Outage Reporting Turns on 900,000 User Minutes. The Arithmetic Differs by Service

A communications outage enters the federal record when it clears the tests written into its own branch of the rule. Many branches apply two at once. The first is duration: the outage has to last at least 30 minutes. The second is size, and size is expressed in a unit that sounds like a headcount but is not one. The figure printed in the rule is 900,000 user minutes. Divide it by the duration floor and the second test becomes a headcount for exactly one case: at 30 minutes, 900,000 user minutes is 30,000 users. Let the outage run two hours and the same threshold is 7,500 users. The threshold does not move. The number of people needed to reach it falls as the clock runs.

That leaves the harder question, which is what the 30,000 is made of. For a wireless switch failure the answer is not a count of affected subscribers. It is a multiplication, and one of the two factors is a nationwide average that is the same whether the dead equipment sat in a dense downtown or in a county with one road through it. The formula is printed in eCFR, 47 CFR 4.9, Outage reporting requirements, one paragraph below the threshold it feeds. Two other paragraphs, in a different section, supply two further ways to produce the same quantity. The threshold is one number. The arithmetic behind it depends on what service the provider is filing as.

The Count That Has to Reach 900,000

The wireless trigger is 47 CFR 4.9(e)(1). Its introductory text reads: “All wireless service providers shall submit electronically a Notification to the Commission within 120 minutes of discovering that they have experienced on any facilities that they own, operate, lease, or otherwise utilize, an outage of at least 30 minutes duration:” Five subparagraphs follow, and the fifth matters as much as the second.

Subparagraph (i) is “Of a Mobile Switching Center (MSC);” — a bare equipment branch with no size test in it at all. Subparagraph (ii) is the one everybody quotes: “That potentially affects at least 900,000 user minutes of either telephony and associated data (2nd generation or lower) service or paging service;” Subparagraph (iii) is “That affects at least 667 OC3 minutes (as defined in § 4.7); or” Subparagraph (iv) is marked [Reserved], an empty slot that still has to be counted when checking that a reading covered the whole list. Subparagraph (v) reads: “That potentially affects a 911 special facility (as defined in § 4.5(e)) or potentially affects a 988 special facility (as defined in § 4.5(f)), in which case they also shall notify the affected facility in the manner described in paragraph (h) or (i) of this section, respectively.”

Two things in that list are worth slowing down for. The verb changes between branches: (ii) says potentially affects, while (iii) says affects, with no qualifier. And the circuit branch converts just as cleanly as the user branch. 667 OC3 minutes divided by the 30-minute floor is 22.2 circuits, so 23 previously operating OC3 circuits down for half an hour produce 690 OC3 minutes and clear the branch, while 22 of them for the same half hour produce 660 and do not. Both branches are products of a duration and a count, and neither count is a tally of people: one counts OC3 circuits, the other counts macro cell sites.

The Chain

The chain starts with a definition that contains no numbers. 47 CFR 4.5(a) says: “Outage is defined as a significant degradation in the ability of an end user to establish and maintain a channel of communications as a result of failure or degradation in the performance of a communications provider's network.” Nothing in that sentence tells a filer how many users were degraded. The next link supplies that, for one trigger.

47 CFR 4.9(e)(2) reads in full: “In determining the number of users potentially affected by a failure of a switch, a wireless provider must multiply the number of macro cell sites disabled in the outage by the average number of users served per site, which is calculated as the total number of users for the provider divided by the total number of the provider's macro cell sites.” The trigger named in that sentence is a failure of a switch. The input is a count of macro cell sites. The multiplier is the provider's own system-wide ratio: every subscriber on the books divided by every macro site on the books.

So the quantity that gets compared with 900,000 user minutes is an estimate built from two inventory totals and one site count. A downtown macro site and a rural macro site enter the product at the same weight, because the average is taken across the provider's entire footprint before any particular outage is considered. The same arithmetic runs in the other direction too: a provider that densifies its network, adding many small-coverage macro sites, lowers its own average users per site, and therefore needs more sites down to reach the same threshold.

