User stitching
User stitching combines data from multiple sources or sessions into a unified view of a user. You need it to combine website and mobile app data, to unify customer interaction data across platforms, or to follow a user's journey through various touchpoints. It is especially useful when a user interacts with a system on multiple devices or platforms and these interactions should end up in a single user profile.
User behavior data is often incomplete, especially on the web, where users can easily delete cookies, change devices or use a different browser. Stitching correctly starts with understanding user identification.
Working with persistent identifiers requires the user's consent. Consent is always bound to specific purposes, which have to be documented and communicated properly.
Learn more about consent management with walkerOS.
Relying on third-party tools and cookies might have been common practice, with benefits, in the past. Data that isn't owned and collected first-party always depends on others.
First-party data collection is the basis for attribution modeling, a complete view of your users and data ownership.
User identification levels
User identification means recognizing a user, typically by associating them with a unique identifier. walkerOS has three levels of user identification:
- id: a unique identifier for a user, typically from internal systems like a CRM.
- device: a device-specific identifier, saved in the storage.
- session: a session-specific identifier for user activities within a single session.
A single user can have different devices with multiple sessions. The id
identifies an individual user across sessions and devices. The device
recognizes a user across multiple sessions on the same device, while session
covers user behavior and interactions within one limited time frame.
The diagram represents one user with two devices across three sessions.
These techniques may require a user's consent and documentation to respect privacy and stay compliant. Ask your data protection officer for legal advice.
Example
Events are individually measured user interactions. Stitching them together creates user journeys, and all event-related data helps to identify a user.
The following table is a minified event log over several days. It shows user activities over time and across sessions, with varying levels of user identification and consent categories.
All events of a run, representing a page load, share one source.trace id. A
hash is a temporary fingerprint. session, device, and user are all simplified
unique identifiers.
| Step | Day | Event | Consent | Trace | Hash | Session | Device | User |
|---|---|---|---|---|---|---|---|---|
| 1 | Monday, 1st | session start | analytics | a | foo | |||
| 2 | Monday, 1st | page view | analytics | a | foo | |||
| 3 | Monday, 1st | consent deny | analytics | a | foo | |||
| 4 | Monday, 1st | page view | analytics | b | foo | |||
| 5 | Tuesday, 2nd | session start | marketing | c | bar | 1 | phone | |
| 6 | Tuesday, 2nd | page view | marketing | c | bar | 1 | phone | |
| 7 | Tuesday, 2nd | consent accept | marketing | c | bar | 1 | phone | |
| 8 | Tuesday, 2nd | session start | marketing | d | bar | 2 | phone | |
| 9 | Tuesday, 2nd | page view | marketing | d | bar | 2 | phone | |
| 10 | Wednesday, 3rd | session start | marketing | e | baz | 3 | phone | olli |
| 11 | Wednesday, 3rd | page view | marketing | e | baz | 3 | phone | olli |
| 12 | Wednesday, 3rd | user login | marketing | e | baz | 3 | phone | olli |
| 13 | Friday, 5th | session start | analytics | f | qux | |||
| 14 | Friday, 5th | page view | analytics | f | qux | |||
| 15 | Friday, 5th | page view | analytics | g | qux | olli | ||
| 16 | Friday, 5th | consent deny | analytics | g | qux | olli | ||
| 17 | Friday, 5th | user login | analytics | g | qux | olli | ||
| 18 | Thursday, 11th | session start | marketing | h | lol | 4 | ph0n3 |
Based on the data, we can at least summarize the following:
- 18 tracked
eventsin total - 8 different
traceids ('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h') - 6 individual
session startevents - 5 temporary
hashes('foo', 'bar', 'baz', 'qux', 'lol') - 4 reliable
sessions(1, 2, 3, 4) - 2 known
devices('phone' and 'ph0n3') - 1 unique
user('olli')