MVT Deterministic Treatment Assignment

Overview 

Deterministic Treatment Assignment (DTA) is an enhancement to Multi-Variate Testing (MVT) that replaces cookie-based assignment with a stable, externally managed ID. This ensures consistent treatment assignment across sessions, reduces biases associated with cookie-based tracking, and mitigates issues related to API integrations, browser limitations (such as third-party cookie deprecation), and session tracking inconsistencies.

Instead of using the mvtid stored in cookies, DTA systematically assigns MVT treatments based on a separate, stable ID. You can choose from:

  • External Session ID - Managed by the client, ensuring session continuity.

  • External User ID - Managed by the client, ensuring persistent cross-session identity.

  • User GUID (rrGUID) – Assigned by Algonomy to uniquely identify users.

  • Arbitrary Parameter – Custom ID or value that can be used in specific cases but is rarely implemented.

Users remain in the same treatment group as long as the chosen ID remains unchanged, eliminating the instability, test bias, and server-side integration issues of the legacy approach. While External Session ID is typically the best choice, the Customer Success (CS) Team should collaborate with Analytics to determine the most suitable seed type, as it also drives MVT test analysis.

Technical Implementation and Architecture

Treatment Assignment Algorithm

  1. ID Selection: The system uses the chosen stable ID as a deterministic seed for treatment assignment.

  2. Hashing and Partitioning: The selected ID undergoes a hashing function to distribute users into treatment groups.

  3. Treatment Allocation: The hashed value is mapped to a predefined MVT treatment bucket.

  4. Consistent Assignment: The treatment allocation remains stable as long as the ID remains unchanged, ensuring repeatability in user experiences.

Data Flow

  1. The system receives a request containing one of the predefined stable IDs.

  2. The ID is passed through a hashing algorithm to generate a deterministic numeric value.

  3. The numeric value is used to index into the treatment lookup table, ensuring consistent assignment.

  4. The system responds with the assigned treatment, which is included in API responses or page-level personalization.

Enabling Deterministic Treatment Assignment (DTA)

DTA can be enabled through collaboration with the Customer Success (CS) Team. This process involves:

  1. Defining the preferred ID type.

  2. Configuring API endpoints to process treatment assignment deterministically.

  3. Ensuring that all running MVTs are completed or stopped before making changes.

  4. Enabling the MVT deterministic user hashing site configuration.

    Graphical user interface, text, application Description automatically generated

  5. Setting the MVT seed type site configuration to select the stable ID:

    1. External User ID

    2. External Session ID

    3. RR User GUID

Note: The MVT seed type setting is used for treatment assignment and is also referenced in Portal Reports. If MVT deterministic user hashing is not enabled, the seed type setting is ignored for assignment but remains relevant for reporting.

Best Practices

Choosing the Right ID

The CS Team should work with Analytics to determine which type of ID is most stable for the site:

  • Are user IDs consistent?

  • Are session IDs stable?

  • Which type of ID is best suited for the MVTs being conducted?

While External Session ID is typically the best choice, consider the following:

  • External User ID is often not populated for a large portion of traffic on many client sites. However, if the client enforces login or consistently provides an External User ID, it may be preferred over External Session ID.

  • User GUID is cookie-based, leading to similar issues as the legacy solution. User GUID should only be used as a fallback if the client cannot systematically and accurately track most of their traffic.

If the client uses a server-side integration (which applies to all Discover and Find clients) and plans to use User GUID, ensure that full cookie pass-through is enabled. This means:

  • The Omnichannel Personalization–managed cookies (rcs cookies) from the customer’s browser are passed through from the client to our server-side API.

  • Our system updates the cookie contents and passes them back to the client.

  • The client correctly updates the cookie in the customer’s browser.

Additional Considerations

The chosen seed type will determine how MVT reporting groups visit data. A visit is defined as a sequence of page views that share the same ID within a session, without long gaps. Gaps exceeding a predefined threshold reset the visit counter.

  • If External Session ID is used for treatment assignment, visit data will be grouped accordingly.

  • If External User ID is chosen, cross-session aggregation is possible.

  • If rrGUID is selected, it enables long-term user tracking across devices.

Comparison: DTA vs. Traditional Assignment

Feature DTA Traditional Assignment
Assignment Mechanism Stable ID-based hashing Cookie-based MVT ID (uses a random-number generator)
Persistence Across Sessions Yes (if user ID-based) No (persists only via cookies)
Bias Risk Low Moderate
Browser Privacy Compliance High (no cookies) Moderate (depends on cookies)

Conclusion

Deterministic Treatment Assignment enhances MVT by providing a stable, scalable, and privacy-compliant alternative to cookie-based tracking. By leveraging stable IDs and a robust hashing mechanism, DTA ensures repeatability in treatment assignment while reducing test biases and enhancing data integrity for personalization strategies.