Back to all articlesAPI integration planning

Assign field-level authority when both systems can edit the same data

Alactic Editorial11 min read

Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.

A salesperson changes a service date in the CRM while a coordinator changes the same date in the operations system. Both edits are valid inputs. Without an authority rule, the integration has no agreed basis for deciding which date should become the shared value.

Assign authority to each field, qualified by workflow state. Specify which system may publish the accepted value, what the other system may submit as a proposal, and when disagreement must stop propagation for review. Also define a fallback for unavailable sources: an outage should not silently transfer ownership.

The practical artifact is a field-authority matrix. It should distinguish source-owned, destination-owned, derived, and jointly reviewed fields—not simply label an entire application the source of truth.

1. Separate permission to edit from authority to publish

An editable screen tells you where someone can enter a value. It does not establish whether that value should overwrite another system. Treat local editability and shared authority as separate decisions.

For integration planning, give a value one of three statuses:

  • Accepted: the value approved for shared use under the current authority rule.
  • Proposed: an edit awaiting validation or acceptance by the authority holder.
  • Disputed: competing values that cannot be settled by the existing rule.

These are recommended design concepts, not assumptions about controls already present in either application. A proposal might live in a supported field, a separate change record, or a managed review queue, depending on the available implementation.

A useful rule reads: “Once a booking is confirmed, operations owns the scheduled service date. CRM edits become change requests and do not replace the accepted date.” A weak rule reads: “Operations is usually correct.” The first defines a decision; the second leaves the integration team to guess.

Do not use the newest timestamp as a substitute for authority. A recent edit may have been made from an old screen, imported from stale data, or generated by a retry. Recency can help establish sequence, but it does not explain whose decision should prevail.

2. Establish the matrix conventions before filling it

Fictional worked example: A service business connects a CRM with an operations scheduling system. Both allow staff to edit customer-facing details. All field names, values, people, and workflow states below are hypothetical; this is a proposed operating design, not an Alactic customer result or product capability claim.

For this example, the CRM is called the source and operations the destination. These labels identify the systems consistently; they do not imply that every field flows from source to destination.

  • Source-owned: the CRM publishes the accepted value. Destination edits are proposals or rejected changes, according to the row.
  • Destination-owned: operations publishes the accepted value. The CRM receives it without gaining authority over it.
  • Derived: a named calculation produces the value from accepted inputs. Neither application’s manual entry wins.
  • Jointly reviewed: no unilateral edit becomes accepted. A defined review process must approve a common value.

Each row also needs a workflow state, conflict example, exception route, and outage behavior. Without those details, a category such as “source-owned” is too broad to guide a failed update.

The matrix below uses a conservative default: keep the last accepted value visible with its freshness status when its authority is unavailable, but do not imply that stale data remains safe for every action.

3. Filled field-authority matrix

The following rows form one matrix, presented as field cards so the decision rules remain readable. They can be copied into a shared document tomorrow.

Customer display name — source-owned

  • State and authority: CRM owns the display name throughout the booking workflow.
  • Accepted path: validated CRM changes publish to operations.
  • Conflict example: CRM contains “Harbor Studio”; operations is edited to “Harbor Studio North.” Keep the CRM value accepted and record the operations edit as a correction proposal.
  • Review exception: if the proposal indicates that the wrong business record is linked, pause updates for that record and route the linking issue separately. Do not attempt to solve identity problems by changing the name.
  • Unavailable authority: retain the last accepted name, mark it stale, and hold correction proposals until the CRM can be checked.

Requested service date — source-owned before confirmation

  • State and authority: CRM owns the requested date while the booking is in draft. This is a customer preference, not a confirmed appointment.
  • Accepted path: CRM changes update the request visible to operations.
  • Conflict example: CRM requests 14 June; operations enters 16 June in the request field. Keep 14 June as the accepted request. Operations can propose 16 June as an alternative.
  • Review exception: a request arriving after confirmation must enter the rescheduling workflow rather than overwrite the confirmed schedule.
  • Unavailable authority: preserve the last accepted request, but hold confirmation if the request’s current status cannot be established.

Scheduled service date — destination-owned after confirmation

  • State and authority: operations owns the confirmed scheduled date.
  • Accepted path: operations publishes the date and booking revision to the CRM.
  • Conflict example: operations confirms 16 June; a CRM user enters 14 June as the scheduled date. Preserve 16 June and create a rescheduling proposal.
  • Review exception: if the CRM indicates cancellation while operations is confirming a date, hold further scheduling changes until the booking state is resolved.
  • Unavailable authority: show the last confirmed date as unverified. Under this hypothetical policy, do not send a new confirmation based on an unverified change.

Service location — authority changes at confirmation

  • State and authority: CRM owns the location in draft; operations owns it after confirmation.
  • Accepted path: the confirmation transition records the accepted location and transfers authority at a specific booking revision.
  • Conflict example: a draft CRM edit arrives after confirmation. Do not let its arrival time override the handover. Treat it as a location-change proposal.
  • Review exception: a material location change after confirmation requires the scheduling coordinator to check whether the existing booking can still be fulfilled.
  • Unavailable authority: keep the last accepted location. Hold dispatch instructions that depend on an unresolved location change.

Estimated duration — derived

  • State and authority: a designated calculation owns the estimate in all active states.
  • Accepted path: calculate from accepted service type and quantity, recording the calculation version and input revisions.
  • Conflict example: the calculation produces a hypothetical 90-minute estimate; someone types 60 minutes in the CRM. Do not publish the manual value as the derived estimate.
  • Review exception: if a specialist estimate is needed, store an explicitly approved override separately, with its reason and applicable booking revision.
  • Unavailable authority: retain the previous estimate as stale. If accepted inputs changed, hold planning decisions that require a current estimate.

