mParticle Practitioner flashcards
mParticle Practitioner · Build recall with a focused review session.
32 cards. Read the term and the definition here. The deck at the top of this page is optional.
- Hybrid architecture definition
- A setup where client-side SDK events and server-side API events feed the same mParticle workspace, sharing one identity model and one output pipeline.
- MPID
- The mParticle ID, a platform-generated canonical identifier assigned to each user profile and used to link identities and events across inputs.
- Client vs server latency
- SDK events batch and flush on size or time thresholds, while server Events API calls arrive immediately, so the two paths reach outputs at different times.
- Data plan enforcement
- Workspace settings that declare allowed event and attribute names and types, letting the platform block or flag undeclared data before it reaches outputs.
- Input vs output feed
- An input feed is a source of data such as an SDK or API; an output feed is a destination that receives processed events and audience updates.
- Kit integration
- A partner integration embedded in the client SDK that forwards events directly from the device rather than from the mParticle platform.
- Environment isolation
- Development and production workspaces use separate API keys and separate data stores, so test traffic never mixes with production user profiles.
- Profile merge condition
- Two profiles merge only when both carry a shared identity key such as email or customer ID; without overlap they remain separate MPIDs.
- Audience definition
- A named group of users whose membership is evaluated in real time against rules over events, attributes, and calculated values.
- Recency window rule
- A condition such as 'performed event in the last 30 days' that admits users on recent activity and drops them when the window expires.
- Calculated attribute
- An aggregate over a user's events, such as total purchases or last-seen date, usable as a criterion inside audience rules.
- Nested audience
- An audience whose rules reference membership in another audience, allowing composition of complex segments from simpler building blocks.
- AND vs OR logic
- AND between rules requires all conditions to hold and narrows membership; OR requires any condition and widens it, so grouping changes results.
- Audience membership event
- A platform-generated event written to the user profile when audience status changes, forwardable to outputs for downstream activation.
- Empty audience cause
- A rule referencing an attribute or event that is never populated will match no users, producing an audience with zero membership.
- Audience activation connection
- A configuration linking a specific audience to a destination account so membership changes are delivered to that partner for targeting.
- Predictive audience
- An audience whose membership comes from a trained model's probability score for each user, filtered by a configured threshold.
- Model target event
- The outcome event a prediction model learns to forecast; it must occur often enough in history for training to be meaningful.
- Feature leakage
- Including data recorded after the outcome event in model features, which inflates measured accuracy and breaks real-world performance.
- AUC interpretation
- A model quality metric where about 0.5 means random ranking and higher values mean better separation between likely and unlikely converters.
- Prediction threshold tradeoff
- A high score cutoff yields a small, precise audience; a low cutoff yields larger reach but includes many low-probability users.
- Score refresh behavior
- Predictive membership is recomputed as scores update, so users can enter or exit an audience without performing any new event.
- Prediction identity dependency
- Model scores rely on the identity graph; fragmented profiles with unmerged client and server events produce scores from incomplete history.
- Predictive audience reuse
- A predictive audience can serve as a rule criterion in other audiences and be connected to outputs like any rule-based audience.
- Identity priority order
- The ranked list of identity types the platform checks when resolving a user, with the highest-priority match anchoring the profile.
- Identify call
- A client or server call that adds or updates identity keys on the current user, helping attach activity to the correct profile.
- Login and logout
- Login switches the current user to a known identity, potentially merging an anonymous profile; logout returns the session to an anonymous state.
- Merge effect on audiences
- When two MPIDs merge, combined history is re-evaluated, so the user can immediately gain or lose audience membership.
- Cross-device identity key
- A durable first-party identifier such as email or customer ID that links a user across devices, unlike device-specific advertising IDs.
- Server event without identity
- A server-side event missing identity keys attaches to an anonymous profile, fragmenting history and weakening audience and prediction quality.
- Duplicate forwarding risk
- Enabling both a kit and a server-side integration for the same partner can deliver the same event twice, inflating counts downstream.
- Output event filter
- Per-destination rules selecting which events are forwarded, preventing schema breaks and unnecessary volume at the partner.