The Gold Standard of Utilities | Counting Overlapping Interruptions Without Double-Counting Customer Time

Imagine Cedar Quay has four customer accounts, A, B, C and D. For this exercise, a customer is either receiving the defined service or not receiving it. The record identifies those states at each account’s connection. We are measuring time without service, rather than time spent repairing equipment.

That boundary matters. A pump could remain under repair after another supply route restores a customer’s service. Conversely, closing a fault ticket might precede confirmed restoration at a particular building. Neither equipment status nor administrative completion automatically supplies the customer’s start and end times.

Write a measurement sentence before doing arithmetic: this record describes minutes without the defined service at these four connections during this observation period. Keep a separate sentence for what the repair log measures. A clear boundary prevents one dataset from quietly taking over the job of another.

The binary model is deliberately simple. Water service can involve pressure and quality conditions; electrical service can involve conditions beyond complete loss. Later chapters examine partial or uncertain states. For now, assume the fictional record has already established when the exercise’s service criterion is unmet.

This is a useful first-principles habit in Mathematics and public reporting. Define the object being counted, the condition that makes it count and the time boundary. A correct sum is only meaningful after those decisions. The existing Water Utility Companies guide explains the wider supply and responsibility questions that a duration calculation alone cannot settle.

Contents · Next chapter

Use intervals to describe time. In this guide, 09:00–10:00 means sixty minutes beginning at 09:00 and ending when service is restored at 10:00. If another interruption begins at exactly 10:00 without a restored period between, the customer’s service remains continuously unavailable.

A mathematical convention can represent the first interval as including its start and excluding its end. You do not need special notation to calculate these examples. The purpose is to make adjacent records agree at their shared boundary, rather than imagining that the clock creates an extra minute at every endpoint.

Do not count printed clock labels. A list reading 09:00, 09:01 and 09:02 spans two minutes between the first and last timestamp. It does not establish three complete minutes of interruption. The timestamp says where a boundary lies, while duration measures the distance between boundaries.

Now distinguish precision from accuracy. A system may store seconds while the original observation was only approximate to a minute. Printing a timestamp with many digits does not make the underlying event boundary equally certain. Preserve the source’s actual resolution and the method used to establish the time.

For a classroom check, ask a learner to compare 09:00–09:20 and 09:20–09:45. With no restored service between, the total is forty-five minutes. If service was actually restored from 09:20 until 09:25, the second interval must instead begin at 09:25. Then the two interruptions contribute twenty plus twenty, or forty minutes. The wording changes the data, so it must change the answer.

Contents · Previous chapter · Next chapter

Customer B appears in two fictional fault records. Fault F1 affects B from 09:00 until 10:00. Fault F2 affects B from 09:30 until 10:30. Adding their listed durations gives sixty plus sixty, or 120 minutes. But B was without service continuously for only ninety minutes, from 09:00 until 10:30.

The common half-hour, 09:30–10:00, appears in both records. Subtracting that thirty-minute overlap gives 120 minus thirty, or ninety. Equivalently, draw the periods on a line and measure their combined coverage once. Both methods answer the same question about B’s time without service.

It would be misleading to infer that either fault was unimportant. Both may be relevant to engineering investigation and restoration work. Deduplicating customer time does not erase causes or responsibilities. It only prevents the same customer-minute from being counted twice in a service-duration measure.

Now change F2 so that it affects B from 10:15 until 10:45, with service restored from 10:00 to 10:15. The durations are sixty and thirty minutes, totalling ninety again. The total is unchanged, but B now experiences two separate episodes instead of one. This is why duration alone cannot describe continuity.

An original repair check follows. A learner writes that two faults always mean two interruptions. Ask them to locate the restored period between the faults. If there is none, the fault count and episode count cannot be treated as interchangeable. If the record does not establish the service state between them, the correct next step is to mark that uncertainty rather than choose whichever count looks tidier.

Contents · Previous chapter · Next chapter

The main Cedar Quay exercise has three fault records. F1 affects A and B from 09:00–10:00. F2 affects B and C from 09:30–10:30. F3 affects C and D from 09:45–10:15. Assume the listed periods accurately describe loss of the defined service at those connections.

Multiplying customers by duration for each fault produces 120, 120 and sixty customer-minutes. Adding these figures gives 300. That is a sum of fault-associated entries, with overlaps retained. It is not yet the deduplicated account of customer time.

