Latency is the time between submitting an SMS and its arrival on the handset, typically a few seconds on a healthy route; it is the metric that matters most for one-time codes.
For a code, latency is the product: a user waiting twenty seconds gives up and asks for another. Latency comes from the gateway's processing, the route's hops, the operator's SMSC and the handset's connection. It rises when a route is congested or when messages queue behind a large campaign.
Separate transactional traffic from bulk so that a marketing send never delays a login code, and measure latency per destination rather than as one average.
At SMSMeteor: the status page publishes API and delivery latency percentiles; transactional messages flagged as such are not queued behind campaigns.
