SMS Notifications

At 3:14 AM your phone buzzes. The text reads: "CRITICAL: checkout-api p99 > 5s, 40% errors, us-east-1. INC-4821." You were in deep sleep 10 seconds ago. Now you are reaching for your laptop.

SMS is the second most reliable way to wake a human (after a phone call). It works when the WiFi is down, when the app is force-closed, when Do Not Disturb is on (if configured correctly), and when the engineer is in an area with patchy data coverage. Twilio routes more than 1 billion messages per day across global carriers and is the backbone for PagerDuty, Opsgenie, and Splunk On-Call SMS delivery.

But SMS is also the channel most likely to silently fail. Carrier queues can back up for 30 minutes during regional events. Do Not Disturb filters drop messages from unknown senders. Short codes get flagged as spam. During the September 2022 Twilio US West outage, a major fintech lost paging capability for 47 minutes while their internal monitoring reported all systems healthy.

The question is not whether SMS works. It does, most of the time. The question is what happens during the 2 percent of the time when it does not, and how your escalation policy handles those gaps.

SMS paging looks simple from the outside: PagerDuty sends a text, your phone buzzes. In practice, the message passes through four separate organizations, each with its own queue and failure mode.

PagerDuty fires more than 4 million phone or SMS alerts per year for its 22,000 customers. The default 5-minute ack SLA assumes every hop works. When one does not, the escalation policy must move to the next responder fast enough that the incident does not age out while the page sits in a carrier queue.

SMS costs roughly $0.0079 per message via Twilio to US numbers. Voice calls cost $0.014 per minute. A team that pages 50 times per day spends about $12 per month on SMS and $50 on voice fallback. Those numbers are trivial compared to a missed SEV1, which is why every mature paging setup pairs SMS with a voice call that fires 30 seconds later.