Reconstruct each account instead. A appears only in F1, so its duration is sixty minutes. B’s two overlapping periods combine into ninety. C’s period in F3 lies entirely inside its F2 period, so its total remains sixty. D appears only in F3, contributing thirty. The correct combined customer duration is therefore sixty plus ninety plus sixty plus thirty, or 240 customer-minutes.

The correction removes thirty duplicate minutes for B and thirty for C. Thus 300 minus sixty equals 240. Keeping the raw total alongside the corrected total is useful: it explains exactly why the number changed. A reviewer can trace the correction to particular overlaps rather than being asked to trust an unexplained adjustment.

Contents · Previous chapter · Next chapter

There is another way to verify the main example. Split the overall period wherever an affected-customer state changes. In each segment, count how many different accounts lack service, then multiply that count by the segment’s duration.

From 09:00–09:30, A and B are affected: two accounts for thirty minutes give sixty customer-minutes. From 09:30–09:45, A, B and C are affected: three for fifteen give forty-five. From 09:45–10:00, all four are affected: four for fifteen give sixty.

From 10:00–10:15, B, C and D remain affected: three for fifteen give forty-five. Finally, from 10:15–10:30, B and C remain affected: two for fifteen give thirty. The sum is sixty plus forty-five plus sixty plus forty-five plus thirty, again 240.

The account method groups the calculation by customer; the timeline method groups it by time. Agreement is a meaningful check because the two routes organise the same information differently. If the totals disagree, inspect an endpoint, an identity or an overlap. Do not average the two answers. A mismatch is evidence of a problem to resolve, not a reason to split the difference.

Contents · Previous chapter · Next chapter

A repeated report is not necessarily a repeated event. A customer might telephone, submit an online form and receive a follow-up ticket for one interruption. Three messages can therefore describe one account’s one service-loss episode.

Keep message identifiers and event relationships separate. The message record preserves who reported what and when. The reconstructed service record identifies the account, relevant period and supporting evidence. Linking messages to an event allows both records to remain useful without converting every contact into another interruption.

Suppose A’s online report arrives at 09:10 and a telephone report at 09:20. Neither time automatically replaces the established 09:00 start in the exercise. Report time and event time differ. The earliest received message is not necessarily the moment service first failed.

The opposite mistake is also possible. Two genuine episodes may be merged because they concern the same account and occur close together. If service was restored between them, preserve that fact under the exercise’s episode definition. Similar wording is not enough to establish duplication.

A good reconciliation note explains the reason for linking or separating records. It can say that two messages refer to the same account and continuous service-loss period, supported by the stated evidence. It should not merely say duplicates removed. That short explanation lets another reader distinguish a sensible correction from an accidental deletion and keeps the original observations available for review.

Contents · Previous chapter · Next chapter

Return to the main example. Four distinct customers are affected. Under the defined continuous-service-loss rule, each experiences one episode, so there are four customer episodes. The repair log contains three faults. All three totals are correct within their own definitions.

The number of fault-account associations is different again: F1 lists two accounts, F2 lists two and F3 lists two, producing six associations. Calling those six different customers would overcount B and C. Calling them six separate customer interruptions would also be wrong under this exercise’s continuous-episode rule.

For a changed case, let A lose service again from 11:00–11:20. The distinct affected-customer count stays four because A was already included. The customer-episode count rises from four to five. Accumulated customer-minutes rise from 240 to 260. Each measure responds differently because each asks a different question.

Write the question above the result. How many different accounts experienced a loss? How many separate episodes occurred at accounts? How much account-time was lost? How many engineering faults were recorded? A report becomes clearer when readers do not need to infer which question a single interruption count was meant to answer.

Real reporting conventions can classify short interruptions, planned work, major events and other cases in particular ways. The exercise’s definitions should not be silently labelled as an official reliability index. The EIA reliability explanation provides authoritative context for why duration and frequency are distinct measures with specified reporting scope.

Contents · Previous chapter · Next chapter

If Cedar Quay contains only the four accounts in the main example, 240 customer-minutes divided by four gives an average of sixty minutes per account. B experienced ninety minutes and D thirty. The average describes the group; it does not replace the individual records.

Now suppose the same affected accounts belong to a system with twenty accounts in total, the other sixteen receiving service throughout the exercise. Dividing 240 by twenty gives twelve minutes per system account. Dividing by the four affected accounts still gives sixty minutes per affected account. These are different denominators, answering different questions.

