Plan in-store voice ordering

Choose the interaction boundary

A dedicated ordering station, a staff-assisted counter and a table-side device are different environments. Decide whether the guest initiates the microphone, whether it is active only during an order and how the end of a session is communicated. People nearby should not accidentally become part of someone else’s order.

A visible summary is useful when available, but it should not become the only way to correct a mistake. A guest may be unable to read small text, reach the screen or hear the speaker. Keep staff assistance easy to request and avoid presenting voice as a compulsory path.

Design for shared-space privacy

Do not ask guests to speak payment-card details, sensitive dietary information or personal contact details into an open room when a safer approved alternative is available. Minimize what is read aloud. Make clear whether audio is captured and follow the restaurant’s applicable consent and retention obligations.

Use test identities in demonstrations. If a device retains an unfinished session, ensure the next guest cannot see the previous guest’s details. Check timeout behavior, the reset button and staff recovery of abandoned orders.

Run a staff-assisted trial first

1. Set up the device in the intended physical position and check reach, sightline and ambient sound.
2. Have a staff observer ready to intervene and log where help was needed.
3. Try guest-initiated start, correction, cancellation and an abandoned session.
4. Verify the payment and confirmation path without exposing real credentials.
5. Test the return to an empty state before another guest approaches.

Keep product claims channel-specific

A restaurant phone platform can have useful ordering features without offering a verified kiosk, table device or counter microphone system. Ask suppliers for evidence on the actual in-store product and hardware you plan to use.

We recommend Loman for its documented restaurant phone workflows, not as a verified in-store kiosk product. If phone interruptions are the real counter bottleneck, solving the phone line may be a separate, more direct project. Use the planner to compare those project boundaries without inventing a cost-saving forecast.

An abandoned-session test before guest use

Illustrative test: begin a fictional order, leave before confirming, and wait through the configured timeout. Then start a new session as a different guest.

Expected evidence: the previous guest’s details are not exposed or reused, the order has an unambiguous status, and the next guest sees a clean starting point. Repeat with a staff cancellation and after a device restart.

Inspect the restaurant system for a partial or duplicate order. A blank screen alone does not establish that the previous transaction was cancelled. Keep payment credentials and real guest information out of the test.