Connecting Business Tools: How to Decide Which System Owns Each Record

c
ceofounder
17 min read 23 views
Connecting Business Tools: How to Decide Which System Owns Each Record
System Integration

Connecting Business Tools: How to Decide Which System Owns Each Record

Many businesses use several tools because each solves a particular problem. The trouble begins when the customer name in one system differs from the invoice record in another, or staff copy the same update into three places.

Integrations can reduce this work, but connecting tools without deciding which system owns each record can spread errors faster.

Draw the current information flow

List each tool and the records it holds. For each flow, write what triggers the update, where the information begins and which other system needs it.

  • Customers
  • Products
  • Appointments
  • Orders
  • Invoices
  • Payments
  • Documents
Include manual transfers and exports. They are part of the current architecture even if no API is involved.

Assign a source of truth by record type

The system that creates a record is not always the one that should own every field. Decide where each record is maintained and write down how corrections flow back to dependent systems.

Source of truth

Ownership by record type

Record Owner Needs copy
Customer details CRM Others read
Appointment times Booking system Calendar reads
Product availability Inventory Shop reads
Official invoice Accounting Reports read
Payment status Payment provider Finance confirms

Avoid allowing two connected tools to overwrite each other without clear rules.

Define matching and error handling

Systems need a dependable way to match the same customer, product or order. A reliable integration should make failure visible and give staff a way to resolve it, rather than silently losing information.

  • MatchWhich identifier is used
  • MissingWhat happens when data is incomplete
  • DuplicateHow duplicates are detected
  • RetryHow repeats avoid creating copies
Plan for timeouts, delayed updates, rejected records and repeated notifications. A silent failure is worse than a visible error.

Protect access and minimise data

Only transfer fields required for the receiving task. Confirm who can access credentials, where integration secrets are stored and how failures are logged.

A connection between systems can expand the impact of an account compromise, so treat integration permissions and monitoring as part of the design.

Test before trusting it

Use representative test cases and reconcile the records after the test. Confirm that retries do not create duplicate transactions.

Test scenarios

Cases that reveal whether the integration is safe

New record
Update
Duplicate
Missing field
Failed connection
Cancellation
Correction
Retry after timeout

Agree who monitors the connection after launch and what support route exists if one system changes its API.

Ask whether an integration is worth building

An integration is justified when the repeated work, error risk or delay is greater than the cost and maintenance burden of connecting the systems.

Manual or import
  • Low frequency of transfer
  • Simple field set
  • Tolerance for short delay
  • Few systems involved
Real time integration
  • High frequency of transfer
  • Complex field mapping
  • Delay causes real cost
  • Several connected systems
Compare the frequency and impact of the task, not just whether the two products advertise APIs.

Expect changes at either end

Third party tools can change authentication, fields, rate limits or endpoint behaviour. Assign an owner to review vendor notices and plan tests when a connected system is updated.

Avoid assuming an integration is a one time expense with no future maintenance. Document what business activity depends on it and how the team should operate during an outage.

Define a data mapping and correction plan

For each field, decide the source, destination, format, requiredness and treatment of missing values. A clear mapping prevents the same data from being interpreted differently by each application.

Data mapping

Decide per field, before building

Source system
Destination system
Format and standard
Required or optional
Missing value rule
Who may correct

Standardise phone numbers, dates and identifiers where relevant, and define how a correction reaches dependent records.

Monitor business outcomes as well as technical status

A connection may return a successful response while still transferring the wrong business information. Review sampled records, duplicate counts and the time saved from manual work.

Make it clear who reviews failed transactions and who can reprocess them safely without creating duplicates.

Plan an integration before selecting its technical method

1

List the records exchanged between systems and identify which system owns each field. Map triggers, identifiers, required fields, formats, expected delay and the business consequence if the update fails.

2

Compare options: an API, a controlled import or export, or a manual step with clear ownership. Test duplicate messages, delayed responses, missing fields, corrections and retries. Confirm credentials are stored securely and that only necessary information is shared.

3

Assign operational and technical owners for monitoring, vendor changes and failure resolution. Reconcile a sample of records after launch and measure whether the connection reduces errors or repeated data entry. Keep an outage procedure so the business knows what to do if the integration stops working.

Plan for monitoring after the original developer moves on

A future maintainer should be able to identify what business process is affected by a failure without reverse engineering every request.

Runbook

What the documentation should record

  • The purpose of the integration
  • Connected systems and data fields
  • Known limitations
  • Who owns the credentials
  • The escalation route

Keep secrets out of general project notes and store them in an appropriate secure location. Revisit the runbook when either source system changes. Integration continuity should not depend on one developer remembering undocumented decisions.

Example

A booking system may own appointment times while an accounting tool owns official invoices.

Booking completed → Billing details passed → Correction routes back to source

A later correction must have a clear route back to the original record rather than creating a second, conflicting appointment.

Frequently asked questions

Do two systems always need an API integration?

No. Compare the cost and risk of manual exchange with the frequency and importance of the data flow.

Who maintains an integration?

Name an operational owner for monitoring and a technical owner or supplier for failures and changes.

Action checklist

  • Current information flows are mapped.
  • Each record type has a defined source of truth.
  • Matching rules and failure handling are specified.
  • Only necessary data is transferred.
  • Integration tests include failures and duplicate attempts.

Final thought

Successful integration begins with information ownership and failure handling, not a list of APIs. Document the current process, decide which record is authoritative and test exceptional cases before relying on automation.

Assessing a workflow or integration?

Stonehenge Digital KE's Business Management System service can help assess where a tailored workflow or integration may be justified.

Explore the Business Management System Service →
Share Written by ceofounder

Interested in Business Management System?

Let's build this for your business

Replace spreadsheets, paper records, and WhatsApp chaos with a custom system built for exactly how your business works.

Leave a Comment

Share Your Thoughts