Do not compare one provider’s average over all accounts with another’s average over affected accounts as though they measure the same thing. An apparent improvement can be created simply by switching denominators. Write the eligible population beside the numerator before calculating.

Changing populations need additional care. If accounts join or leave during a reporting period, an official method determines the appropriate population measure. This guide does not supply a universal choice between beginning, ending or average customer counts. Its lesson is to keep that choice visible and consistent with the question.

For an independent check, retain 240 customer-minutes but use thirty system accounts. The average becomes eight minutes per system account. No interruption has been repaired by this arithmetic. Only the comparison population changed. A learner who can explain that difference understands why a summary can move without any change in the experience of the affected people.

A second check works backwards from a published average. If a fictional system reports twelve minutes per account across twenty accounts, multiplying gives 240 customer-minutes under the stated definition. That reconstruction does not reveal which accounts were affected or how many episodes they experienced. Many different histories can produce the same total.

Ask a learner to invent two histories that match it. One account losing 240 minutes and four accounts each losing sixty both produce 240 customer-minutes. If the observation period is only ninety minutes, the first history is impossible within that period. The time boundary therefore adds a constraint that the total alone did not express.

This is a useful inverse-problem lesson: recovering one total from an average is possible when the denominator is known, while recovering the complete distribution generally is not. A report should not turn a mathematically reconstructed sum into an invented story about individual customers.

Contents · Previous chapter · Next chapter

A district-level message saying restoration has begun does not identify every account’s end time. If ten customers regain service at 10:00 and another ten at 10:30, using 10:00 for all twenty understates the latter group’s interruption.

Consider an independent fictional event beginning at 09:00. Ten accounts are restored after sixty minutes and ten after ninety. The total is ten times sixty plus ten times ninety, or 1,500 customer-minutes. Treating all twenty as restored after sixty minutes gives 1,200, an understatement of 300.

This example uses complete restoration at each defined connection. A different exercise might include limited supply, reduced pressure or another partial service state. Those conditions need their own definitions and evidence. A single binary duration figure cannot explain every aspect of service quality.

Keep the status description understandable. State which accounts have confirmed restoration, which remain affected and what is not yet verified. There is no need to invent an overall completion time merely because a dashboard expects one field.

The engineering sequence can still be recorded separately: component isolated, repair completed, testing undertaken, service verified. Those stages may be important, but they are not interchangeable timestamps. The Electricity Utility Companies guide is the wider route into reliable supply and recovery. The focused arithmetic here helps readers ask whether a service-duration claim actually follows the users through the recovery stages.

Contents · Previous chapter · Next chapter

Suppose a record confirms that account C lacked service at 09:30 and was still affected at the 10:30 reporting cutoff. Its final restoration time is not known. The record establishes at least sixty minutes of interruption by that cutoff, rather than a completed sixty-minute episode.

Do not fill the missing restoration time with the cutoff and then present it as an observed recovery. The cutoff describes the end of observation. Restoration describes a change in service state. They can coincide, but the record must establish that coincidence.

A useful provisional account can report accumulated customer-minutes observed up to a stated time, together with the number of ongoing episodes. Later information can extend the record and establish final durations. Keep the provisional and completed measures labelled so that a later increase is not mistaken for an arithmetic correction.

There is also a difference between no loss observed and no observation available. If connection status was unavailable for part of a period, entering zero lost minutes assumes continuous service without evidence. Depending on the task, the unknown period may need to remain outside a confirmed total or be represented through explicit bounds.

For a classroom bound, imagine a ten-minute unknown interval for one account, with all other information complete. Its additional lost time lies between zero and ten minutes under the binary exercise. That is an uncertainty range, not a probability distribution. The record gives no basis for saying that five minutes is the most likely answer.

If an unknown interval overlaps a period already confirmed as unavailable for the same account, it adds no new uncertainty to the total customer-time within the overlap. The service state there is already established. Reconcile evidence by account and interval before adding uncertainty ranges, just as you reconcile fault periods before adding durations. This prevents an uncertainty budget from double-counting the same unresolved minute under two separate records.

Contents · Previous chapter · Next chapter

A revised total should be accompanied by a reason. Perhaps an account identifier was corrected, an overlapping interval was merged or a restoration time was verified. These are different changes, and they can affect different measures.

Imagine the first report totals 300 customer-minutes from fault entries. Reconciliation identifies B’s thirty duplicate minutes and C’s thirty duplicate minutes. The revised customer-time total is 240. A clear note says that customer periods were deduplicated; it does not imply that sixty minutes of actual service loss vanished.

