Next Best Action for Email Teams: 12 Worked Examples (Not Enterprise Slides)

Updated 1 November 2026.

Definition

Next best action is a real-time decisioning approach that uses customer data, business rules and prediction to choose the most valuable action for each individual at a given moment, where the set of possible actions includes sending nothing. CDP.com's glossary definition is the same in substance: "a real-time decisioning strategy that uses customer data, business rules, and artificial intelligence." The phrase "including doing nothing" is the part enterprise slides leave out and email teams need most.

The waterfall every action passes

EngageMKTG's applications page shows the order; it holds for any stack:

1. Profile validation. Is this a real, reachable, unified profile (not a duplicate across brands)?
2. Consent. Does this person's consent cover this message type and this brand, today?
3. Status. Not suppressed, bounced, complained or unsubscribed in any system; not inside a sunset flow.
4. Frequency. Caps per profile, per brand, per channel, recency since the last send.
5. Selection. Among the actions still allowed, which is best now? This is NBA.
6. Send or suppress. Execute, and log the reason.

If any of gates 1–4 fails, selection never runs. The twelve examples below are selection decisions, each with the gate that would have stopped it.

12 worked examples

Each example gives the signals, the candidate actions, the decision, the guardrail that bounds it, and what to measure. Timings are typical; set your own.

1. Cart abandoned, complaint on file Signals: cart abandoned 2 hours ago; one spam complaint 14 months ago on another brand's stream. Candidates: abandoned-cart email now; SMS; nothing. Decision: skip email (status gate: complaint in any system = no marketing email); allow on-site reminder only. Measure: complaint rate on the abandoned-cart flow before and after the gate.

2. Past buyer, quiet nine months Signals: two orders, last 9 months ago; no click in 180 days; consent on file. Candidates: today's campaign; revival cohort; sunset flow. Decision: defer today's campaign; enroll in the next day's revival cohort (value-reminder email, 1/30 pacing). Guardrail: revival stop conditions (hard bounces 2%, complaints 0.1%/0.3%). Measure: 45-day click-or-order rate on the frozen cohort.

3. New signup, minute zero Signals: form submission 30 seconds ago; source = pricing page. Candidates: welcome now; wait for the weekly digest; sales call. Decision: send the welcome within minutes with one action; schedule the day-1 value email only if the action is not completed. Guardrail: day-0 sequence frequency (max one message in the first 2 hours). Measure: activation rate and time to value by cohort (the day-zero guide).

4. Unengaged 90 days, no purchase ever Signals: 7 emails received, no open or click in 90 days; no orders. Candidates: campaign; re-permission; nothing. Decision: skip campaigns (exclude the unengaged tier); send re-permission email 1 of the sequence, in a daily batch. Guardrail: the sunset policy's disposition on day 21. Measure: unsubscribe and complaint rate on the batch; recoveries.

5. Frequency cap reached Signals: three marketing emails in the last 7 days (cap = 3). Candidates: fourth email (a flash sale); defer; nothing. Decision: defer 24 hours; if the sale ends first, skip. Guardrail: cap is hard, not advisory. Measure: unsubscribes per 1,000 for capped vs uncapped cohorts.

6. Replenishment window Signals: consumable bought 40 days ago; typical repurchase interval 45 days. Candidates: replenishment reminder now; wait for the window; promotion. Decision: send the reminder at day 42 without a discount; promotion only if no order by day 50. Guardrail: one reminder per cycle. Measure: order rate per 1,000 vs holdout.

7. Open support ticket Signals: ticket opened 3 hours ago, status "waiting on us." Candidates: today's promotion; nothing. Decision: skip marketing until the ticket is resolved or 72 hours pass. Guardrail: status gate reads the support system. Measure: complaints and unsubscribes among ticket-holders mailed vs not.

8. Price drop on a viewed product Signals: product viewed twice in 7 days; price dropped 15% today; consent for marketing email. Candidates: price-drop alert; wait for weekly campaign. Decision: send the alert if the frequency gate allows; otherwise fold into the next campaign. Guardrail: no alert if a cart exists (the cart flow owns it). Measure: conversion vs the weekly-campaign holdout.

9. Paid trial ending (newsletter or SaaS) Signals: trial ends in 3 days; 4 of 6 issues read, or activation event completed. Candidates: upgrade email now; wait for expiry; nothing. Decision: send the upgrade nudge at day −3 to engaged trialists; skip for trialists with no engagement (they get the last-notice email at expiry instead). Guardrail: one nudge, one expiry notice. Measure: free-to-paid conversion vs holdout (beehiiv's 2025 platform median is 0.62%; your own baseline matters more).

10. Duplicate profile across two brands Signals: same address subscribed to Brand A and Brand B; both have a send today. Candidates: two emails; one; none. Decision: send one (the brand with the more recent engagement); defer the other a day. Guardrail: profile validation gate unifies the two records first; frequency counts across brands. Measure: unsubscribe rate on multi-brand profiles before and after.

