Designing the human half of self-service ordering

The problem

The brief was direct: there was no process to train and upskill teams on kiosks once they were installed. Troubleshooting documents existed but were never sent to restaurants or collated. The training gap was a symptom. Four gaps surfaced in the first pass.

What I did

• Channel performance. Sales mix by store format and asset type, penetration and adoption, which showed the swing between drive-thru-led stores and the highest-usage franchisee group.
• Guest-feedback mining at two scales. The source was guest feedback running to hundreds of thousands of responses, comments and ratings. It was read at two scales: every self-service comment in a three-month pull read in depth, and a wide sweep across a much larger store population to measure how often the channel surfaced at all. The absence signal was the finding: the channel was near-invisible in the wide sweep, and where it appeared, staff recognition was the top positive theme.
• Benchmarking. Four franchisee groups running parallel floor-role programmes, compared as archetypes, plus a scan of markets with older kiosk footprints.
• Pilot measurement. Mirror-store surveys on menu-board removal, store-level adoption tracking, and a monthly cut benchmarking a pilot store pair.
• Mapping and design. The omnichannel journey on the agency's spine, seven insights, one concept converged from the four franchisee designs, the floor-role standard, the training ecosystem, labour-deployment arithmetic, the placement playbook, the troubleshooting FAQ and the governance roadmap.

What the research found

Installation is not adoption. The machine was the constant; the human was the variable.

The role already existed, four times. The best of the four franchisee designs converged into one standard rather than a fifth invention.

The failures were about people, recorded as machine complaints. A first-timer who gave up, a cash guest excluded, a failed kiosk, a dirty foyer nobody owned: each was logged as a kiosk problem.

Awareness of the flagship programme was zero in the sample. None of the restaurants sampled had implemented, or heard of, the programme meant to fix this.

Menu boards were not the comprehension problem. The mirror study on menu-board removal showed comprehension held without them where a host was present, which freed wall space and settled a long argument.

Three decisions

What was delivered

The floor-role standard as it was written: fields, not prose, so it can be versioned, reviewed and checked.

sop: QSR-FHR-001   Foyer host floor-role standard   version 1.0
owner: channel_lead
triggers:
  - kiosk_installed AND weeks_since_install < 26
  - adoption_share < target        # share of in-store sales via kiosk
presence: { required: peak_hours, rosterable: true, reporting: restaurant_manager }
scripts:
  greeting: "Hi, I'm [name], your ordering host. Would you like faster, cashless ordering at the kiosk?"
  assist: help_navigate_without_judgment
  direct: cash_and_specific_piece_requests_to_counter_queue
pre_peak_checks: [menu_availability_synced, collection_screen_functional, kiosks_cleaned, receipt_and_payment_stock_ok]
duties:
  primary: [greet, assist, direct, drive_adoption]
  secondary: [foyer_cleanliness, high_touch_surface_cleaning_hourly]
exceptions:
  cash_guest: counter_queue_with_smile
  out_of_stock: script_out_of_stock + counter_fallback
  tech_failure: first_line_ladder   # retry, error-code catalogue, manager reboot, stand-alone reprint
kpis: [adoption_share, feedback_staff_recognition, feedback_cleanliness, cash_wait_complaints, first_line_recovery_rate]
review_cadence: monthly

Also delivered: the channel review and benchmark report, the omnichannel journey map and the kiosk order workflow below, the training ecosystem (huddle guide, job aids, station stickers linked to scenario videos, the card above), deployment cards and the redeployment calculator, the implementation playbook with the kiosk-count formula and the submission and approval flow for new installations, the troubleshooting FAQ with the error-code catalogue, the placement self-assessment template, and the governance roadmap.

Kiosk order workflow: guest frontstage, restaurant backstage, measures and fail points per step; click any cell to expand the design detail.

Stage / lane	Enter & greet	Order	Pay	Prepare	Collect & enjoy
Guest (frontstage)	Greeted within seconds The guest enters and is met at the door: before the redesign nobody owned this space.	Guided first order Assist script: help navigate, never judge, offer the counter as an equal path.	Card at the kiosk Kiosks take no cash by design; cash guests are directed to the counter queue with the stealth register.	Waits at pick-up queue Collection screen shows order status; “Checking kitchen” delays are acknowledged and reassured.	Order verified at hand-off The presenter confirms the order; the feedback link lands on the receipt.
Foyer host (frontstage)	Greeting + cash-or-card ask No customer enters without being greeted; the direct script sends cash and specific-piece requests to the counter.	Assist + pre-peak checks Menu availability synced, collection screen functional, kiosks cleaned, receipt and payment stock ready.	Cash path to the counter Cash guests are never abandoned at the machine; the stealth register absorbs the queue.	Ready flip within 60 s The order is flipped to ready inside the 60-second order-to-collect rule.	Hand-off + parting recognition The guest leaves recognised, not ignored: the moment that drives the feedback theme.
Restaurant (backstage)	Rostering + cleanliness The foyer host is a rostered role: the time & attendance platform now supports it; hourly high-touch cleaning.	POS receives order; menu sync Menu availability is refreshed nightly; express-kiosk exclusions and availability mismatches are scripted.	Payment state verified Pin-pad timeouts and failed payments have defined responses before the guest escalates.	Kitchen display + packer One packer/presenter per redeployed cashier assembles the order to the display.	Collect / recovery First-line ladder: retry → error-code catalogue (100–699) → manager reboot → stand-alone reprint.
Measures	Greet rate; cash routed Every guest greeted at peak; cash guests routed, not turned away.	Assisted first order Guided first order completes in ~60 seconds.	Counter wait ≤ 5 min Guest through the system inside the 5-minute golden rule.	Order-to-collect ≤ 60 s Ready flip measured per order.	Feedback theme mix Staff recognition top positive theme; channel near-invisible in the wide sweep.
Failure points	F-01 No one owns the door No named foyer role: nobody greets; the space between door and machine is unowned.	F-02 First-timer struggles Guest gives up at the kiosk and asks staff for help: with no assist script.	F-03 Cash guest excluded long counter waits; cash customers have no path but the queue.	F-04 Order forgotten / tech failure Order not flipped; all-kiosk outages and payment timeouts with no team response.	F-05 Machine blamed in feedback A service failure is recorded as a kiosk complaint; no recovery script.

Working files, anonymised: the benchmark report. The brief, the channel analysis, the journey map, the blueprint and the analysis report stayed working documents with the business.

How it would be measured

What I would do differently: run the post-implementation phone check as part of the programme from the first installation, rather than as the check that found nobody had heard of it.

Sponsor, scope and gates