Technology · Latency
Real time has
a stopwatch.
A screen can use a live listener and still feel late. The guest experiences one path from a tap to the field device showing an actionable task, so the useful software latency budget includes every hop to that screen.
The definition
Measure from intent to actionable state.
The clock starts when the guest presses Request. It stops when the right field device shows a clear task that a valet can act on. Server processing time is only one interval inside that journey.
Valletto does not publish a universal latency number because one has not been measured across every property network, handset, and notification state. A buyer should ask every vendor, including us, to run the test in the real environment.
The budget
One promise, five intervals
Break the journey apart
→A median for the server interval cannot describe a delayed notification or a sleeping device. Stop the platform clock at the actionable render, then measure valet acknowledgement separately so staffing and attention do not distort software latency.
Measure the tail, not only the happy path
Averages can hide the requests that arrived after the guest reached the curb.
→Illustrative shape only. Record your own median, P90, P95, and non-delivery rate rather than borrowing a benchmark.
Run the curb test
Use synchronized timestamps or a filmed clock so every interval can be reconstructed without relying on memory.
States
Foreground and locked
Repeat with the valet app open, backgrounded, and the phone locked. Delivery behavior can differ.
Networks
Good, weak, and changed
Test property Wi-Fi, cellular in the garage, and a transition between them.
Load
One request and a rush
A quiet test proves correctness. A burst reveals queues, fan-out limits, and screen ordering.
Outcome
Visible and actionable
Do not stop at a delivery log. Confirm the correct device displayed the correct task in the right context.
The procurement standard
Ask for the distribution, the test setup, and the failures.
A defensible answer names where the timer starts and stops, which device states were included, how many trials ran, and how non-deliveries were counted. 'Instant' and 'real time' are not measurements.
The takeaway: if a real-time claim cannot be expressed as a reproducible curb test, it is branding rather than a latency budget.
