Acoustic Campaign Certification cheat sheet

Behavioral scoring aggregates explicit actions (opens, clicks, page views, form submits) into a numeric engagement value; scores decay over time so recency outweighs volume.

A behavior event record minimally contains contact ID, event type, timestamp, channel, and campaign or asset ID; missing any of these breaks attribution.

Recency-frequency-monetary (RFM) framing: recency = days since last interaction, frequency = event count in window, monetary = value attributed to conversions.

Engagement tiers are typically defined by threshold bands (e.g., high/mid/low) applied to a rolling window; tier boundaries must be fixed in advance to keep segments comparable.

Behavioral triggers fire on a state change (first open, Nth click, cart abandonment) rather than on a schedule; scheduled sends ignore real-time intent.

Deduplication rule: one contact should generate at most one instance of a given trigger per suppression window to prevent send storms.

Cross-channel behavior must be normalized to a common contact key; unmatched anonymous events stay unlinked until an identity merge occurs.

Behavioral data quality checks: null event type rate, timestamp skew, duplicate event IDs, and orphan contact IDs should each be monitored.

A profile is a contact record plus linked attributes: identity fields, consent state, preferences, channel addresses, and computed attributes.

Explicit attributes come from declared data (forms, preference centers); implicit attributes are inferred from behavior and must be labeled as inferred.

Consent is stored per channel and per purpose; a profile can be opted in for email and opted out for SMS simultaneously.

Identity resolution merges records on deterministic keys (verified email, hashed phone) before probabilistic matching; merges are reversible via an audit trail.

Profile completeness is measured as the share of non-null fields among required attributes for a given program, not across all possible fields.

Preference center choices override default program settings; suppression lists override both and are checked at send time.

Profile updates should be event-sourced so the current state can be reconstructed and disputed changes audited.

Golden record rule: when duplicate profiles conflict, the most recently verified value wins for identity fields; behavioral history is unioned, not overwritten.

Intent signals are observed actions indicating near-term interest: repeat product page views, pricing page visits, content downloads, and search queries.

Signal strength = frequency × recency weight × action weight; assign higher action weights to high-cost behaviors (demo request) than low-cost ones (single open).

Intent scores require a decay function; a common form is exponential decay where weight halves every N days.

Threshold rule: only contacts crossing a defined intent score promote into a high-intent segment; below-threshold contacts stay in nurture.

Signals must be time-bounded; an event older than the lookback window contributes zero to the current score.

Negative signals (unsubscribe, spam complaint, long inactivity) reduce or zero out intent and can suppress outreach regardless of positive score.

Anonymous intent (unmatched web sessions) can be aggregated at account or domain level but must not be attributed to a named contact until identity resolution.

Signal-to-action mapping should be documented: which score band triggers which orchestration path, so scoring changes are testable.

A segment is a saved, dynamic query over profile and behavioral attributes; membership recomputes as underlying data changes.

Static (snapshot) segments freeze membership at build time and are appropriate for one-time sends or holdout groups.

Include and exclude rules combine with AND across rule groups and OR within a group; precedence must be explicit to avoid ambiguous logic.

Segment size sanity check: a segment that grows or shrinks more than an expected percentage day over day usually signals a broken rule or data feed.

Suppression segments (unsubscribes, global opt-outs, competitors) must be applied at send time, not baked into targeting logic.

Holdout segments (typically 5-10 percent) are excluded from treatment to measure incremental lift against a comparable control.

Overlap analysis between segments prevents duplicate messaging when a contact qualifies for multiple active campaigns.

Segmentation on inferred attributes carries higher error risk; validate inferred-field segments against a sample before scaling sends.

An orchestration program is a directed flow of nodes: entry criteria, wait steps, decision splits, channel sends, and exit rules.

Entry criteria should be mutually exclusive across concurrent programs to prevent a contact entering two competing journeys.

Channel priority rule: choose the channel by contact consent, prior response, and urgency; fall back to the next consented channel on non-delivery.

Wait steps measure elapsed time from the prior node; decision splits re-evaluate attributes at the moment of evaluation, not at entry.

Frequency capping applies across all channels in the program; a contact who hits the cap exits or is deferred, not silently skipped.

Exit rules must include conversion, unsubscribe, and goal-completion conditions so contacts stop receiving steps after the desired action.

Cross-channel attribution requires a shared contact key and a defined attribution window; last-touch and multi-touch models yield different results.

Test design: hold a control group, change one variable per test, and predefine the primary metric and minimum sample before launch.