Skip to main content
Waaru
WhatsApp for D2C ecommerce

Handle return requests on WhatsApp with clear approval

Tanay PatelAI and full-stack engineer
The short answer

A WhatsApp return-request workflow should collect the order reference, item and reason, then route the request to someone who can apply your store policy. Make receipt, approval and refund status distinct. In the reviewed Waaru Shopify connector, return creation and refund issuance are not supported actions.

What can the first reply reasonably promise?

Tell the customer that you can help them submit a request. Do not promise eligibility before checking the purchase and the applicable policy. A calm acknowledgement gives them a next step without committing the store to a decision it has not made.

Ask for the relevant order reference and item. If several items are involved, make that explicit rather than assuming the entire order should be returned. Then ask for the reason in the customer's own words or through clear choices with an option to explain.

Avoid treating every request as a product defect. Wrong size, damage, a missing item and a changed preference may need different checks. The first flow should identify the job, not decide every exception.

How should you verify the order and policy?

Use your store's customer verification process before disclosing order details. The order reference should help locate a record, but should not be the sole basis for sharing someone else's purchase information.

Check the purchased item and the policy that applies to it. Decide which information the workflow can safely read and which decisions remain with support. Do not fill gaps in the policy with an AI-generated answer.

Shopify distinguishes return and exchange management from other order tasks. Provider support for a process does not mean every connector exposes it. Use the Shopify order-tracking workflow for read-only status checks, and keep return approval as a separate step.

What does an honest request acknowledgement look like?

Example reply, not a policy promise: “I have your request to return the shirt from this order because the size is not right. Support will check the item and your order against the return policy. Please wait for the return instructions before sending it back.”

The internal summary should include the requested item, the stated reason, the verification status and anything still missing. If a photo is genuinely needed, explain what it should show and avoid collecting unrelated personal information.

Do not label this conversation “refund completed.” The request may not yet be approved. A customer should be able to understand whether they are waiting for a decision, shipping instructions, inspection or a payment update.

How should you build the flow in Waaru?

Start with Logic Flow collection and routing. Add a path for missing information and one for a person who needs staff immediately. Keep the initial request available in the inbox even if a later integration step fails.

Connect the intended store through business integrations. Review the action catalogue rather than assuming that any Shopify task can run. The reviewed connector supports order reads and optional tagging, not return creation or refund issuance.

Where an available tag action helps support find a pending request, treat the tag as a queue marker only. It is not approval or proof of a refund. The authorised store operator must complete those actions through the appropriate system.

Who handles exceptions and uncertain outcomes?

Name a person or team that can decide policy exceptions. A damaged item, a disputed delivery or a request outside the ordinary process should not bounce between generic menus. Preserve the customer's explanation for that reviewer.

If a write times out, inspect the source record before repeating it. This matters for any connected action you use, even a queue tag. Do not build an uncontrolled retry loop around an unknown result.

Tell the customer when more checking is required. If support performs a refund in the store's system, verify that record before reporting its status. Do not invent bank processing times or claim that a payment has reached the customer just because a refund request was submitted.

What should the team test before release?

Use this implementation checklist with support:

  • Verify the store connection and customer-match process.
  • Test one item, several items and an incorrect order reference.
  • Confirm where the current return policy is maintained.
  • Route missing evidence and policy exceptions to a named owner.
  • Check that receipt, approval and refund wording stay distinct.
  • Test a failed lookup and a guest asking for a person.
  • Review late replies against the messaging restrictions.

Keep a record of the questions support still needs to ask after handoff. If the same essential detail is repeatedly missing, improve the intake question. If the question is rarely useful, remove it. The D2C workflow page helps connect return handling with order enquiries and human support.

FAQ

Frequently asked questions.

Can Waaru issue Shopify refunds through this workflow?

Not through the reviewed Shopify connector. Return creation and refund issuance are not supported actions. An authorised operator must use the appropriate store process.

Should the chatbot automatically approve eligible returns?

Only a separately verified, explicitly authorised decision process could do that. This guide covers collecting a request and handing approval to support, not automated approval.

Is a return-request tag proof that the return is accepted?

No. A tag can help organise work. It does not confirm policy approval, receipt of the returned item or a refund.

Research notes

Evidence checked for this article.

  • Shopify documents return and exchange management as a distinct order process.

    Shopify · Accessed 2026-09-16

About the author

Tanay Patel

AI and full-stack engineer

Tanay builds production AI systems, agent workflows, and the full-stack software around them. His work focuses on reliable state, recovery, and human handoff.

Read Tanay's engineering notes

Put this into practice on your own number.

The platform is free. AI features are paid separately.