Amplitude

Amplitude models product usage as immutable events attached to a resolved user identity, with identity stitching across devices and behavioral cohorts as first-class objects you can chart, compare, and sync to downstream tools. Identity resolution runs on three signals: a device ID the SDK generates automatically, the user ID your product assigns at signup or login, and an internal Amplitude ID minted once the platform can conclusively tie them together. Merging happens on the backend, so anonymous events a user generated before logging in get connected to their profile retroactively — you do not have to alias identities by hand the way some competitors require.

That identity model has sharp edges worth knowing before you instrument. Once set, a user ID cannot be changed, so pick a stable internal identifier rather than an email address. User IDs are case-sensitive — “User123” and “user123” become two different people. And never write placeholder strings like “None” as a user ID, because Amplitude will happily merge every user carrying that value into one profile.

Governance lives in Amplitude Data: a tracking plan where you document event names, properties, and ownership, monitor event volume and property types for unexpected changes, and rename, redact, or reshape data after ingestion without redeploying code. Autocapture can collect website interactions before you manually instrument events, which is useful for getting started but no substitute for a deliberate taxonomy. Warehouse connectivity is genuinely two-way: the Snowflake source supports full, timestamp-based, append-only (CDC), and mirror sync modes — though mirror sync disables Amplitude’s enrichment (ID resolution, property syncing, taxonomy validation) to keep the warehouse as source of truth, a trade-off to understand before choosing it.

Experiment (feature flags plus A/B testing) and Session Replay hang off the same event taxonomy, so an experiment’s exposure events and a replay’s timeline share definitions with your funnels. That coherence is the pitch — one identity graph and metric layer across analysis, testing, and replay. Session replay volume is capped by plan tier, so heavy replay use is effectively an enterprise conversation.

Pricing is event-based across the board: a free plan with 2M events per month and no time limit, self-serve Plus that scales to 70M events, and sales-quoted Growth and Enterprise above that, with add-on packages priced as a percentage of your platform plan. All plans include unlimited seats.

Use it when a product org needs deep retention and cohort analysis with mature governance, SSO, and enterprise controls. Look past it when event-based pricing gets punishing at your volume, or when you would rather keep raw events in your own warehouse and model them there — mirror sync narrows that gap but costs you Amplitude’s enrichment layer.