From there the chain stops being labeled. The site count becomes a user count under 4.9(e)(2), but 47 CFR 4.7(e) sorts user minutes into two branches and assigns the wireless switch figure to neither by name: (e)(1) routes telephony to assigned telephone number minutes, while (e)(2) covers “all other forms of communications” with duration multiplied by end users potentially affected. Either way the product is measured against 900,000, and the comparison is where the duration factor does its quiet work: the same estimate of users clears the threshold at 30 minutes only if it reaches 30,000, at 60 minutes only if it reaches 15,000, at 120 minutes only if it reaches 7,500, and at 180 minutes only if it reaches 5,000. If it clears, three clocks start: a notification within 120 minutes, and then the two reports in 47 CFR 4.9(e)(4) — “Not later than 72 hours after discovering the outage, the provider shall submit electronically an Initial Communications Outage Report to the Commission. Not later than 30 days after discovering the outage, the provider shall submit electronically a Final Communications Outage Report to the Commission.”

The last link is the form itself. The filing manual, FCC, Network Outage Reporting System User Manual Version 3, instructs the filer at the wireless user field: “Provide the number of potentially affected wireless users.” One number goes in that box. The regulation supplies the arithmetic for that number in the switch-failure case, so for that case the figure is the output of the average rather than a tally of people who lost service. Whatever reaches that box is then inherited downstream: the 72-hour initial report, the 30-day final report that may revise it, and any later count of reportable outages.

Three Formulas, One Threshold

47 CFR 4.7 holds six lettered definitions, and the one that governs the threshold is (e), user minutes. It has two numbered branches. The first, 4.7(e)(1), reads: “Assigned telephone number minutes (as defined in paragraph (c) of this section), for telephony, including non-mobile interconnected VoIP telephony, and for those paging networks in which each individual user is assigned a telephone number;” For that class of service the unit is not people. It is telephone numbers.

And the paragraph it points at, 4.7(c), carries two computations rather than one. The first multiplies the duration by the sum of assigned numbers and administrative numbers potentially affected. The second is introduced in the regulation's own words: the quantity “can alternatively be calculated as the mathematical result of multiplying the duration of an outage, expressed in minutes, by the number of working telephone numbers potentially affected by the outage”. And the term that second method turns on is defined twice in the same section, in different words: inside 4.7(c) it is the telephone numbers, including DID numbers, working immediately prior to the outage, while 4.7(f) defines working telephone numbers as the sum of all telephone numbers that can originate, or terminate telecommunications. Two methods, one paragraph, two definitions of the term one of them depends on, and no requirement that any of it land on the same figure for the same event.

The second branch, 4.7(e)(2), covers everything else, and its closing sentence adds a third formula outright: “For interconnected VoIP service providers to mobile users, the number of potentially affected users should be determined by multiplying the simultaneous call capacity of the affected equipment by a concentration ratio of 8.” That sentence sits inside (e)(2), not in a paragraph of its own, which is easy to miss when scanning designators. Run the same arithmetic as before: to reach 30,000 potentially affected users at the 30-minute floor, affected equipment needs a simultaneous call capacity of 3,750. Capacity, times 8, against a threshold written in user minutes.

Three provisions, then, produce the quantity the threshold judges: telephone numbers times duration, macro cell sites times a system-wide average, and simultaneous call capacity times 8. The wireline trigger in 47 CFR 4.9(f) shows how little the surface wording changes between them. Its first branch is “Potentially affects at least 900,000 user minutes of either telephony or paging;” — the identical figure, feeding on the telephone-number unit rather than the site-average estimate. Its four branches are the user-minute branch, the OC3 branch, the special-offices branch that points to 4.5(a) through (d), and the 911 and 988 branch that carries the report deadlines inside it.