Now suppose later evidence moves D’s restoration from 10:15 to 10:20. D contributes thirty-five rather than thirty minutes, and the total becomes 245. That revision concerns the measured endpoint, not duplicate removal. Preserve both explanations if they occurred in sequence.

Separate correction from method change. If a new reporting method excludes a particular category, comparisons with an earlier total may need a consistently recalculated series or a visible break. A lower number produced by a new exclusion is not automatically an improvement in service.

For practical learner work, keep a short revision table with the previous value, changed record, reason and new value. A good reviewer should be able to reproduce the movement. This habit transfers into laboratory records, school survey summaries and project accounts: retaining the reason for a changed number makes the result more trustworthy than quietly replacing the original figure and hoping nobody notices.

Contents · Previous chapter · Next chapter

Two systems can accumulate the same customer-minutes through different patterns. One may have one long interruption affecting a small group; another many shorter interruptions spread widely. Equal duration summaries do not establish equal inconvenience or equal service experience.

An original comparison makes this visible. System P has ten accounts each losing sixty minutes, giving 600 customer-minutes. System Q has sixty accounts each losing ten minutes, also giving 600. If both systems serve 100 accounts, each averages six minutes per system account. The affected proportions are 10% and 60% respectively.

Neither is automatically better in every respect. The importance of an interruption depends partly on the user’s task and alternatives, which this table does not measure. Keep those consequences separate from the duration arithmetic rather than inventing a universal inconvenience score.

Official indices make defined choices about duration, frequency and inclusion. The EIA Electric Power Annual reliability notes illustrate the importance of reading those definitions. The Cedar Quay examples are educational measures, not submitted regulatory statistics.

Before comparing real summaries, match the reporting period, service boundary, eligible customer population, interruption inclusion rules and treatment of ongoing records. If one condition differs, state the limitation. Numbers remain useful when their scope stays attached; they become misleading when a convenient similarity in units conceals a different underlying account.

Contents · Previous chapter · Next chapter

A school may lose electricity and internet access during the same period. Adding electricity customer-minutes to internet customer-minutes can describe a defined sum of service-specific losses. It does not automatically describe the number of minutes during which the school could not perform one activity.

Suppose electricity is unavailable from 09:00–09:30 and internet access from 09:20–10:00. An activity requiring both services is prevented by either loss. Its unavailable period is the combined interval 09:00–10:00, or sixty minutes. Adding thirty and forty gives seventy because the common ten minutes appears twice.

An activity that can continue without internet has a different dependency. Its lost time cannot be inferred from the same combined-service calculation. Define the task before choosing whether to combine service states through either-service failure or some other condition.

This is the same union principle in a changed context, but the objects have changed. Earlier, we combined fault periods for one customer’s one service. Here, we combine different service requirements for one activity. The arithmetic transfers only after the meaning of each interval is established.

The Urban Systems guide connects these dependencies at a wider scale. Keep that relationship useful without pretending that a single duration statistic can express every downstream effect. A dependency explanation needs the task, required services and actual conditions, alongside the time record.

Contents · Previous chapter · Next chapter

For a small exercise, draw the periods. For a larger fictional record, use a consistent procedure: group by account and service, order valid intervals by start time, and compare each period with the end of the current merged interval.

If the next start lies before the current end, the periods overlap. Extend the merged end only if the next end is later. If the next period lies completely inside the current one, it adds no duration. If a verified restored gap separates them, finish the current interval and begin another.

The procedure depends on the episode convention for touching intervals. In this guide, touching periods with no service restored between form one continuous episode. A different official method may distinguish events using additional criteria. State the rule instead of assuming that a software default supplies the correct service definition.

Check input quality first. An end earlier than a start may reveal a date error, a crossing of midnight or another unresolved problem. Two accounts sharing a similar name should not be merged without a reliable identity. A timezone difference can also make apparently overlapping local times misleading.

Finally, inspect a few reconstructed customer histories and cross-check the overall timeline. A spreadsheet can perform the arithmetic while still producing the wrong measure if its grouping or interpretation is wrong. The learning goal is a procedure the student can explain: what is grouped, what is merged, which records remain uncertain and why the resulting sum matches the stated question.

Contents · Previous chapter · Next chapter

Use this independent fictional record. Account R loses the defined service from 13:00–13:40 and again from 13:20–14:00. Account S loses service from 13:30–13:50. Account T remains served throughout. Ask for distinct affected accounts, continuous customer episodes and customer-minutes.

