude when you need to understand user journeys and product engagement deeply. Three key use cases: first, building funnels to see conversion rates through a multi-step process, like onboarding or checkout, and identifying drop-off points. Second, creating behavioral cohorts to group users by their actions (e.g., 'power users' who used a feature 10+ times) for retention analysis or targeted messaging. Third, running A/B tests and managing feature flags to measure the causal impact of changes on user behavior.

When not to use it

Amplitude is overkill for simple content websites where basic traffic metrics like page views and bounce rate are sufficient. It is not a replacement for a data warehouse; its strength is interactive analysis, not complex joins across disparate data sources. Because it relies on a well-maintained tracking plan, it's not a 'set and forget' tool and requires ongoing governance, making it a poor fit for teams unwilling to invest in data hygiene.

One canonical example

A product manager for a music streaming app wants to improve user activation. They create a funnel in Amplitude to track the user journey from signup to first song play. The steps are: 1. Event: 'User Signed Up', 2. Event: 'Viewed Onboarding', 3. Event: 'Searched for Song', 4. Event: 'Song Played'. The funnel shows a 40% drop-off between 'Searched for Song' and 'Song Played', indicating a potential issue with the search results or playback experience that the team can now investigate.