Integration Mistakes That Cause Data Problems
The five most common integration mistakes that lead to bad data, and how to prevent each one. Perth, Melbourne, Sydney and Brisbane.
The five most common integration mistakes that lead to bad data, and how to prevent each one. Perth, Melbourne, Sydney and Brisbane.
Who this is for
IT leaders and operations managers who've experienced data quality problems from system integrations or want to avoid them.
Question this answers
What are the most common integration mistakes that cause data problems, and how do we prevent them?
What you will leave with
Integrations are the most common source of data quality issues in business systems. Not because the technology is unreliable, but because the planning is often incomplete. Data flows between systems silently, and problems accumulate before anyone notices.
Here are the five mistakes we see most often, and how to prevent each one.
What happens: The same customer, order, or transaction appears multiple times in the target system because the integration creates new records instead of updating existing ones.
Why it happens: The integration does not properly match incoming records against existing ones. The matching logic relies on fields that are not unique (like name) instead of unique identifiers (like email or account number).
How to prevent it:
What happens: The same data (e.g., customer address) exists in multiple systems and conflicts. Nobody knows which version is correct.
Why it happens: Both systems allow editing the same data, and there is no rule for which one "wins" when they disagree.
How to prevent it:
What happens: A record fails to sync (API error, validation failure, timeout) and nobody notices. The data gap grows silently over days or weeks.
Why it happens: The integration was built to handle the happy path only. Errors are logged somewhere but not monitored or acted on.
How to prevent it:
What happens: Data arrives in the wrong format. Dates break (DD/MM/YYYY vs MM/DD/YYYY), phone numbers lose their leading zero, currency amounts lose decimal places.
Why it happens: Data mapping did not account for format differences between systems. Testing used clean sample data that did not reveal edge cases.
How to prevent it:
What happens: The integration works fine for months, then gradually degrades. Volume changes, API updates, data format shifts. Problems accumulate without detection.
Why it happens: The integration was built and forgotten. There's no dashboard, no health checks, no regular review.
How to prevent it:
The five most common integration mistakes are duplicate records, no source of truth, no error handling, data format mismatches and no monitoring. Integrations cause data problems less because the technology is unreliable and more because the planning is incomplete. Data flows between systems silently, so problems build up before anyone notices. Most of these problems are preventable with proper planning.
An integration creates duplicate records when it makes new records instead of updating existing ones. This usually happens because matching relies on fields that are not unique, such as a name, rather than unique identifiers such as an email or account number. Prevent it by defining a unique identifier before the build, using upsert logic that updates a record if it exists and creates it if new, and adding duplicate detection rules.
A source of truth is the one system designated as the master for each shared data element, such as customer addresses. Without one, both systems allow editing the same data and nobody knows which version is correct when they disagree. Make the data read-only in the other systems, or sync changes back to the master, and document where to edit what. Having no clear source of truth is the number one cause of integration data issues.
Integration error handling should be designed as part of the build, not added afterwards, because every integration will fail at some point. Failed records should go to a dead-letter queue, alerts should notify the right person when failures occur, and transient errors should be retried with exponential backoff. Without this, a record that fails through a validation failure or timeout goes unnoticed and the data gap grows silently for days or weeks.
Data format mismatches are prevented with explicit transformation rules, because formats should never be assumed to match. Document the format requirements for every field in both systems, test with anonymised production data that includes edge cases, and validate data at the point of ingestion so bad data is rejected. Typical breaks are DD/MM/YYYY dates read as MM/DD/YYYY, phone numbers losing their leading zero and amounts losing decimal places.
Ask the author
Ask it here and it comes straight to the founder. No sales call, no obligation, and a real answer even if the answer is that you do not need us.
Kasun Wijayamanna
Founder, replies within one business day
Tell us what you are comparing, replacing, or trying to improve. We will come back with a practical recommendation and realistic scope.
Built here. Your data stays here.
Thanks for reaching out. We will get back to you within one business day.
See what else we do