Compare a kiosk pilot with staffed ordering by measuring the same things on similar shifts: completed orders, guest queue time, staff assistance and abandoned attempts. Set your decision thresholds before the pilot starts. A busy screen alone does not show whether service improved.
Use the results to decide whether to expand, adjust the setup or keep staffed ordering. The comparison also gives you better inputs for a cost decision, but it does not replace an installed-cost quote or a break-even calculation.
Choose comparable shifts and record the differences
Pick a baseline period with staffed ordering and a pilot period with the kiosk available. Compare the same day of the week and service window where practical. A Tuesday lunch should not be judged against a Saturday rush without noting the difference.
Keep the menu, prices, promotions and staffing assignments as consistent as possible. Record anything you cannot hold steady: weather disruptions, a large group, a sold-out item, a staff absence or a kitchen equipment problem. These notes help explain a result; they are not a reason to discard an inconvenient shift.
Choose a unit of measurement before collecting data. Count a completed order when it reaches the normal accepted-order record, not when a guest starts tapping. One guest party may place one order or several, and one order may contain meals for several people. Record parties separately if you want to understand how many groups approached the ordering area.
For both periods, assign a shift lead to collect:
- Completed orders: the accepted-order count for the same service window, from the order record in your POS.
- Queue time: the time from joining the ordering line to reaching a cashier or an available kiosk. Sample at preselected intervals and record the sample size.
- Assistance: the number of kiosk attempts needing employee help and the approximate time the helper spends away from other work.
- Abandoned attempts: attempts that stop before an accepted order, with the reason if an employee can observe it without guessing.
- Service conditions: staff on duty, channel availability, unusual demand and kitchen delays.
Use the same queue definition and sampling schedule for the staffed baseline. For example, a shift lead might time every fifth party joining the line rather than selecting only the longest waits. Record whether the line leads to a cashier, a kiosk or a choice of either; a shared line can make channel-specific comparisons harder.
Separate what you observe from what you assume. An order count is an observed record. A guess that a guest would have bought elsewhere if the kiosk had not existed is an assumption. Review the ordering and kitchen records together so a front-counter improvement is not mistaken for faster food production.
Use a decision worksheet, not a single sales number
The following example is hypothetical. Assume each column represents a comparable two-hour lunch window, with the same menu, prices and staffing. A shift lead times 20 parties per window using the same sampling rule. The owner sets the thresholds before the pilot: at least 90 completed orders, average sampled queue time no more than four minutes, and no more than one in five kiosk attempts needing help. The owner also requires a workable fallback for failed attempts.
The two kiosk outcomes are alternatives, not consecutive results or a forecast. The assistance rate uses kiosk attempts as its denominator; completed orders include orders accepted at either the kiosk or the counter.
| Measure | Staffed baseline | Kiosk outcome A | Kiosk outcome B |
|---|---|---|---|
| Completed orders | 90 | 96 | 84 |
| Average sampled queue time | 5 minutes | 3 minutes | 6 minutes |
| Kiosk attempts needing help | Not applicable | 8 of 50, or 16% | 18 of 45, or 40% |
| Abandoned kiosk attempts | Not applicable | 2 | 7 |
| Fallback observed | Counter available | Counter accepts redirected guests | Counter available, but redirection is inconsistent |
Outcome A meets the stated operational thresholds and has a working fallback. It supports considering a wider pilot, provided the result holds across additional comparable shifts and the installed-cost quote is acceptable. It does not establish that the kiosk made the extra six orders or that labor costs fell.
Outcome B misses every numerical threshold, and its inconsistent fallback needs attention. Keep the current setup from expanding while you investigate. Watch where guests need help, check whether the menu or instructions are confusing, and repeat the comparison after a specific change. If results still miss the thresholds, staffed ordering is the stronger choice under these assumptions.
Do not add abandoned attempts to completed orders or call each one a lost sale. A guest may complete an order at the counter, leave without ordering or return later. Record the path where it is observable, and avoid counting a redirected guest twice as a completed order.