11. Hard bounce recorded in another system Signals: the CRM shows a hard bounce last week; the ESP still lists the address as active. Candidates: send; suppress. Decision: suppress (status gate reconciles bounces across systems; Klaviyo-style automatic suppression after a hard bounce is the model). Measure: hard-bounce rate on the next send.

12. Compliance status failing in Postmaster Tools Signals: DMARC alignment failing on one stream this morning; spam rate 0.08%. Candidates: proceed with today's volume; hold. Decision: hold all marketing on the affected stream until the Compliance row passes; transactional continues. Guardrail: provider signals are a program-level gate, not a per-profile one. Measure: time to green; no complaint spike.

These twelve run as rules and a reason log on your existing ESP. Lifecycle Decisioning builds the waterfall and the log around one measured use case first. See the application →

Building NBA without an engine

An ordered list of exclusions and one "best action" rule covers most of the examples:

1. Exclusions, in order: suppressed anywhere → complained anywhere → open ticket → frequency cap reached → inside a sunset flow → unengaged tier (campaigns only).
2. Triggers, by priority: welcome (minute 0) → cart → price drop → replenishment → trial ending → revival cohort → campaign.
3. One rule: the highest-priority trigger that passes every exclusion wins; everything else waits.
4. Log it: a sheet or a custom property per profile with the date, the winning action and the exclusion that stopped the others.

What an engine adds: per-profile scoring among candidates of similar priority, real-time reads of consent and status rather than nightly rebuilds, the reason log as a system of record, and holdout measurement built in. The comparison guide covers when that is worth paying for.

Measuring NBA honestly

• Holdouts. Keep 5–10% of eligible profiles on the old logic for 60–90 days; compare orders, revenue, unsubscribes and complaints.
• Uplift, not propensity. A high propensity to buy is not evidence the email caused the purchase. CDP.com cites BCG's observation that 20–40% of propensity-driven programs show negligible incremental lift when re-measured; treat the range as directional, but run the holdout.
• Count the skips. Report how many sends the gates prevented and what happened to complaints and unsubscribes. For most email programs, that number is the first-year ROI of decisioning.
• Frozen cohorts for revival. Examples 2 and 4 are measured on the reactivation playbook's cohort method at 45 and 90 days.

FAQ

What is next best action in email marketing? A per-person decision at send time: given this profile's data, your rules and a prediction, which single action is most valuable now, including no action. It replaces segment-level campaign scheduling with per-profile selection inside consent, status and frequency gates.

Do I need AI or a CDP to do next best action? No. Ordered exclusions and prioritized triggers in an ESP implement most of it. A decision engine adds per-profile scoring, real-time status reads, a reason log and holdout measurement.

What is the difference between next best action and a journey? A journey is a predefined path with branches. NBA re-evaluates each person's options at each decision point and can choose nothing. The comparison guide has the full table.

How do I measure next best action? With a holdout on the old logic and uplift, not propensity: orders, revenue, unsubscribes and complaints for treated vs held-out profiles over 60–90 days, plus the count of prevented sends.

What are good examples of next best action in email? Skipping a promotion for a profile with an open support ticket; deferring past a frequency cap; routing a quiet past buyer into a daily revival cohort instead of a campaign; sending one email, not two, to a profile subscribed to two of your brands; holding all marketing while a Compliance status row fails.

Why does "do nothing" matter so much? Because complaints and unsubscribes come from the sends that should not have happened. Gmail's limit is three complaints per thousand delivered; the skip decisions are what keep a program under it.

If examples 2 and 4 describe your list, start with the audit. The free List Health Audit maps the dormant tiers and scans for the addresses that would fail the status gate. Get the free audit →

Sources

• CDP.com, Next best action glossary and Next best action use cases: definition; BCG uplift observation (directional). https://cdp.com/glossary/next-best-action/ and https://cdp.com/articles/next-best-action-use-cases/ (checked 1 Nov 2026)
• Braze, Decisioning engines: how AI powers customer engagement: signals, rules as guardrails, explainability. https://www.braze.com/resources/articles/decisioning-engine (checked 1 Nov 2026)
• Google, Email sender guidelines: spam-rate thresholds. https://support.google.com/a/answer/81126
• Klaviyo Help, Understanding active profile management: automatic suppression after a hard bounce. https://help.klaviyo.com/hc/en-us/articles/24263920096027
• beehiiv, The State of Newsletters 2026: 0.62% median free-to-paid conversion. https://www.beehiiv.com/blog/the-state-of-newsletters-2026
• EngageMKTG, Applications (decision waterfall), Dormant Email List Reactivation Playbook, Email Sunset Policy, Agentic Marketing (internal).