R’s periods combine into 13:00–14:00, contributing sixty minutes. S contributes twenty. Two different accounts are affected, with two continuous customer episodes. Total customer-minutes are eighty. Over all three system accounts, the average is eighty divided by three, approximately 26.7 minutes per account.

Now add confirmed restoration for R from 13:40–13:45, and correct its second loss period to 13:45–14:00. R contributes forty plus fifteen, or fifty-five minutes, across two episodes. With S’s twenty minutes, the total becomes seventy-five. The distinct affected accounts stay at two while customer episodes rise to three.

Explain why the total fell while the episode count rose. The restored gap removed five minutes of service loss but split R’s experience into two periods. There is no contradiction because the measures describe different features.

For a final change, suppose S’s restoration is unknown at 13:50. You can no longer report its twenty minutes as a completed episode merely because the earlier exercise used that end time. Rewrite the conclusion as an observation through the stated cutoff and identify what remains unresolved. Transfer is demonstrated by revising the method when the evidence changes, rather than preserving the answer because the account letters look familiar.

Contents · Previous chapter · Next chapter

A learner who reports 300 rather than 240 customer-minutes in Cedar Quay may understand multiplication perfectly. The unstable step is treating fault-associated entries as disjoint customer-time. Repeating multiplication drills would miss the actual problem.

Begin with B alone. Draw the two sixty-minute periods, highlight their common half-hour and ask whether B can lose the same service twice during that same minute. Once the learner explains the overlap, return to the four-account table and let them reconstruct C independently.

A different learner may find 240 correctly but divide by four when asked for an average over twenty system accounts. Their difficulty concerns the denominator. Keep the valid numerator and repair the population choice. The correction should match the mistaken move rather than replace the whole solution.

An extension learner can reconcile both the customer and timeline methods, then introduce an unresolved endpoint or a changed episode convention. Ask which result remains valid under the change. This develops judgement about the method’s conditions instead of rewarding arithmetic speed alone.

Finish with a short explanation containing the measured quantity, calculation, comparison population and limitation. A strong answer can be concise because the distinctions are clear. The gold standard is a learner who can choose and defend a suitable counting method in another service record, including recognising when the available information does not yet support a completed total.

Contents · Previous chapter · Next chapter

Does merging overlapping periods remove a real fault?

No. Preserve the fault record for investigation. Merging changes the account of customer-time so that the same customer-minute is counted once. Fault history and service experience remain connected but distinct records.

Can I add customer durations together?

Yes, after each customer’s relevant intervals have been reconstructed under a consistent definition. Different customers losing the same minute each contribute to customer-minutes. The duplication problem concerns counting the same customer’s same service-loss minute more than once.

Is one long interruption the same as several short ones?

They can have equal total duration while differing in frequency and consequences. Keep the episode pattern alongside the accumulated duration when that distinction matters to the question.

Does a missing restoration time mean service stayed unavailable forever?

No. It means the record does not establish the final endpoint. Report confirmed accumulated time through a stated cutoff and preserve the ongoing or uncertain status. Do not invent either restoration or indefinite continuation.

Are these calculations official SAIDI or SAIFI figures?

No. They teach interval merging, identities and denominators through simplified fictional data. Official reliability reporting has defined customer populations, event criteria and other rules. Use the responsible authority’s current methodology for an actual submission or comparison.

What should I check before trusting a lower revised number?

Ask what changed: observed service, corrected records, duplicate removal, population, inclusion rule or method. The revision explanation should identify the changed evidence and show how it affects the calculation.

Contents · Previous chapter · Next chapter

For the service chain, return to The Gold Standard of Utilities. It connects sources, networks, customer records, maintenance and recovery. Follow Electricity Utility Companies for the power-service route and Water Utility Companies for the supply, quality and responsibility route.

For connected consequences, use Urban Systems. For a different utility question, the established Electricity Demand Management guide concerns demand and service conditions; interruption duration is not a substitute for that operating account.

Authoritative context comes from the EIA explanation of duration and frequency measures and Electric Power Annual reliability notes, reviewed on 21 October 2026. These sources support the distinction between measures and reporting scope. Every Cedar Quay calculation and transfer task is an original fictional teaching example.

Continue with the Gold Standard education parent and How X Works. A useful next exercise is to explain one corrected total in four sentences, preserving the original record, the overlap, the revised measure and its limit.

Contents · Previous chapter · Return to Utilities