AI for Ecommerce Email and SMS Operations

Deliverability Break Detection

Email and SMS deliverability degrades in stages. A 2% bounce rate is normal. A 4% bounce rate signals a list quality problem. A 6% bounce rate means ISPs are filtering. Most brands notice only when revenue drops.

AI can monitor bounce, complaint, and unsubscribe rates in real time and flag thresholds before they compound. The rule is simple: if bounce rate exceeds 3% on a single send, or complaint rate exceeds 0.5%, or unsubscribe rate exceeds 0.3%, pause the next send and audit the list segment.

For SMS, the equivalent is carrier rejection rate. If more than 5% of messages are rejected by carrier networks on a single send, the number or content is flagged. Rejection often precedes account suspension by 48 hours.

The operator still decides whether to investigate or pause. AI surfaces the signal. The human owns the decision.

Monitor bounce, complaint, unsubscribe rates per send
Set thresholds: bounce > 3%, complaint > 0.5%, unsubscribe > 0.3%
For SMS: flag carrier rejection > 5%
Alert operator within 2 hours of send completion
Log all flags in a single audit table for trend analysis

Send Fatigue Detection and Frequency Capping

Send fatigue is invisible until it appears in unsubscribe spikes or revenue decline. A customer who receives 4 emails in 7 days has a 12% unsubscribe risk. At 6 emails in 7 days, the risk jumps to 28%. At 8 emails in 7 days, it reaches 40%.

AI can track send frequency per subscriber across all campaigns (promotional, transactional, editorial, SMS) and flag when a customer is approaching fatigue thresholds. The operator then decides whether to suppress, delay, or consolidate sends.

The rule: if a subscriber receives more than 5 marketing emails in 7 days, or more than 3 SMS in 7 days, flag them in a fatigue cohort. If they receive more than 7 emails in 7 days, auto - suppress them from the next promotional send and notify the operator.

Fatigue suppression should be temporary (7 - 14 days) and tied to send type. A customer fatigued on promotional email may still want transactional SMS.

Track sends per subscriber per 7 - day window
Flag at 5+ emails or 3+ SMS in 7 days
Auto - suppress at 7+ emails in 7 days (notify operator)
Separate fatigue rules by send type (promo, transactional, SMS)
Review fatigue cohort size weekly; adjust thresholds if > 15% of list

Segment Definition and Drafting

Most segment drafts are either too broad ("all customers") or too narrow ("customers who bought product X in the last 30 days and have not opened an email in 60 days and have a lifetime value > $200"). The first wastes send volume. The second reaches 200 people.

AI can generate segment definitions based on operator intent. The operator says: "I want to re - engage customers who have not purchased in 90 days." AI drafts a segment: "Customers with last purchase > 90 days ago AND last email open > 30 days ago AND email hard bounces = 0." The operator reviews, adjusts, and approves.

The draft should include segment size, expected open rate (based on historical cohort data), and revenue potential. If the segment is smaller than 500 people or expected revenue is < $500, the operator should reconsider.

Segment drafts should be version - controlled. Each draft is timestamped, attributed, and stored. When a segment performs poorly, the operator can trace back to the definition and refine it.

Operator states intent; AI generates 3 - 5 segment definitions
Each definition includes: size, expected open rate, revenue potential
Operator reviews, edits, and approves before deployment
Minimum viable segment: 500 people, $500 revenue potential
Store all segment versions with timestamp and performance data

Campaign Performance Baseline and Anomaly Detection

A campaign that achieves 22% open rate is good. A campaign that achieves 22% open rate when the account average is 28% is a problem. Most operators do not track the baseline.

AI can establish a rolling baseline for open rate, click rate, conversion rate, and unsubscribe rate per send type and segment. When a send falls 2 standard deviations below the baseline, it is flagged as an anomaly.

Common causes: subject line change, send time change, list quality degradation, content change, or ISP filtering. The operator investigates and decides whether to adjust future sends.

Baseline should be recalculated monthly and exclude outliers (holiday sends, one - off campaigns).

