A workflow brief often describes the intended exchange: a request is submitted in one application and a corresponding item appears in another. The less obvious decisions begin when the second application does not confirm the update. What should the customer see? Is the work safe to repeat? Who can establish what actually happened?
Before agreeing the first release, give unresolved exchanges a visible state, an evidence trail and a named owner. The following checklist is a planning aid for a scoped integration discussion. It is not a claim that a particular third-party service, account or permission is available.
1. Describe one exchange, including its boundary
Use an illustrative example: an approved service request should create a work item in an operational tool. This is a planning example, not an NG Labs client result or an implemented integration.
Name the source record, the destination record and the event that starts the exchange. Identify which system owns each field. If the portal owns the customer's request but the operational tool owns its assigned status, write that division into the brief. Decide which changes may travel back and which remain read-only.
Also name the access owner. Before promising the connection, confirm who can approve the required account access and whether the intended interface and scope are available. Keep credentials out of the initial brief.
2. Separate an incomplete exchange from a failed business decision
A rejected request and an unconfirmed update are different product states. In the example, an approver may deliberately reject the service request. Separately, the request may be approved while creation of the work item remains unresolved.
Write the state names in language that the team and customer can understand. A useful starting map is: ready to send, awaiting confirmation, confirmed, needs investigation and closed by an authorised operator. Adapt it to the actual workflow rather than treating those names as a required architecture.
For every state, specify what a person sees and what they may do next. Avoid displaying a completed handover before the agreed evidence of completion exists.
3. Define a repeated action before adding a retry control
Ask what counts as the same intended exchange. Does a second click refer to the original request version, or to a changed request that requires another approval? What should happen when confirmation arrives after someone has begun investigating?
Make those questions part of the acceptance examples. Agree what evidence an operator checks before repeating an action, who may authorise it, and how the record will show that decision. A retry button alone does not answer those product questions.
4. Give investigation enough context without exposing private records
List the information needed to connect an unresolved exchange with its source record and destination response. For example, agree a reference, relevant version, state, event time and concise outcome category. Decide who may see that context and what should be excluded from routine customer-facing messages.
Define the customer message separately from the operator view. A customer may need to know that their request is retained and an update is being checked. The operator may need more context to decide the next action. The actual logging, retention and access design belongs in the implementation scope and review.
5. Assign recovery and closure
Name the first person or role responsible for unresolved work, the person allowed to repeat or correct it, and the evidence needed to close the item. Include access loss, a changed destination field, an outdated source version and an unavailable service in the review discussion.
Agree the operating boundary explicitly. A project contact address does not establish an around-the-clock monitoring commitment. Record the support scope, decision maker and escalation path that are actually agreed for this product.
A recovery decision table for the brief
Use the table below to expose decisions, not to prescribe a particular integration design. It extends the illustrative service-request example above; it describes no live customer system.
| Observed condition | Decision to agree | Responsible role to name | Evidence before closure |
|---|---|---|---|
| No confirmation is available | Keep the exchange unresolved; agree what an operator checks before any repeat | Investigation owner | The identified source version and the checked destination outcome |
| The person repeats the action | Decide whether this is the same intended exchange or an authorised new version | Authorised operator | The recorded version and the agreed repeat decision |
| Confirmation arrives after investigation starts | Decide how it is matched to the unresolved item and who reconciles the result | Reconciliation owner | A matching reference and a recorded resolution |
| A role or destination permission changes | Agree who can restore access and whether the exchange remains pending | Account/access owner | Approved access and the outcome of the authorised recovery |
| The source record has changed | Decide whether a new review is needed before another exchange | Business decision maker | The accepted current version and its approval |
For each row, add what the customer sees, who may perform the next action and which person can establish completion. Review the table with the people responsible for both systems. Mark a deliberately manual recovery step as manual; do not imply that the application already automates it.
Before retrying, agree what each field means
A connection can accept a message while the resulting record is still wrong for the business. Before adding a repeat control, agree a field contract: the meaning, ownership and permitted representation of the values that affect the work. Keep these proposed rules distinct from the evidence of an actual check.
