Integration questions to answer before an estimate

Define the trigger and outcome

Connect the CRM is too broad. A useful description is: after an order is confirmed, create a deal, transfer its items and assign a manager. For each connection, describe the trigger, the data and the result staff will see. Decide separately whether changes must flow back to the original system.

Confirm access before promising a date

Documentation does not prove that an operation is available to your account. Check the plan, permissions, test environment, limits and the provider’s support contact. Share credentials securely; the brief only needs to identify availability and ownership. Approval by an external provider can become the main scheduling dependency even when development is straightforward.

Decide which system owns each value

If a customer’s phone number changes both on the website and in the CRM, the integration needs a conflict rule. Assign a primary source for each data type and define how disagreements are resolved. Not everything needs two-way synchronisation. Sending orders to an operations system and receiving confirmed statuses may be sufficient.

Plan for duplicates and delays

A notification can arrive twice, and a provider can respond late. Reprocessing must not create a second order or perform the same action again. Plan a stable operation identifier, visible pending states and retry rules. The user should understand that a request was received even while the external result is still pending.

Make failures visible to an owner

A technical log does not help the manager waiting for a shipping document. Decide who receives an alert, where the status is visible and how an action can be retried safely. Critical operations need a manual resolution path. Exceptions deserve a workflow rather than disappearing into the gap between two systems.

Write down acceptance scenarios

Include normal exchange, an unavailable provider, duplicate events, invalid data and recovery after failure. Describe the expected outcome and who will approve it. Separate implementation costs from provider fees and ongoing connection support. The scope then becomes testable, while operating costs and responsibility for future changes remain clear.