Direct answer
Decide which system owns each field, and let only that system write it. Key every sync on the external record ID instead of a name or email match, make each sync safe to repeat, and resolve conflicts with a clear rule based on when each side last changed the value. Then run a scheduled reconciliation that compares both systems and reports drift, and check for duplicates before any email or call goes out. Without these, two systems slowly tell two different stories about the same customer.
How two systems end up owning one record
Most integrations start as a one-way push from the app into the CRM. Then sales starts editing stages in the CRM, the app starts reading those stages back, and an automation in the CRM starts sending emails based on fields the app writes. Nobody decides who owns the lifecycle stage, so both systems write it, and whichever ran last wins.
The result is rarely an error. It is a board that shows one stage while the CRM shows another, a contact who gets the same sequence twice, or a returning customer who gets nothing at all. In the production issues we reviewed across about 25 projects in October, CRM and integration state drift appeared in four.
Four drift patterns we keep finding
The returning contact who never enters the workflow
A form handler creates a CRM contact for new people and quietly stops when the contact already exists. Any CRM workflow that starts on a form submission then skips everyone who has been in touch before, which is often the warmest lead you have. The form post succeeded, so nothing looks wrong.
A reconciliation job that restarts outreach
A scheduled job checks for open follow-up tasks and creates a sequence when it finds none. If someone cancelled those tasks on purpose, the next run treats the lead as new and starts the sequence again, so the same person gets the same emails twice.
Two systems with different rules for the same stage
The app and the CRM each calculate a lead or deal stage from their own rules. When those rules disagree, for example on what happens after every quote is declined, or whether a closed stage can ever reopen, the board shows one answer and the CRM shows another, and both look authoritative.
An external status that never comes back
The app hands work to another service, such as a scheduling or e-signature tool, but never handles the callback that says it finished. The record stays “scheduled” or “sent” forever. The external system knows the truth, and the app never asks.
Five rules that prevent drift
- One system of record per field. Write down, field by field, which system owns it. The app might own application status while the CRM owns the sales stage. The other side reads that field and never writes it.
- Sync on external IDs, idempotently. Store the CRM’s record ID on your record and match on it. HubSpot’s contacts API guide calls email the primary unique identifier for avoiding duplicate contacts, and its batch upsert endpoint creates or updates records keyed on email or a custom unique property. A keyed write like that lets a retried sync update instead of duplicating. Our article on idempotent retries covers the same pattern for agents.
- Resolve conflicts with timestamps and ownership. Compare last-modified times before writing, and never let an older value overwrite a newer one. If the field belongs to the other system, don’t write it at all.
- Reconcile on a schedule and report drift. A nightly job reads both sides and lists every record where they disagree. Then fix the rule behind each one. Also listen for the other system’s events, and verify them; our webhook integrity checker helps with that.
- Dedupe before outreach. Before any sequence starts, check for an existing or recently cancelled sequence for the same person, and keep a durable suppression record so a reconciliation run can’t restart one.
The same discipline applies to insurance systems
Agencies face the same problem between their quoting tools, portals and the agency management system. The management system usually holds the policy of record, so integrations should treat it as the owner of those fields, sync on its identifiers and reconcile regularly. We have built an Applied Epic integration in production.
Where to start
Pick the field that causes the most arguments, usually the lifecycle stage, and run a one-off comparison between the two systems. The size of that list tells you how much drift you already have, and it often explains complaints about missed or repeated emails.
Related Zenveus services: Applied Epic Integration, the InsurTech Audit and the 48-Hour Verdict