A person typing on a smartphone displaying a list titled 'The naughty list'. Email segmentation examples: eight patterns, each a decision plus a rule
Photo by RDNE Stock project on Pexels

Careers

Part of Notes on email segmentation from the field

Email segmentation examples: eight patterns, each a decision plus a rule

Eight email segmentation examples, each a decision plus a minimum rule: the welcome pattern field by field, plus controls and porting rules for your ESP.

What to take away

  • Eight patterns cover most subscriber work: welcome, product fit, renewal, preference, geography, engagement, service risk and accessibility format.
  • Each pattern is a decision plus a minimum rule. The welcome pattern below is the one carried through to named fields, a window, exclusions, expiry, a failure state and an owner.
  • Named public examples sit inside those patternsSpotify and Netflix under engagement, Starbucks and Airbnb under geography, Adobe under renewal, Delta under service risk, Duolingo under trigger-and-exit, and HubSpot, Klaviyo and Salesforce Marketing Cloud under platform porting.
  • No measured lift appears below. Test before you quote a number.

What each of the eight patterns decides

The practitioner notes on email segmentation from the field show how these patterns behave once real data arrives.

PatternDecision it makesMinimum rule
WelcomeWho receives onboarding mailConfirmed subscription inside 7 days, minus customers and prior complainers
Product fitWho hears about a fixCurrent product record, service facts split from promotion
RenewalWhen a reminder leavesRenewal date, time zone, cancellation state
PreferenceWhich topic or format was requestedLatest effective choice plus a fallback
GeographyWhich market rules applyVerified service country plus an unknown bucket
EngagementHow often you sendConfirmed clicks, purchases, replies, declared interests
Service riskWho receives outage factsImpact record, no promotion, expires at resolution
AccessibilityWho receives an alternate formatDeclared request, restricted access, no promotion

Two constraints hold across all eight: evidence beats inference, and every rule needs an exit. Treat the table's minimum rules as starting points. Before one runs, give it a named field with a source, a threshold or window, exclusions, an expiry and a named owner — the welcome pattern below shows that completed shape.

The CDC planning chapter on segmenting an audience separates primary from influencing audiences. Borrow the demand for a stated objective, not its categories.

Example: the welcome rule, field by field

  1. Trigger fieldsubscription_confirmed_at, written only by double opt-in.
  2. Windowconfirmed inside the last 7 days.
  3. Exclusionsexisting_customer is true, a complaint inside 180 days, or welcome_journey_completed is true.
  4. Expiryanyone past day 7 drops out, and events older than 24 hours are held back rather than treated as fresh.
  5. Failure state and ownerif the event feed stalls for 6 hours, pause the send and route the incident to the lifecycle manager.

Welcome audience eligibility

Confirmed subscription within seven days?

Yes

include in welcome audience

No

exclude from welcome audience

The same shape applies elsewhere. Duolingo sends streak reminders after a lesson event and stops when the streak breaks or reminder settings change.

A support update selects owners of an affected version from transaction records rather than browsing signals. The email automation case studies show how rarely a published lift survives contact with a second team's data.

Engagement tiers that do not lean on opens

A frequency rule built on confirmed clicks, account actions with notice, replies, purchases and declared interests outlasts one built on opens. Those signals are inputs, not tiers: each tier needs a named field, a threshold and a window before it can set how often you send. Opens are a limited signal now that several mailbox providers prefetch images. Suppressions and total contact pressure override any score.

Spotify reads plan type, confirmed plays and saved tracks. Netflix reads viewing history and device type. The costliest email segmentation mistakes usually begin with a score nobody can explain.

Renewal, service and geography decisions

A renewal reminder selects contracts approaching a verified renewal date in the billing system. It excludes cancelled and disputed accounts, respects account time zone, and rechecks state within 24 hours of send. Adobe Creative Cloud expiry notices are the public shape of this pattern.

An outage creates a segment from operational impact records, with restricted access, no promotion and an expiry at resolution. Delta Air Lines delay mail follows that pattern.

Market rules follow a verified service country plus an explicit unknown bucket. Starbucks store choice and Rewards tier, along with Airbnb destination and check-in dates, rest on declared or booked values.

Controls to attach to every segment

  • Purpose and the single decision, in one line.
  • Fields with source, owner and a freshness limit.
  • Inclusion, exclusion, unknown bucket and overlap priority.
  • Expiry date, plus a recheck before send.
  • Test case, expected range and incident path.
  • Privacy and fairness review with a named reviewer.

Assign data and response owners with the NIST Privacy Framework starting guide. It outlines a voluntary risk process and is not legal clearance.

Porting a pattern to your ESP or CRM

Most platforms split each rule in three places. The window sits in the trigger filter, exclusions sit in a suppression list, and expiry sits in a scheduled re-evaluation. HubSpot carries lifecycle stage, list membership and job function. Klaviyo carries predicted CLV and churn risk. Salesforce Marketing Cloud carries contract role and account data.

The guide to how to improve email segmentation for business teams covers that mapping work.

Accessibility and preference requests

The GOV.UK Service Manual's user-needs guidance says teams should research all kinds of users and treat unsupported opinions as assumptions. Do not infer disability from a proxy when a declared preference or an inclusive default serves the task.

A format request uses tightly limited access, serves only that request, and never promotes. Bilingual commercial mail to Quebec brings CASL consent and unsubscribe duties into the same review.

Common questions

Are these rules ready to send as written?
No. Replace purpose, population, permissions, timing and market first, then give every field a source and a freshness limit.
Can geography be inferred from an IP address?
Possibly, but precision, permission and proxy risk need qualified review. A verified service country with an unknown bucket is safer.
How do I know a segment worked?
Set one expected range before launch, hold a control group, and record the outcome over the same window.

More in Careers

Latest from Practice Desk