Customer-facing service scope — jointly reviewed

  • State and authority: after confirmation, sales and operations jointly review material scope changes.
  • Accepted path: a named booking coordinator publishes the exact revision both reviewers approved.
  • Conflict example: CRM adds an extra room while operations records that it is excluded. Neither edit overwrites the other as the accepted scope.
  • Review exception: missing approval, contradictory instructions, or an altered proposal after approval keeps the change disputed. A revised proposal needs fresh acceptance.
  • Unavailable authority: retain the last agreed scope and mark the proposed change unresolved. If the pending work depends on it, escalate before execution.

4. Make ownership changes explicit workflow events

The service-location row is the hardest case because its owner changes. Define a handover event rather than relying on a label that each system may update independently.

In the hypothetical workflow, confirmation should establish a booking revision, the accepted location at that revision, and the new authority holder. If the two applications cannot complete that transition together, use an explicit pending-handover state and reconcile it before accepting changes under the new rule.

Consider a CRM location edit made in draft but delivered after confirmation. The integration needs to know which state and revision the edit was based on. “Received after confirmation” is not the same as “authorised after confirmation.”

Record enough context to distinguish them:

  • Originating system and field.
  • Proposed value and the accepted revision it was based on.
  • Workflow state when the proposal was made.
  • Change identifier and processing status.
  • Authority-rule version used for the decision.

If the systems cannot supply reliable revision information, document that limitation. A reasonable fallback is to route ambiguous handover edits to review, not invent a sequence from uncertain timestamps.

5. Define exceptions without creating a second source of truth

An exception should have a narrow scope, a responsible reviewer, and an exit condition. “A manager can fix it” is insufficient because the fix may be overwritten on the next synchronization.

For each review item, preserve the accepted value, competing proposals, their origins, relevant workflow state, and the reason propagation stopped. Assign a named role to decide, plus a backup route if that role is unavailable.

For jointly reviewed fields, separate approval from publication. In the fictional scope example, sales and operations approve the same proposed revision; the booking coordinator records the decision and publishes it. This avoids two reviewers independently writing different versions of their agreement.

Choose one of three explicit outcomes:

  1. Accept a proposal: publish it through the authoritative path and close the competing proposals.
  2. Keep the existing value: reject the proposals with a reason so they are not repeatedly retried.
  3. Change the ownership rule: approve a new rule version and assess affected records before applying it.

A one-off override should not silently change permanent ownership. Give it an applicable record and revision, a reason, and an expiry or completion condition. Keep the original rule visible.

6. Plan for unavailable and disputed authorities separately

Unavailable means the designated authority cannot currently be checked or updated. Disputed means there is no accepted decision about which value or authority should apply. Retrying may help the first condition; it does not settle the second.

Choose an outage response by consequence. A stale display name may be acceptable for internal viewing. An unresolved service location may make dispatch inappropriate. Avoid a single fallback that either blocks everything or treats every cached value as current.

A practical recovery sequence is:

  1. Mark affected values as stale, pending, or disputed; do not relabel them accepted merely because a write was attempted.
  2. Retain proposals durably with their original identifiers and base revisions.
  3. When connectivity returns, read the authority’s current value and workflow state.
  4. Re-evaluate queued proposals against that state. Do not replay obsolete edits blindly.
  5. Publish eligible changes through the authoritative path, then verify the resulting value.
  6. Send unresolved conflicts to review with the original context intact.

If the authoritative system is retired or unavailable for an extended period, ownership transfer should be a deliberate migration decision. Name the replacement authority, choose a cutover point, reconcile outstanding changes, and prevent the old system from resuming publication under an obsolete rule.

7. Test disagreements, not just successful synchronization

A successful update proves that data can move. It does not prove that the correct party controls the field.

Before rollout, exercise the matrix with fictional records and retain the observed outcome for each test:

  • Competing edits: edit the same field in both systems. Verify that the accepted value follows authority, not arrival order.
  • State transition: delay a draft edit until after confirmation. Verify that the handover rule still applies.
  • Outage and replay: queue a proposal, change the authoritative record, then restore connectivity. Verify that recovery checks the new revision.
  • Blank value: distinguish an intentional clearing from an omitted field. Define whether clearing is permitted for that field.
  • Repeated update: process the same change again. Verify that it does not create repeated review items or new business actions.
  • Derived-value correction: change an accepted input. Verify that the estimate is recalculated or marked stale rather than left apparently current.

Also check the human outcome: can a coordinator see what is accepted, what is pending, and what action is blocked? A technically correct rejection is still operationally weak if nobody can understand or resolve it.

8. Start with a bounded ownership review

For a first review, choose a hypothetical pilot of ten shared fields involved in one workflow. Assign each a category, state rule, conflict example, review owner, and unavailable-source response. Separate fields that only look equivalent, such as requested date and scheduled date, before deciding who owns them.

Measure completeness and behavior rather than claiming success from update volume. Track how many pilot fields have a complete rule, how many test conflicts produce the expected accepted value, how many review items lack an owner, and how many recovered changes bypassed revision checks.

Set the release condition before testing: every pilot field has a documented authority rule, every review exception has an accountable role, and every tested outage preserves the distinction between accepted and pending values. Where the integration cannot enforce a rule, record the gap and restrict the affected workflow rather than implying the conflict is resolved.

API integrationField ownershipData governanceConflict handling

Put the next conversation to work.

Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.

Explore Communications