When to run this checklist
Run it once the change has been designed and before it is switched on: a new connector between the CRM and the marketing platform, a form tool, an ecommerce system or anything else that creates or updates contact records; a new field added to an existing sync; a field whose direction changes; a new rule for which records cross; a new matching or merge rule; or a connection reconnected after a release or a change of credentials. Run it again for each later change, however small it looks, because the failures it guards against come from details such as a field set to replace rather than append, or a consent field left on the connector's default rule. The academy capstone on planning a hot tub retailer's year of programme change practises placing this gate inside an acquisition, so the bought business's customer list crosses only after its connector passes, in a year that also moves tagging server-side, changes the bidding goal and redefines revenue.
The person who ticks it is the owner of the sync, with the team accountable for each field being changed and whoever owns data protection for the consent lines. The method sits in a handful of pages. The syncing guide writes the contract, the tutorial on documenting marketing integrations keeps the register every connection belongs in, and the consent map tutorial names the changes that must update the map before they go live. The prevention routines of the diagnoses for values changing back after a sync, duplicate contacts, blank lead sources, customers receiving prospect messages and unsubscribed contacts still receiving email all end in the same habit: seeded records read after every change to the sync.
Neighbouring lists and procedures take the cases this one does not. A new form, lead source or routing rule is tested end to end by the lead routing QA checklist. A workflow that reads a changed field is relaunched through the email automation flow launch checklist where it sends email. A tool joining the stack for the first time comes in through the marketing tool onboarding SOP, and replacing the CRM itself is the tutorial on moving to a new CRM. A one-off list load is the contact import tutorial, and the lead management audit scores the sync as part of a running programme rather than gating a change.
Before you start
- The change written down: which tools, objects and fields it touches, and what will be different afterwards.
- The integration register row for the connection, or the blank row it will need.
- The field ownership table from the sync contract, with the owning system, accountable team and direction for each synced field.
- The consent map row for every purpose the change reads or writes.
- The seeded records in each system: an opted-out contact, a customer with a won deal, a contact with an open opportunity and a record with a hand-corrected value.
- A place to test where nothing reaches customers, reports or ad platforms, and the test record convention.
- The automation inventory, to find workflows that read the fields being changed.
- The change log where the change, its date and the way back will be recorded.
The checklist
The groups follow the order of the work: record the change, settle the owners and direction of each field, close the doors through which duplicates and lost suppressions arrive, protect consent and the customer flag, prove it with seeded records, then make sure someone will see if it breaks. Recording comes first because the integration registerOne standing record of every connection between marketing tools: what moves, which way, keyed how, on what schedule, with whose credentials and owner. row is where the sync's owner is named and, by the end, its failure check written. A line in the first five groups that cannot be ticked holds the change; the last group can be finished on the day of go-live, but not skipped.
Scope and the register
Field owners and direction
Matching, merges and deletions
Consent and customers
Testing before go-live
Watching and records
Where each check is explained
The widget keeps the lines short so they can be ticked; the reasoning lives in the guides. Each check links to the page that explains it, and several of the diagnoses repeat the same point from the side of the failure.
If an item fails
The consent group is the hard stop. A change that lets a stale subscribed value replace a later objection, or that copies a deletion in a way that removes someone's entry from the suppression list, does not go live until it is fixed. The syncing guide records the position the site relies on: the GDPR, as the site's glossary states it, makes the right to object to direct marketing absolute, and the UK regulator's direct marketing guidance is the reference for honouring opt-outs promptly, which is why the conflict rule for consent is that the strictest value wins, wherever it was recorded. The consent map tutorial traces each withdrawal and objection through every tool for the same reason: an objection has to be applied in every channel it covers. That is general direction rather than legal advice, and the rules for each purpose are for whoever owns data protection to set. If the seeded opted-out contact becomes reachable in any tool during the test, stop and trace it before anything else; an opt-out that arrives by phone or reply in the meantime still goes through the opt-out request SOP.
The field and matching groups hold the change too, but they are usually fixed in the design rather than the connector. A field that both systems may write has no owner; choose one, make the other side read-only, and write the decision in the ownership table, where a RACI matrix keeps one accountable team per row. A matching rule that would create a new record where one already exists turns into a merge queue that somebody has to clear by hand, and the duplicate contacts diagnosis explains why the suppressed state must survive every merge. Original source is written once on creation; if the connector would overwrite it, map the incoming value to the most recent source field instead. Corrections a person made are values the sync must never write over: under the GDPR's accuracy principle, as the syncing guide states it, an old value the sync restores over a correction leaves the business holding data the person already asked it to fix, which makes it more than a reporting fault.
