Dotdigital Certification cheat sheet

Program builder logic is built from three node families: triggers (entry points), rules/conditions (branching), and actions (send, delay, update contact, webhook); a program runs per contact, not per batch.

A trigger evaluates on the event that fires it; a condition node re-evaluates at the moment the contact reaches it, so a contact can take a different branch on a later pass through the same program.

Delay nodes pause on a duration or until a fixed calendar time; contacts waiting in a delay do not consume send capacity until they exit the node.

Re-entry rules determine whether a contact can start a program again: allow re-entry, block re-entry while active, or block permanently; default behaviour blocks duplicate active enrollments.

Goal/exit criteria remove a contact from the program as soon as the goal is met, preventing later nodes from sending a redundant message.

Send limits and quiet hours are enforced at the action node; a blocked send is queued or dropped depending on the node setting, not retried automatically.

Test mode sends to nominated seed addresses and skips contact-record writes; live sends write to the contact's activity history and drive reporting.

Automation reporting counts entries, completions, goal conversions, and drop-offs per node; the drop-off figure is the difference between contacts entering and leaving a node.

Dynamic content blocks render at send time from contact data, so a value changed after the send is queued will not appear in that send.

Fallback (default) values must be set for every merge field; a missing value renders the fallback or blank, never the field token.

Personalization rules can key on data fields, custom objects, or behavioural attributes; rule order matters because the first matching rule wins.

Recommendation blocks typically use viewed, purchased, or browsed category signals; a cold-start contact with no history needs a fallback product set.

Personalization is evaluated per recipient, so a single campaign can produce many rendered variants; preview renders use a nominated contact.

Conditional content in email uses the platform's conditional syntax, not raw scripting; unsupported logic fails at render and can blank the block.

Zero-party data (preferences declared by the contact) is more reliable for personalization than inferred data and should be captured with an explicit consent record.

Real-time web personalization targets known contacts via a recognized cookie or identity link and anonymous visitors via segment or behaviour rules.

Loyalty tiers are usually defined by a rolling 12-month points or spend threshold; tier recalculation should be scheduled, not event-driven, to avoid oscillation.

Points liability = unredeemed points balance x redemption value; it is a financial exposure and should be reported alongside engagement metrics.

Points expiry rules must be stated in the terms; a common pattern is expiry after N months of account inactivity rather than a fixed calendar date.

Win-back segmentation uses recency bands (for example 90/180/365 days since last order or last click) and should exclude contacts already in an active win-back program.

Churn risk scoring combines recency, frequency, monetary value, and engagement decay; the score is only useful if a defined action fires at each threshold.

Referral programs need unique codes per advocate and fraud checks on the same device, address, or payment instrument.

Retention reporting should compare cohorts by acquisition month; a flat overall churn rate can hide a worsening recent cohort.

Suppression of recent purchasers from acquisition messaging prevents discount cannibalization and improves margin per send.

Email sends require a verified sending domain with SPF, DKIM, and DMARC aligned; authentication failures lower deliverability regardless of content.

SMS/MMS requires an approved sender ID or short/long code per country; message length is segmented (for example 160 GSM-7 characters per segment, fewer with Unicode).

WhatsApp business messaging uses pre-approved template messages for business-initiated conversations; free-form replies are allowed only inside an open customer service window.

Mobile app push requires a device token per install; tokens expire or rotate, so inactive tokens must be pruned to keep delivery rates accurate.

Web personalization and on-site pop-ups depend on a tracking script; consent state gates whether the script may identify a visitor.

Ads and retargeting audiences are synced from segments to the ad platform; match rates depend on hashed identifiers and consent for marketing use.

Channel choice follows urgency and consent: transactional email or push for low urgency, SMS/WhatsApp for time-critical, and app push only for installed, opted-in users.

Every channel needs its own opt-out mechanism; an email unsubscribe does not automatically suppress SMS or push, and consent must be recorded per channel.

The contact record is the primary entity; related data (orders, custom objects) joins on a stable identifier such as email, customer ID, or a hashed key.

Data fields have types (string, number, date, boolean); a type mismatch on import causes the value to be rejected or truncated, not coerced silently.

Insight/data-collection scripts capture behavioural events (page view, product view, cart) tied to an identifier; identity resolution links anonymous and known profiles.

Segment builder combines AND/OR rule groups over fields and behaviour; segment membership is recalculated on a schedule or on data change, so counts are not real-time.

Engagement metrics: open rate = unique opens / delivered; click-through rate = unique clicks / delivered; click-to-open rate = unique clicks / unique opens.

Delivery metrics: bounce rate = bounces / sent; complaint rate is measured against delivered mail and is the metric mailbox providers weigh most heavily.

Attribution for revenue reporting needs a stated model (last click, first click, or custom); comparing revenue across models without stating the model is misleading.

Consent and preference data belong on the contact record with source, timestamp, and channel scope; reporting on consent coverage depends on those fields being populated.

AI features in a marketing platform operate on the customer data already stored; data quality and consent scope therefore bound what any AI output can safely use.

Generative copy assistants produce drafts; brand, legal, and factual accuracy remain the sender's responsibility, and outputs should be reviewed before send.

Predictive scoring (for example predicted next order date or churn probability) is a model output with a confidence band; thresholds should be tuned against a holdout group.

Send-time optimization picks a per-contact delivery time from historical engagement; contacts with little history fall back to a segment or account default.

AI segment or audience suggestions should be validated against a control segment to prove incremental performance before scaling spend.

Conversational or agentic flows must respect the same consent and suppression rules as manual sends; an AI-triggered message is still a marketing message.

Model outputs drift as behaviour changes; predictions should be re-scored on a schedule and monitored for accuracy decay.

Human-in-the-loop review is required wherever an AI action is irreversible: sending, discounting, or writing to the customer record.