Keep the operational worksheet separate from the cost comparison. These examples include no fixed costs, variable costs or exception costs. Before expansion, request an installed quote covering hardware, software, installation and support, then ask about payment processing, replacement equipment and other charges that may vary with use or exceptions. Compare quoted processing fees separately from the all-in operating cost. Use the kiosk break-even guide after you have credible volume and cost inputs.
Observe failures without disrupting guests
A pilot can look successful if it counts only orders that reach the POS. Assign the shift lead to watch for attempts that stop at the kiosk and record a short reason: guest chose the counter, could not find an item, needed an unsupported payment method or left for an unknown reason. Do not infer a reason when nobody knows.
The counter employee should offer a staffed path when a guest cannot finish. The shift lead records that handoff separately from an accepted counter order. This prevents an abandoned kiosk attempt from being treated as a sale while still recognizing that a guest may have been served another way.
Track assistance even when an order succeeds. A kiosk that needs frequent employee intervention may move work from the register to the dining room rather than free up capacity. Note the helper's role and whether help interrupted order taking, food running or another task. Avoid converting a few observed minutes into a promised staffing reduction.
Also watch the kitchen. If accepted orders increase but tickets wait longer for preparation, the ordering channel may have shifted the bottleneck. Compare order acceptance with kitchen timing where your records allow it, and note when the kitchen is already at capacity. A restaurant POS system can be part of the record you review, but confirm which timestamps and reports are available in your setup.
Accessibility belongs in the failure review. Ask whether guests can read and reach the screen, understand the next step and use a staffed alternative without extra friction. Invite feedback without making individual guests responsible for diagnosing the system. Document the issue and the practical fallback.
Set acceptance criteria and demonstrate the workflow
Before buying more units, write down the minimum result, who owns each measure and when the owner will review it. Use thresholds that reflect your service: a high-volume counter may prioritize queue time, while a restaurant with complex custom orders may put more weight on assistance and accuracy. Include several comparable shifts rather than relying on one good lunch.
Set a review date before launch, such as after the agreed pilot shifts are complete. At that meeting, choose one of three actions: expand to a limited next site, change a named issue and repeat the comparison, or stop expansion. Record the reason and the person responsible for the next action. If conditions changed substantially during the pilot, schedule more comparable shifts before making the call.
Run a vendor demonstration before live service. Agree on the test order and expected result in advance. Fleksa's kiosk page describes browsing, item customization and order review, with suggested add-ons possible; confirm the exact options and setup you intend to buy. Ask how the cart workflow handles edits before an order is submitted.
- Build and revise an order. Ask the demonstrator to add an item, make a relevant customization, review the cart and change or remove an item. Pass if the final review matches the agreed choices before submission.
- Check the handoff. In a simulation, have the demonstrator show where the order would appear for staff and how an error would be identified. Pass if the shift lead can locate the expected information and explain the next action.
- Interrupt the flow. Simulate a guest who cannot complete an attempt. Pass if the counter employee can describe a clear staffed fallback and the shift lead knows how to log the abandoned attempt without counting it as an accepted order.
- Check the quote and setup. Ask which hardware, payment methods, installation and support are included, and what happens when a device or connection fails. Pass if the vendor identifies the proposed setup and outstanding items in writing.
The shift lead owns demonstration exceptions: log each failed or unanswered item, assign it to the vendor or restaurant owner for resolution, and retest before live service. A demonstration pass is not a live-service pass. Practice in a simulation that creates no real card payment, guest message or courier dispatch. Then run the staffed live pilot and measure actual service using the agreed worksheet.
For your next step, list your menu, hardware, payment and fallback requirements. Compare them with the Fleksa restaurant kiosk and confirm current support and a written setup quote before choosing a workflow.
Frequently asked questions
How many shifts should a kiosk pilot include?
Use several comparable shifts across the service periods you care about. Agree on the number and review date before starting, and add shifts if unusual conditions make the comparison unreliable.
Should abandoned kiosk attempts count as lost sales?
No. Count them as attempts that did not become accepted kiosk orders. A guest may order at the counter instead; record that outcome separately when you can observe it.
What if the kiosk reduces queue time but needs frequent help?
Review both measures against your preset thresholds. If assistance exceeds your limit, investigate the cause and repeat the pilot after a change rather than treating the shorter queue alone as a pass.