The testing group decides the date, not whether the change happens. If there is nowhere to test without data reaching customers, reports or ad platforms, find one first: the test environments guide treats changes whose effects leave the business, such as a merge, a bulk update, data sent to an ad platform or a change to consent handling, as one-way doors to be tested before anyone walks through them. A seeded record that comes back wrong is a gap found in minutes; read the record's history, name the writer, fix the rule and run the seeds again. Where a source such as an ad platform's lead form is involved, the diagnosis for LinkedIn leads not syncing shows how to trace one source's mapping and connection.
The watching group can be finished on the day, but a sync that goes live with no named reader of its error queue and no failure check is a sync that will break without anyone noticing. Write the check with the alert's recipient in the register row, record the change in the Change LogA dated record of changes made to tracking, accounts, reports, lists or rules, with who made each and why, so later movements in the figures can be traced. with the way back, and read the counts on both sides in the first days. The data flow and integrations guide explains the mechanics, such as append against replace and identifier mismatch, behind most drift the counts reveal.
An office furniture leasing firm changes its sync (illustrative)
Two new fields, three catches and a clean go-live
Haxtrumbo Leasing, an invented office furniture leasing firm, keeps about 6,200 contacts in its CRM and syncs about 4,100 of them to a separate marketing platform. Its marketing lead wants to add a lease renewal date to the sync for a renewal campaign and to let job titles flow both ways, because contacts update them through a preference form. The sales operations manager owns the sync and runs this list with her, with the office manager who handles data protection on the consent lines. The ownership table shows the first catch: sales already edits job titles in the CRM, so a two-way field would have two writers. They keep the CRM as owner, send preference-form changes to the CRM first and flag them as manual so the sync never overwrites them. The second catch is in the connector: its new mapping would replace original source on every load, which would have rewritten the source of every contact touched by the renewal campaign; it is moved to the most recent source field. The third catch is the consent field, still on the connector's latest-edit rule. In the test account the seeded opted-out contact, given an older subscribed value from a stale form record, becomes reachable; the rule is changed to the most restrictive value and the seed stays suppressed on the second run. The seeded customer stays out of both prospect sequences, the hand-corrected record keeps its corrected title after two sync cycles, and 2 test submissions read back with their sources. The change goes live on a Tuesday morning; by Friday the error queue holds 14 items, all one cause, a renewal date in a format the platform rejected, and the synced and suppressed counts show the same gap as before the change.
Common mistakes
- Treating a sync change as a settings tweak. A field mapping decides what thousands of records will say, and a change with no register row, test or log entry is impossible to explain later.
- Letting the connector's defaults decide. Whatever the connector does out of the box was not chosen for your fields: a most-recent-edit rule lets an old consent value beat a newer objection, and replacing a field that should append wipes what it held. Decide each field's rule instead.
- Two writers for one field. If both systems may write it, whichever save came last wins, and the value will flip back and forth.
- Testing with real customers. A test that copies customer records or sends to them creates the problem it was meant to catch; use marked test records and keep them out of reports, as the CRM troubleshooting section shows from the failures.
- Reading the settings rather than the records. A connector that looks right can still overwrite a value; the seeded records show what actually happened to the Consent RecordThe stored evidence of a marketing permission: the channel, the wording the person saw, the date and the source of each opt-in, kept with the contact record., the Lifecycle StageA CRM field recording where a contact stands in the relationship, such as subscriber, lead, opportunity or customer. and the corrected fields.
- Forgetting the workflows downstream. A changed field can change who enters a sequence; run seed contacts through every workflow that reads it.
- Deleting to tidy up. A deletion that travels through the sync can remove the record that kept someone suppressed, and with it the evidence of their objection; DeduplicationRemoving double counts when one conversion is reported by more than one source. and retention decisions belong with the data protection owner, not with the connector, and test data stays free of real Personal DataAny information relating to an identified or identifiable living person, such as a name, email address, location data or an online identifier..
The checklists library holds the lists that sit beside this one, from lead routing QA to email automation flow launches.
