Data Retention

mParticle provides data retention policies for the data you store. Two main data classes provide information about your users and the events they trigger: profile data and event data.

Profile data

Data about users are stored as attributes of individual profiles. These attributes include identities, device types and IDs, and several custom attributes such as membership status and demographic information. An attribute value may not be current, depending on how often it is updated.

For more information about how profile data is associated with users, see Store and Organize User Data.

Event data

Events describe actions that your users have taken. Event data is stored as attributes of event types. The value of an event data attribute is valid at the moment the event was triggered. For example, the event “Sign up” could have an event attribute of “membership tier,” the membership status at the time of signing up.

A complete description of the mParticle event schema is in JSON Schema.

Data retention limits

The maximum period that mParticle stores profile and event data is governed by your long-term data retention policy, which is defined in your contract. Usually, the long-term data retention policy is the same for event and profile data. However, you can have different ranges for added control and flexibility for events and profile data. For example, during org and account setup, you can set long-term data retention for events at two years and reduce the time profile data is available to 12 months.

Another factor may affect the data that is available for audiences: the Real-Time Audience Storage Lookback specified in your contract. Typically it is set to 30, 60, or 90 days, but can be changed. It is also overridden by Unlimited Lookback. See Data retention and Unlimited Lookback for details.

Data retention and Unlimited Lookback

Unlimited Lookback is a premium feature that extends your audience and calculated attribute lookback to your long-term retention for events. Without this feature, audience lookback is limited to the Real-Time Audience Storage Lookback specified in your contract.

Data retention for personalization features (audiences, journeys, and calculated attributes) behaves differently depending on whether or not your account has the Unlimited Lookback feature enabled:

• Unlimited Lookback:

  • Uses long-term data retention for events.
  • All events in the date range for your long-term data retention for events are available for data evaluations.
  • In the criteria builder, the number of days you specify for recency can go back to the long-term data retention for events.
  • When a personalization feature (journey, audience, or calculated attribute) is activated, it is initialized using data available for the entire range specified by your long-term data retention for events. For the initialization of personalization items, a longer lookback usually means larger data volume, which may incur an additional expense.
  • To help with cost estimation, you can set an estimated lookback value. This informational value alerts anyone creating a calculated attribute when a specified date range exceeds that organization’s expected date range.

• Without Unlimited Lookback:

  • mParticle uses the Real-time Audience Storage Lookback for real-time evaluation, which is specified in your contract and is often set to 30, 60, or 90 days.
  • The behavior described for Unlimited Lookback features is not available.

Examples with no Unlimited Lookback

• If your long-term retention for events is set to two years, you can view events that occurred up to two years ago.
• If your long-term retention for profiles is set to seven days, if a user has no activity for more than seven days, then the profile data for that user expires and isn’t available for calculated attributes, journeys, or real-time profile enrichment.

Examples with Unlimited Lookback

• If your long-term retention for events is set to two years, audiences, journeys, and calculated attributes are initialized using data up to the long-term retention for events and long-term retention for profiles.
• When choosing a recency or frequency value in the criteria builder for audiences or journeys, you can specify a value up to the long-term retention for events.

Date range measurement

mParticle determines the beginning of a date range for retention purposes differently for event and profile data:

• Event data: The event batch timestamp (timestamp_unixtime_ms) added to the top level of every batch, representing the time the batch was received by mParticle. Note this is different from the timestamp associated with an individual event.
• Profile data: The last time a profile was updated by an inbound data stream. For example, a profile may be created, attributes may be added, updated, or deleted, or other profile information may change. Each change triggers a timestamp change.

Events and profiles need different ways of calculating age because an event isn’t usually updated. Since several processes may modify a profile’s timestamp, the date is measured differently than event data.

Date range example

Assume the following facts:

• A screen view event (screen_view) occurred and is time-stamped 1657934165001 (6 June 2023, at 21:29:24).
• This event was received by mParticle (ingested) in a batch time-stamped 1657934165102 (6 June 2023, at 21:30:55).
• Your long-term data retention for events is two years.

In this example, mParticle keeps the event available until 1749270655 (6 June 2025, at 21:30:55).