Ecommerce email practices that protect deliverability

Ecommerce email breaks differently from B2B email, and the reason is the acquisition path. A B2B subscriber usually chose to subscribe. An ecommerce subscriber usually bought a lamp and did not uncheck a box. That difference produces higher complaint rates, faster list decay and more spam-trap exposure — and it means the practices that protect your deliverability are mostly about restraint rather than about tactics.

Separate engagement practices from infrastructure practices

Two categories, two owners, two failure modes:

• Engagement practices — consent, segmentation, frequency, creative, flows. These change your complaint rate.
• Infrastructure practices — authentication, sending domains, stream separation, warm-up, suppression enforcement. These change whether you can reach anyone at all.

A brand with excellent flows and a single sending domain shared between receipts and promotions is one bad campaign away from undelivered password resets. A brand with perfect infrastructure that mails its whole list twice a day will fail on complaint rate. Both have to be true.

Get consent right at the checkout

This is the single most impactful practice in ecommerce email, and the most commonly broken.

• Do not pre-tick the marketing box. Under GDPR and ePrivacy it is not valid consent, and in every market it produces subscribers who do not remember subscribing — which is the definition of a future complaint.
• Keep transactional and marketing consent separate. A purchase gives you the right to send a receipt and a shipping update. It does not give you the right to send a Black Friday campaign.
• Record the evidence. Timestamp, source page, consent text shown, IP. When a complaint or a subject-access request arrives, this is the only thing that helps you.
• Use double opt-in for pop-up and competition signups. These are the two sources most likely to contain typos, throwaway addresses and spam traps.
• Send the welcome email within minutes, while the person still remembers you. The engagement from a fast welcome is the best reputation fuel a young domain gets.

Segment on behaviour, not demographics

Four dimensions cover almost every ecommerce case: purchase recency, purchase frequency, category bought, and lifecycle stage (never purchased, first purchase, repeat, lapsed).

Personalisation beyond a first-name merge means recommending from the category someone actually bought in. A merge tag that renders "Hi there," is worse than no greeting, and a recommendation block showing products someone already owns is worse than no recommendation.

Cap frequency, and mean it

Set a hard cap per subscriber per week across all flows and campaigns, and enforce it centrally rather than per-campaign. Without a global cap, a shopper who abandons a cart during a sale week can receive a cart reminder, a sale campaign, a category email and a price-drop alert in two days. That is how an engaged buyer becomes a complaint.

Also suppress across flows: a purchase must immediately stop the abandoned-cart series, and an unsubscribe must stop everything marketing-shaped in every tool you own, including the reviews app and the loyalty app.

Sunset disengaged subscribers on a schedule

Removing addresses feels like losing reach and is the single most reliable way to raise placement. Mail sent to people who never open teaches the provider that your mail is unwanted, and long-dormant addresses are where recycled spam traps live.

Run it as a rule, not a project: one re-engagement attempt at twelve months of inactivity, then suppression. Keep the record for suppression purposes; stop sending.

Control discount fatigue

A list trained to wait for 30% off will not buy at full price and will complain when the discount is smaller. Practical limits: do not put a discount in every subject line, reserve the deepest offers for engaged segments, and use non-discount reasons to email — restocks, how-to content, early access, order milestones. Deliverability relevance is direct: offer-only programmes see engagement collapse between sales, and engagement collapse is what providers measure.

Sale events and volume spikes

Seasonal peaks combine your highest volume with your widest audience, which is exactly the wrong combination for a young or mid-reputation domain. Plan backwards from the peak: start raising volume six to eight weeks out so the ramp is gradual, tighten the audience for the biggest sends rather than widening it, and split large sends across several hours. Treat 421 and 450 deferrals as an instruction to slow down, not a bug to retry through.

Infrastructure items ecommerce brands skip

• Authentication for every sending domain and every app. Reviews tools, loyalty apps, helpdesks and shipping notifiers all send as you. Each needs DKIM, and each include: eats into SPF's 10-lookup limit.
• Transactional on its own subdomain, so campaign complaints cannot delay receipts.
• Tracking links on your own subdomain via CNAME, never a public shortener.
• Suppression enforced at send time. A suppression list that exists but is not applied is the most common cause of a sudden relapse.
• `List-Unsubscribe` and `List-Unsubscribe-Post` headers on every bulk send, with the endpoint acting on a POST within two days and no login.
• HTML under about 100KB, so Gmail does not clip the message and hide the unsubscribe link behind "View entire message".

What not to do

1. Do not import the full customer table and call it a list.
2. Do not treat a checkout as a marketing opt-in.
3. Do not run a win-back loop forever; one attempt, then suppress.
4. Do not judge the programme by delivery rate — acceptance is not placement.
5. Do not add a new sending domain or ESP in November.
6. Do not buy a list to boost a seasonal campaign. Nothing above survives that decision.

MailPilot covers measurement and warm-up: it warms mailboxes over OAuth (Gmail, Outlook) or SMTP/IMAP including SendGrid, Amazon SES, Mailgun and Postmark, checks SPF, DKIM and DMARC on delivered mail, tracks per-mailbox reputation daily, and reports real per-provider inbox placement from a seed-inbox network — so you can see a complaint-rate problem forming before a sale week rather than after it.