CRM and Marketing Automation Hygiene - Duplicates, Consent, and Deliverability

Q: How often should we dedupe?

At least quarterly for active databases, more often after imports or events. Automated matching catches most; humans resolve edge cases.

Q: What is the minimum for email consent?

It depends on jurisdiction and message type. In practice: clear opt-in for marketing, documented for lists, and honoured preferences - plus easy unsubscribe on every marketing message.

Q: Why does deliverability get worse over time if we do nothing?

Stale addresses, complaints, and bounces accumulate; ISPs infer lower engagement. Cleaning lists and sending only to engaged segments usually helps more than template tweaks.

Q: Is CRM hygiene the same as marketing automation hygiene?

Overlapping. CRM is the long-lived record; automation is behaviour and campaigns. They must agree on identity, consent, and contactability.

Q: Where do we start this week?

One dedupe pass, one consent audit for your top three journeys, and auth checks (SPF/DKIM/DMARC) for domains you send from. Measure bounce and complaint rates next month as a baseline.

Implementation gets the pipes connected; hygiene keeps the water drinkable. Without it, sales sees three records for one person, unsubscribes still get promos, and your domain reputation slowly slides until even good lists underperform. This note covers duplicates, consent, and deliverability as ongoing disciplines - not one-off clean-ups. If you are still choosing tools and scope, start with essential considerations for marketing automation implementation.

Duplicates - merge rules beat periodic panic

Duplicates usually come from multiple entry points - web forms, imports, events, partner lists, manual adds - each with slightly different email formatting or missing company names.

Define a match key - almost always email for B2B lead objects, sometimes email + a second field for consumer or shared inboxes. Decide who wins on conflict: newest activity, CRM as source of truth, or marketing tool - and document it.

Merge, do not only hide - suppressing dupes in reports still leaves reps calling the wrong “primary” contact. Use your CRM or middleware’s merge with field-level rules so history and consent live on one record.

Schedule a recurring job: monthly or quarterly dedupe pass for new patterns (e.g. +tag aliases, personal vs work email). Tag suspected dupes for human review when automatic merge is risky.

Map message types to lawful basis and audience - transactional, service, nurture, cold outreach, partner co-marketing. If your automation cannot express that split, you will eventually email the wrong segment “because the workflow said so.”

Store proof where stakeholders can find it: timestamp, source (form, import, event), and what the person opted into. Imports and list buys need the same rigour as web forms - “we have a spreadsheet” is not a consent model.

Preference centre should be real: categories the business actually uses, synced to CRM and automation before you add a new journey. A single global “unsubscribe from everything” is baseline; granular prefs reduce complaint rates when you grow.

Train anyone with list upload access on no cold addition to nurture tracks without a defined rule - that is how one campaign poisons trust for the whole domain.

Deliverability - reputation is a balance sheet

Authentication - SPF, DKIM, DMARC aligned for the domains you send from. Marketing platforms and IT need the same from-domain story; “no-reply@” on a domain you do not authenticate still fails checks.

List quality - remove hard bounces, repeated soft bounces, and complaint addresses. Re-engagement campaigns belong in a controlled segment, not the main weekly send.

Volume and velocity - sudden tenfold sends from a cold IP or domain warm-up look like abuse. Scale gradually after platform changes or domain moves.

Content and frequency - high complaint rates hurt everyone on the shared IP. If engagement drops, frequency and relevance before blaming “the algorithm.”

Transactional vs marketing - keep password resets and invoices on streams that are not throttled the same as campaigns; do not mix marketing footers in a way that confuses user expectations.

CRM and automation in sync

Two-way sync can create loops or overwrites - e.g. marketing updates a field the CRM overwrites from ERP nightly. Field ownership and last-write-wins rules per field prevent “mystery” data.

Opt-outs must propagate in near real time to every active channel: CRM flags, ad audiences, and partner tools fed from the same customer ID where possible.

Test a full loop after any schema change: form submit → automation → CRM → report. Small schema drift is how “marketing says opted in, CRM says do not contact” starts.

Who owns hygiene

Marketing usually owns journey logic and list segments; RevOps or CRM admin often owns data model and merge policy; legal or privacy signs off on consent copy and retention. If no one owns a monthly metrics review (bounce rate, complaint rate, duplicate count), hygiene backslides within a quarter.

Frequently asked questions

How often should we dedupe?

At least quarterly for active databases, more often after imports or events. Automated matching catches most; humans resolve edge cases.

It depends on jurisdiction and message type. In practice: clear opt-in for marketing, documented for lists, and honoured preferences - plus easy unsubscribe on every marketing message.

Why does deliverability get worse over time if we do nothing?

Stale addresses, complaints, and bounces accumulate; ISPs infer lower engagement. Cleaning lists and sending only to engaged segments usually helps more than template tweaks.

Is CRM hygiene the same as marketing automation hygiene?

Overlapping. CRM is the long-lived record; automation is behaviour and campaigns. They must agree on identity, consent, and contactability.

Where do we start this week?

One dedupe pass, one consent audit for your top three journeys, and auth checks (SPF/DKIM/DMARC) for domains you send from. Measure bounce and complaint rates next month as a baseline.