System Integration Planning: APIs, Data and Ownership

1 つの交換をエンドツーエンドで実行します。

Web サイトから内部アプリケーションに移動する注文など、代表的な取引所を選択します。送信システム、受信システム、フィールド、トリガーをリストします。どのシステムが各値を所有するかを決定します。 2 つのシステムが同じレコードを更新できる場合は、インターフェイスを構築する前に、競合する更新がどのように処理されるかを文書化します。

遅延メッセージや繰り返しメッセージを設計します。

統合は、同じリクエストを 2 回受信するか、受信側が作業を完了した後に応答を失う可能性があります。識別子、再試行、調整について実装チームと話し合ってください。保留中の交換または失敗した交換をユーザーがどのように確認するか、および重複したビジネス レコードを作成せずに誰が安全に交換を修正できるかを定義します。

インターフェースをレビュー可能にします。

受け入れられた入力と拒否された入力、予期される応答、およびアクセス要件の例を記録します。一方のシステムが他のシステムを黙って壊すことがないように、変更プロセスを組み込みます。運用可能な統合を検討する前に、両方の所有者との契約を確認し、代表者の障害をテストしてください。

Walk through the life of an exchanged record

Choose one record, such as an order, and follow its creation, update, cancellation and correction. At every boundary, identify the sending system, receiving system and source of truth. Confirm the actual interface available rather than assuming that an application exposes the data you need.

Review authentication, field meanings, identifiers and expected timing with both owners. Test a duplicate message, a delayed update and an unavailable destination. Decide how to distinguish a rejected request from an uncertain outcome. Arrange reconciliation so the operating team can find records that disagree after an interruption. These decisions should be captured before implementation begins.

An interface contract planning sheet

This is a discussion template, not a substitute for the technical contract. Attach the relevant API specification or file format after the owners agree the behaviour.

• Purpose and direction: why the exchange occurs and which system initiates it.
• Record identity: stable identifiers, field definitions and ownership of each value.
• Trigger and timing: event, schedule, expected delay and ordering assumptions.
• Access and limits: authentication, permissions and constraints confirmed by the provider.
• Failure handling: timeouts, duplicates, retries, rejection reasons and human escalation.
• Verification: test examples, reconciliation checks, versioning and named support owners.

Do not equate a response with a completed business action

A successful technical response may acknowledge receipt without confirming that a downstream process finished. Agree what each status means and how completion is checked. Conversely, a timeout may leave an uncertain result, so retrying a consequential action without checking can create duplication.

Avoid relying on field names alone. Two systems may use the same label for different concepts. Keep mappings and interface changes visible to both teams, including who approves a breaking change.

Should an integration be real time?

Choose timing around the business need and the capabilities of the connected systems. A scheduled exchange may be sufficient for some tasks; others need an event-driven flow. Document the acceptable delay and recovery approach before choosing the implementation pattern.

ILLUSTRATIVE PLANNING EXAMPLE

Example: connecting orders and inventory

An order system sends confirmed purchases to inventory. Decide which system owns stock counts, when a reservation occurs and how cancellations reverse it. Ask what happens when the same message arrives twice or inventory is temporarily unavailable. Assign an owner to each interface and agree representative integration tests.

Create an interface sheet covering fields, triggers, permissions, duplicate handling and recovery. Confirm assumptions with both system owners before estimating the connection.

FURTHER READING

Official resources to explore

Explore the related official guidance below. The templates and examples are practical starting points to adapt to your own project.

• MDN: overview of HTTP