The Clocks Are Not the Same Either

Three different notification windows sit in the same section. Wireless under 4.9(e)(1) and wireline under 4.9(f) both run on 120 minutes. Interconnected VoIP providers get two windows, and the division between them is instructive. 47 CFR 4.9(g)(1)(i) requires a notification “Within 240 minutes of discovering that they have experienced on any facilities that they own, operate, lease, or otherwise utilize, an outage of at least 30 minutes duration” that potentially affects a 911 or 988 special facility. The other branch, 4.9(g)(1)(ii), runs on 24 hours, which is 1,440 minutes, and its first condition reads: “That potentially affects at least 900,000 user minutes of interconnected VoIP service and results in complete loss of service; or”

Note where that sentence ends. The user-minute figure is not the whole test in this branch. It is joined by a second condition in the same clause, complete loss of service, so a degradation that leaves service partly up falls outside this branch however large the user-minute figure is. The same 900,000 figure therefore does different work in 4.9(e)(1)(ii), in 4.9(f)(1), and in 4.9(g)(1)(ii)(A): in the first two it stands alone as the size test, and in the third it is one of two conditions joined by a conjunction. The follow-up filings diverge as well: 4.9(e)(4) and 4.9(f)(4) each require an initial report at 72 hours and a final report at 30 days, while 4.9(g)(2) requires only the final report.

Where the 911 Branch Changes the Test

Subparagraph (v) of the wireless trigger, and branch (4) of the wireline trigger, both point at 47 CFR 4.5(e), which opens: “An outage that potentially affects a 911 special facility occurs whenever:” Four numbered conditions follow, and the count matters. Three of the four carry the 900,000 user-minute test. The second one carries only a clock.

Condition (2) reads: “There is a loss of 911 call processing capabilities in one or more E-911 tandems/selective routers for at least 30 minutes duration; or” There is no size figure in it. Half an hour of lost call processing in a single selective router satisfies the condition, and through 4.9(e)(1)(v) or 4.9(f)(4) that makes the event notifiable without any user count being produced at all. By contrast, condition (3) reads: “One or more end-office or MSC switches or host/remote clusters is isolated from 911 service for at least 30 minutes and potentially affects at least 900,000 user-minutes; or” Same facility type, same clock, and the size test is back.

This is the part of part 4 where the estimate stops being the gate. For the three conditions that keep the user-minute test, a provider's average users per site still decides whether a 911 impact is reportable, exactly as it does for a general wireless outage. For the selective-router condition, the measurement problem disappears, because the condition is written about equipment and a duration instead of about people. An earlier piece on this site followed the same split in a different metric, where cell sites out of service are published as a percentage of a denominator that moves, and another followed one agency carrying two thresholds for the same event. The pattern repeats because the remedy is the same in each case: write the trigger about something countable.

One more detail belongs here, because it is the kind that gets over-read. The reserved subparagraph, 4.9(e)(1)(iv), is an empty branch, and the printed Code of Federal Regulations page for the companion definitions section, U.S. Government Publishing Office, 47 CFR 4.7, 2023 printed edition, confirms that the concentration-ratio sentence belongs to 4.7(e)(2) rather than standing on its own. Neither point changes a threshold. Both change what a careful reading has to account for before claiming the list was covered.

Where This Doesn't Apply

The arithmetic traced here governs three of the nine lettered paragraphs of 47 CFR 4.9. The other six — cable in (a), tandem facilities in (b), satellite in (c), signaling system 7 in (d), and the two special-facility notification paragraphs in (h) and (i) — were not read line by line for this piece, so nothing above should be carried into them.

The site-average formula in 4.9(e)(2) is written about a failure of a switch. It is the method for that trigger, and this piece does not extend it to every event in which macro cell sites go dark. A reader working an actual filing should also check 47 CFR 4.11 for the form requirements that 4.9(e)(5) incorporates, which were outside this reading.

Two gaps in the evidence are worth stating plainly. The filing manual was read through the first 100,000 of its 112,224 characters, so the quoted field instruction is what the manual asks for at that field and nothing here describes the remaining pages. And the FCC bureau report on the February 22, 2024 wireless outage, which would have shown these provisions applied to a real nationwide event, returned a host error on two attempts and was not read; no figure from any filed outage appears above. Every number in this piece is either printed in 4.5, 4.7 or 4.9, or is a division or multiplication of those figures.

Finally, the direction of the bias in the site-average estimate is not something this reading settles. Whether multiplying dead sites by a system-wide average overstates or understates a particular outage depends on where the sites were and how loaded they were at the time, and the rule asks for neither input. What the rule does fix is that the answer compared against 900,000 is a product of inventory totals. The threshold is sharp. The quantity it judges is an estimate, and knowing which of the three formulas produced it is the difference between reading the number and believing it.