For an illustrative print request, a portal owns the approved quantity and finish while an operational tool owns the assigned work status. This is a fictional scope example, not an implemented connection. A missing quantity must not silently become one printed item simply because a destination has a default. The teams need to decide whether the exchange is refused, remains unresolved or follows an explicitly agreed alternative.
| Decision | What to record | Illustrative question |
|---|---|---|
| Meaning and authority | Business meaning, source of truth and the role allowed to change the value. | Does quantity mean printed items, packs or another unit? |
| Representation | Source/destination names, allowed values, units and any transformation. | Which source finish code maps to which supported destination finish? |
| Omitted, empty or null | A separate outcome for each input the interface can represent. | Does an omitted note leave the old value unchanged, while an empty one clears it? |
| Dates and versions | Calendar date or timestamp, applicable timezone and the approved source version. | Is the requested date a business calendar date or a specific instant? |
| Unsupported value | Visible unresolved state, decision owner and evidence before a deliberate repeat. | Who reviews a finish code added after the original mapping was approved? |
| Useful completion | What must be found in the destination and who may close the exchange. | Can the reviewer identify the right request version and expected business values? |
The example questions are decisions to resolve, not universal API behaviour. If an interface cannot express a required distinction, record that limitation and agree a bounded manual step or a scope change. Do not describe a value as transferred correctly merely because a request was accepted.
Review three controlled cases
- Expected values: use a synthetic approved request with an agreed quantity and supported finish. Write the expected destination values first, then record what the reviewer actually found.
- Missing versus empty: compare an omitted note with an explicitly empty note in separate authorised test records. Check each against its agreed meaning; leave the observation blank until the check has happened.
- Changed or unmapped value: specify what happens when a new finish has no approved mapping. Name the unresolved message and responsible owner. Agree how a mapping change and the relevant request version will be reviewed before another attempt.
Use an authorised preview or test environment and synthetic values. This planning exercise does not authorise writes to a live customer system or prove that access to a third-party account is available. It also does not implement validation, permissions or duplicate prevention.
Review row: source reference/version → contract revision → expected destination outcome → observed evidence → unresolved difference → decision owner → next action. Keep credentials and private customer data out of the brief. Agree any handling, retention and access requirements with the people responsible for the actual systems.
The downloadable recovery brief below now includes a field-contract record. For an editable Markdown version, use the existing NG Labs integration recovery template. These are planning aids, not monitoring, support or security commitments.
6. Demonstrate the awkward journey
Review these illustrative acceptance examples alongside the successful exchange:
- The source request remains identifiable when the destination does not confirm creation.
- The customer sees the agreed unresolved state and next step.
- A repeated action is reviewed against the same intended request and version.
- A late confirmation is handled according to an agreed rule.
- A person without the required role cannot repeat or close the exchange.
- The responsible operator can explain the recorded recovery decision.
These examples are requirements to agree and test, not a security certification or a promise of a particular implementation.
Bring the recovery decisions to the scope
Use the blank recovery brief below to record the exchange, ownership, repeated-action decisions and closure evidence in one place.
Bring this record to a web application or client portal scoping conversation with NG Labs. Start with one exchange that matters to the first useful release. Expand the integration after its ownership, access and recovery decisions are reviewable.
NG Labs is a web application and digital product brand of North Growth Lab LLC.
Use the recovery brief
Download the blank plain-text brief. Complete it without credentials or private customer records.
NG LABS — INTEGRATION RECOVERY BRIEF Planning template. No integration, monitoring or support commitment is implied. Exchange / business purpose: Source record, reference and authoritative fields: Destination record and authoritative fields: Trigger and relevant version: Account and access owner for each system: Evidence of confirmed completion: Unresolved state shown to the customer: Repeated-action rule and authorised operator: Late-confirmation reconciliation rule: Changed-source-version review rule: Investigation context, access boundary and exclusions: Recovery owner and evidence before closure: Explicitly manual recovery steps: Agreed support hours, scope and escalation: Acceptance examples and reviewer: NG Labs is a brand of North Growth Lab LLC. Planning template from https://nglabs.cc/.