Calculate rolling 12 - month baseline per send type
Flag sends > 2 SD below baseline
Common anomalies: subject, send time, list quality, ISP filtering
Recalculate baseline monthly; exclude outliers
Log all anomalies and root cause in a single table

Retention Cohort Tracking and Churn Prediction

A customer who has not purchased in 180 days is at 70% risk of churn. A customer who has not opened an email in 90 days is at 55% risk of churn. Most brands do not separate these signals.

AI can track retention cohorts by purchase recency, email engagement, and SMS engagement. When a customer enters a high - churn cohort, the operator can deploy a targeted win - back campaign.

The rule: if a customer has not purchased in 120 days AND has not opened an email in 60 days, they are in the "at - risk" cohort. If they have not purchased in 180 days OR have not opened an email in 90 days, they are in the "churn" cohort.

Win - back campaigns should be limited to 3 - 5 sends over 30 days. If no engagement, suppress for 60 days and re - evaluate.

Define at - risk: no purchase in 120 days AND no email open in 60 days
Define churn: no purchase in 180 days OR no email open in 90 days
Deploy win - back campaign: 3 - 5 sends over 30 days
If no engagement after win - back, suppress for 60 days
Track win - back ROI by cohort; adjust thresholds if ROI < 2x

List Hygiene and Compliance Automation

Hard bounces, complaints, and unsubscribes must be removed from the list immediately. Soft bounces should be retried 2 - 3 times before removal. Most brands do not automate this.

AI can monitor bounce and complaint feedback loops from ISPs and automatically suppress addresses. The operator reviews suppression logs weekly and decides whether to investigate or accept.

Compliance rules vary by region. GDPR requires explicit consent and easy unsubscribe. CAN - SPAM requires unsubscribe processing within 10 days. CASL requires explicit consent and identification of the sender. AI can flag non - compliant sends before they go out.

List decay is normal. A healthy list loses 0.5% - 1% per month to hard bounces and complaints. If decay exceeds 2% per month, the acquisition or engagement strategy needs review.

Auto - suppress hard bounces, complaints, and unsubscribes
Retry soft bounces 2 - 3 times before removal
Monitor bounce and complaint feedback loops
Flag non - compliant sends (GDPR, CAN - SPAM, CASL)
Track list decay monthly; alert if > 2% per month

Operator Workflow and Decision Points

AI surfaces signals. The operator makes decisions. The workflow should be clear: signal arrives, operator reviews in 2 hours, decision is logged, action is taken.

Daily: review deliverability alerts, send fatigue flags, and anomalies. Weekly: review segment performance, retention cohorts, and list hygiene. Monthly: recalculate baselines, review compliance, and adjust thresholds.

All decisions should be logged with timestamp, operator name, and rationale. This creates an audit trail and helps the team learn from past decisions.

Daily: deliverability, fatigue, anomalies
Weekly: segment performance, retention, hygiene
Monthly: baselines, compliance, thresholds
Log all decisions with timestamp, operator, rationale
Review decision log quarterly to identify patterns and refine rules

Questions

FAQ

What is the difference between send fatigue detection and frequency capping?

Send fatigue detection identifies customers who are approaching or have exceeded safe send limits based on historical unsubscribe and revenue data. Frequency capping is the automated rule that suppresses or delays sends to fatigued customers. Detection is the signal; capping is the action.

How often should segment definitions be reviewed and updated?

Segment definitions should be reviewed after each send to check performance against the baseline. If open rate, click rate, or conversion rate falls below expectations, the definition should be audited and adjusted. Major reviews should happen quarterly when baselines are recalculated.

What is a healthy list decay rate?

A healthy list loses 0.5% - 1% per month to hard bounces and complaints. This is normal and expected. If decay exceeds 2% per month, the acquisition strategy (low - quality sources), engagement strategy (too many sends), or list quality (old data) needs review.

Should transactional emails be included in send fatigue calculations?

No. Transactional emails (order confirmation, shipping notification, password reset) should be excluded from fatigue calculations because customers expect and want them. Only marketing emails (promotional, editorial, re - engagement) and SMS should count toward fatigue thresholds.

Want this on your account?

Thirty minutes. Bring the number that keeps you up.

Book a demo

AI for Email and SMS Operations: Detection, Fatigue, and Segmentation