Chapter 1 / 9
What is Amplitude, and which question does it answer?
Amplitude is a product analytics tool: it measures what users do inside an app, where they get stuck and why they come back. Finding the source is not its job; slicing behaviour once the source is known is.
- You can explain the split between an MMP and product analytics in one sentence.
- You can tell which tool should answer a given client question.
- You can explain what Amplitude does not do, and why it does not replace Adjust.
Two different questions, two different data shapes
In Adjust, the central object is the install record. Each row assigns a device to a source, and the hard part is getting that match right. In Amplitude, the central object is the user. Each user has a time-ordered event stream, and the hard part is comparing those streams with the right questions.
Side by side
| Adjust (MMP) | Amplitude (product analytics) | |
|---|---|---|
| Main question | Which channel, campaign and creative does this install or event belong to? | Where do users get stuck, who comes back, which behaviour drives what? |
| Central object | Install and attribution record (device-based) | User and their event stream (person-based) |
| External data | Click/impression data from ad networks, SAN integrations, SKAN postbacks | None. Only what you or your integrations send |
| Event definition | Defined in the dashboard first, sent with a token | Created the first time a name arrives |
| Typical output | Campaign reports, ROAS, cohorts, postbacks, raw data | Funnels, retention, path analysis, segments, cohorts |
| Relationship with networks | Sends postbacks to networks, feeds their optimisation | No direct attribution relationship with networks |
| Typical user | UA and performance marketing teams | Product, growth and data teams |
Adjust also measures post-install events and offers cohort reports. The difference is why it stores those events: Adjust keeps them to report by source and to send back to networks. Amplitude keeps them as each user's behavioural sequence, which is why you can ask ordered questions such as "people who did X and then Y".
What Amplitude does not do
Amplitude does not perform attribution matching with ad networks. It cannot take click data from Meta, TikTok or Google and assign an install to them, it does not manage SKAN conversion values and it does not send install postbacks to networks. That is why it does not replace an MMP in mobile marketing; it consumes the MMP's result.
On the web, Amplitude can capture UTM parameters and referrer information and analyse campaigns. That is not the same as deterministic network matching on mobile.
The scope and names of Amplitude's web attribution and marketing analytics features have changed over time. Check the current product page before telling a client "Amplitude does this".
Which tool does this question go to?
Classifying client questions is the fastest way to learn both tools. A question about source goes to Adjust, a question about behaviour goes to Amplitude, and a question that mixes both goes to the integration.
| Client question | Tool | Why |
|---|---|---|
| "How many installs came from TikTok last week?" | Adjust | Assigning installs to sources |
| "At which onboarding step do users drop off?" | Amplitude | Ordered step analysis (funnel) |
| "What is the ROAS of the Meta campaign?" | Adjust | Cost and revenue joined with source |
| "Do users of the search feature come back more?" | Amplitude | Behavioural segment and retention |
| "Is D7 retention better for TikTok users than for Meta users?" | Both together | Source from Adjust, retention computed in Amplitude |
| "Are these installs fraudulent?" | Adjust | Fraud prevention is an MMP job |
| "Did the new checkout screen improve conversion?" | Amplitude | Before/after comparison or A/B test |
| "What do users do before they purchase?" | Amplitude | Path / journey analysis |
The Amplitude product family
Amplitude is not a single screen but several products built on the same event data. Start with analytics; the others use the same data model.
| Product | What it does | When to learn it |
|---|---|---|
| Analytics | Charts, dashboards, cohorts | First |
| Data / governance tools | Tracking plan, event and property management, validation | Together with Analytics |
| Experiment | A/B tests and feature flags | When needed |
| Audiences / CDP features | Syncing cohorts to other tools | When needed |
| Session Replay | Watching user sessions like a video | When needed |
Product names, packaging and which feature sits in which plan change often. Check the current pricing page for the free plan's event limit and included features.
A frame worth remembering
In Adjust, the hard part is solved while data is being collected: getting the match right. In Amplitude, the hard part is solved before data is collected: getting the event design right. With badly designed events, even the best chart gives the wrong answer. That is why the next chapter starts with the data model.
Write down five questions a client you manage asked in the last month. Classify each as Adjust, Amplitude or both, and give the reason in one sentence. If you can't classify a question, there is usually a definition gap or a missing event.