Should I install the Amplitude SDK?

“Do I need to install the Amplitude SDK or can I just pipe in events from our warehouse?”

I’ve heard this question on so many customer calls. You may have wondered this too and it warrants a longer answer. Amplitude supports warehouse import, the HTTP API, batch uploads, streaming, CDP integrations, and SDKs for web, mobile, and server. That range of options makes it seem like every ingestion path is equal and interchangeable, like different roads to the same destination.

They're not. Each ingestion option unlocks a different slice of the platform. Skipping the SDK means locking yourself into certain capabilities and out of others. Many teams make that call, then discover what's missing a year later when a product manager asks a question the data isn’t able to answer. If the SDK had been installed at the outset, a quick change via remote config could have solved the problem.

SDK and warehouse import aren't rivals

It’s worth noting that many customers collect events with Snowplow or an in-house pipeline. Those events get sent to their warehouse before anything reaches Amplitude. The assumption is that they can just export that data from the warehouse to Amplitude to get all the benefits of our platform.

While there is genuinely valuable data that can be imported into Amplitude from a warehouse, the Amplitude SDK captures product behavior at a resolution no warehouse table holds, like rage clicks, DOM snapshots, and exposure events. You can layer on the SDK for Session Replay, autocapture, and exposure tracking if you’re already using Snowplow or your own pipeline for event collection.

What each ingestion path gets you

When reading this table, think about the different roles in your organization and what they need. Marketers can’t learn which campaigns are working without acquisition attribution data. Growth teams without accurate exposure tracking will get false signals from their experiment results. Product teams have way more empathy for the customer when they can inspect dropoff rates in a conversion funnel and see the user frustration in a session.

Better data, better agents

An AI agent can only apply logic to data that exists. Point it at transaction records alone, and it can tell you that order volume is dropping. Point it at interaction-level behavior data and replays, and it will pinpoint the exact screen where people gave up. Feeding our product agent, Amplitude Wave, as much contextual data as you can collect will generate better opportunities and more impactful product improvement suggestions.

When warehouse import is the right call

To be clear, the Amplitude SDK isn’t the right answer for every team. Orders, payments, subscriptions, refunds, and entitlement changes are usually reconciled in the warehouse and aren't subject to client-side loss. Teams that only need that level of information can import that data from their warehouse or export Amplitude data to their warehouse to do joins.

In certain regulated industries having a second collection layer is often not possible because your warehouse carries governance the SDK doesn't inherit: canonical schemas, lineage, access control, correction workflows. If your team has already fought to make one CDP or event bus the single approved path, reopening that discussion might not be worth it.

So, should you install the Amplitude SDK?

In most cases, yes.

Autocapture, Session Replay, in-product experiments, pre-login activity, and the behavioral depth that makes AI analytics useful all require the SDK. Skipping it means missing out on those capabilities.

If you want to install the SDK, our documentation covers setup for each platform.