Rules
Part of Notes on email segmentation from the field
How to improve email segmentation by rewriting vague labels
How to improve email segmentation by turning labels like high intent into rules with a recorded event, a lookback window, exclusions and an expiry.
What to take away
- A label that records no event, window or expiry decides nothing, so rewrite it as a rule.
- Let eligibility exclusions override every relevance rule.
- Pair a plain-language definition with an executable one, and version both.
- Test boundaries, nulls and shared addresses before the rule goes live.
- Give every rule an expiry and an owner.
Find the labels that decide nothing
Labels such as high intent or engaged sound precise but often record no event, no window and no end date. List the labels your team uses, and ask what a recipient must have done for each one to apply. Where nobody can say, the label is the problem.
How to improve email segmentation begins with one faulty or unproven decision, not another attribute. Preserve the current rule, audience snapshot, source times, campaign exposure, and baseline before changing anything. Remove a segment if it changes no meaningful decision.
Rewrite one label as a rule
Before: high intent means anyone labeled engaged, with no recorded event or expiry. After: category interest means the subscriber has current marketing permission, is absent from the suppression register, and has a recorded click on a link tagged to that category during a documented lookback window.
Record the event source and window, and let eligibility exclusions override the relevance rule. Write the expected action beside the rule, using the campaign planning checklist as the prompt for purpose, owner, and guardrails.
Lifecycle and purchase-history fields need the same precision. An eligible prospective customer has no completed purchase recorded, and an eligible customer has a completed purchase recorded. These are example definitions, not universal categories, so choose fields that fit the stated purpose and notice.
Let exclusions win
Centralize permission, objection, complaint, invalid address, market, age or sector restriction, product status, legal hold, and incompatible-offer rules, as the list building best practices describe for permission. Give exclusions priority over promotion across every sending destination.
Eligibility exclusions to centralize
- Permission and consent status
- Objection and complaint flags
- Invalid or undeliverable addresses
- Market and age or sector limits
- Product status and legal hold
- Incompatible-offer rules
A withdrawal must beat any relevance score. The FTC's CAN-SPAM guide says you must honor a recipient's opt-out request within 10 business days, so a rewritten rule that re-admits an objector is a defect.
Specify the rule completely
Pair plain-language and executable definitions. Specify operators, windows, time zones, thresholds, precedence, overlap, identity, late events, fallbacks, and expiry. Store test cases and version history.
Where a rule depends on a behavioral event, check how that event is defined upstream. The behavioral trigger journey ideas show which events a journey needs before it can fire. Map each field to its source and update time, and use an explicit unknown state rather than a guess.
Test the difficult records
Check boundaries, nulls, stale values, shared addresses, duplicate identities, conflicting sources, localization, recent objections, delayed transactions, tiny groups, and simultaneous segments. Reconcile counts before activation.
Google Cloud's BigQuery data-quality scan guide documents product rules, scheduled checks, results, and alerts for supported tables. These vendor features do not prove a segment is correct. Use them to automate defined checks, then inspect identity conflicts, boundary times, missing values, and known hard cases by hand.
Give every rule an expiry
A rewritten rule should say when it stops applying. The UK Information Commissioner's Office states in its storage limitation guidance that, even if you collect and use personal data fairly and lawfully, you cannot keep it for longer than you actually need it.
Set a purpose-specific retention period and deletion trigger for consent evidence, preference changes, suppression entries, and engagement data. No single period fits every purpose and jurisdiction. Preserve suppression information still needed to honor an objection, and see the CASL subscriber journey rules for covered Canadian messages.
Hand the rule to the platform
Common email marketing platforms include Mailchimp, Klaviyo, HubSpot, and Braze. These are examples, not endorsements. Verify each platform's current segmentation, suppression, consent, and audit capabilities rather than assuming they are equivalent.
Keep platform-independent logic in a shared rule specification, and test exclusions in both the source data and the final sending destination. Assign an owner to rule changes and a rollback route, then reconcile the activated audience with the approved definition before sending. The email marketing analytics page covers the measurement side.
Review a rewritten rule
Before a rewritten rule goes live, a second person should be able to reproduce its audience from the written definition alone. If they cannot, the definition is still a label.
Compare the old and new audiences, and explain every difference in plain words. Records that drop out should fail a named condition, and records that enter should meet one.
Label rewrite table
Use the table as a worked reference. Each vague label needs the pieces in the middle column before it can be trusted.
| Vague label | What the rewrite must record | Proof it works |
|---|---|---|
| High intent | Tagged click, lookback window, exclusions | Record-level reconciliation |
| Engaged | Named event and expiry, not a score alone | Independent reproduction |
| Prospective customer | No completed purchase recorded | Count matches the source |
| Customer | Completed purchase recorded | Repeat dry run |
Common questions
Should every label be rewritten at once?
No. Start with the label that drives the most decisions, release the rewrite to a limited cohort, and compare the result before touching the next one.
Can an automated data check prove the segment correct?
No. Automated checks cover the rules you defined, so inspect identity conflicts, boundary times and exclusions by hand as well.
What proves the segment improved?
A controlled comparison shows a better customer or business outcome without worse errors, unfair exposure, complaints, pressure, privacy risk, or service demand.
