The phone was part of the system.

The sheet below is invented, but most organizations of any size keep one like it. It lists the systems the business depends on, where each one runs, which vendor supplies it and who owns it. It is reviewed every year, and it is usually right about everything it contains. The note clipped to it is invented too. It is the kind that was written in a great many offices last Wednesday afternoon.

Last Wednesday, January 14, Verizon's wireless network failed across much of the United States for about ten hours. USA TODAY reported that the trouble began about 12:30 p.m. Eastern; NBC News put it at about noon. Verizon said at 10:15 p.m. that the outage had been resolved.¹²⁵ Downdetector, which counts reports rather than affected customers, logged more than 1.5 million of them, more than for a Verizon outage last August.² Phones dropped into SOS-only mode, and several major cities told residents to use another carrier to reach 911.²³ A Verizon spokesperson tied the failure to a software issue and said the company had no indication of a cyberattack.³ On Thursday Verizon offered affected customers a $20 credit, and wrote: "This credit isn't meant to make up for what happened. No credit really can."¹ The Federal Communications Commission said it was continuing to investigate.⁵

My argument is that the lesson for most organizations is not about Verizon. Dependency inventories are built around the systems on your list: applications, servers, cloud services and vendors, each with an owner. Wednesday's failures ran through the phone in every pocket, which sits inside the login process, the payment process and the field process without appearing on the list at all. A component nobody lists has no owner, so it has no fallback and no test.

This is not an argument that every organization needs a second carrier. That is one option among several, and a second carrier can bring shared failures of its own. Nor is it a claim about what went wrong inside Verizon. As of this writing the company has said only that a software issue was to blame and that it is conducting a full review, and guessing beyond that would be unfair.³

What actually stopped

The analyst firm Forrester made the distinction the day after. For most IT teams, it wrote, cellular service is a backup path for the corporate network, so an outage barely registers. For some enterprises it is the lifeline of the business itself: field and fleet operations, asset tracking, payment terminals and kiosks.⁶ Forrester estimated the total cost to the US economy at about $500 million. That is an analyst's estimate, drawn from the effects of similar past outages, not a measured loss.

The reported effects were ordinary, which is the point. Sports teams told fans to carry their tickets in a form that worked offline. People missed purchases because a two-factor code sent by text never arrived.⁴ Local reporting found rideshare and delivery drivers who lost the apps that hand them orders and the GPS that guides them, and some simply went home; hours lost that way are not made up later.⁷ None of those failures was in the systems themselves. The ticketing systems, the payment systems and the dispatch apps were running. What failed was the link that every one of them assumed would be in someone's hand.

Systems you run, systems you use

A register like the one above answers the question it was built for: which systems does the organization run? The outage asked a different one: which systems does the work use? The second list is longer, because it includes things no team operates. The phone is the clearest example. It is the second factor for administrators and finance approvers. It is the radio inside a card reader or a delivery tracker. It is where field staff receive their day and where customers receive their reminders. Often the phone number is also an identity, copied into the sign-in system, the customer database and the alerting tool, a fact said everywhere instead of once, so a single dead network breaks it in many places at once.

Because the phone belongs to everyone, it belongs to no one. Nobody's register shows it, so nobody has planned what the process does without it, and nobody has tried.

The credit and the cost

Verizon's credit was priced against the phone line, which is what a carrier sells. The company said that $20 covers, on average, several days of service, and that small business customers would be contacted directly about their credits.¹ That is a fair measure of the service that was missing. It is no measure of what the missing service was holding up. Forrester noted that carriers often commit to five-nines availability, which allows no more than about five minutes of downtime a year.⁶ Wednesday's outage used more than a century of that budget in one afternoon. Whatever a contract promises, the exposure that matters is the process on your side of it, and only you can size it.

Objections, considered

Outages this large are rare. Large ones are, but the dependency is constant, and smaller outages test it often. As a Johns Hopkins business school professor told USA TODAY, outages should be treated as "a routine risk, not a rare event."⁴

Staff phones are personal; they are not our system. Many are. That is exactly why they fall off the register, and it is why the fallback has to be planned by the organization rather than left to each person.

The credit pays for the phone line. Nobody pays for the process the phone was quietly holding up.

What to check this week

• List where a phone signal is a precondition. Second-factor sign-in, above all for administrators and finance approvers; card readers and kiosks on cellular links; field, delivery and scheduling apps; customer texts and mobile tickets; the alerts that wake the on-call engineer.
• Name a fallback that does not use the same network. An authenticator app over Wi-Fi, a hardware key, printed backup codes, a wired or second link for the payment terminal, or a paper procedure that someone has actually seen.
• Run one "phones off" hour. Pick a process, assume no cellular service, and write down what stops. That list is the missing row in the register.
• Read what the carrier contract actually promises. Then stop using the credit as the measure of your exposure.

Back to the register. Every row on it was correct on Wednesday, and every system it listed was running. The line that mattered was the one nobody had written: the phones themselves, owner nobody. It takes one meeting to add it, and an afternoon to find out what it is holding up.