The Cost of Waiting: The Number Nobody Counts

Monday, 08:15, at Redbrook Engineering. Lena, on the four-person desk, picks up ticket 4412. It was raised on Friday at 14:05 by Farah in sales: "Laptop will not start, black screen, fan runs." Lena has it working by 08:40, a loose memory module. The ticket's record is clear and short: twenty-five minutes of hands-on time, resolved first thing the next working day, well inside its target. Farah's record of the same ticket is different. She had a quote due to a customer at 09:00 on Monday. She spent the last two hours of Friday on a borrowed machine that did not have the quoting tool on it. She came in at 07:30 to try again and rang the desk's voicemail twice. She sent the quote at 10:15 with an apology. Her manager spent twenty minutes of Monday morning asking whether IT could "do anything". The desk's books show twenty-five minutes. Farah's show about four hours, her manager's another half, and nobody has written either of those down anywhere.

Waiting cost is the working time requesters lose while a ticket is open. These are hours in which they cannot do their job, do it more slowly, or spend chasing, working around and explaining. It is paid by the requester, by their manager, and sometimes by everyone downstream of them. It never appears in the desk's figures because the desk's figures are about the desk. This article is part of our The Economics of Support series. This article is about counting that cost honestly, in hours. Every number is a worked example from a fictional desk rather than a benchmark from anywhere.

It covers why the number is never counted, why open time is not the same as waiting time, and the three kinds of waiting. Next come the blocked process, where several people wait for one ticket, and the arithmetic in hours, from a sample. Then come why the result dwarfs the desk's own time and the worked example at a fictional desk. What shrinks the number, which is mostly not speed, and what never to do with it follow. A short war story and the waiting-cost worksheet to copy come last, in that order.

Why Nobody Counts It

Nobody counts waiting because it is paid by people who do not keep the books. The desk records its own minutes on a ticket, because the ticket system does that for it. The requester records nothing, because there is nowhere to record it and because, from where they sit, the lost morning was not a cost but a Monday. Finance sees the desk's salaries and the requester's salary. It never sees that one of them spent four hours producing nothing because the other spent twenty-five minutes on Monday rather than Friday. The cost is real. It is large. And it is distributed in small pieces across the whole organisation, which is the most reliable way for a cost to stay invisible.

The desk has a second reason not to count it, which is that the number is uncomfortable. A desk that counts waiting discovers that its own hours are the smaller part of the cost of support. The desk also discovers that the figure it has been reporting as its workload is a fraction of what the organisation spends on tickets. That feels like an admission. It is in fact the opposite. A desk that can show the waiting line is the only part of the organisation with any evidence for what fixing a ticket faster is worth. And it is the evidence that gets the second technician approved. The desk that does not count waiting is arguing for itself with one hand behind its back. (It is usually the hand holding the numbers.)

The last reason is that it is harder to count than the desk's time. And the difficulty is used as an excuse not to try. The desk's minutes are in the ticket. The requester's hours have to be estimated, from a sample, with a classification and a range. Estimated is not the same as invented. The cost map in Where the Money Actually Goes has a line for waiting that this article shows how to fill.

Open Time Is Not Waiting Time

Open time is the stretch from a ticket being raised to its being resolved. It is the figure the ticket system gives you. Waiting time is the part of that stretch in which the requester actually lost something. It is always shorter, often much shorter, and never the same for two tickets. Farah's ticket was open for about sixty-six hours by the clock on the wall and about seven by the clock that matters. That is because the weekend is not waiting. The hours after she gave up on Friday are only partly waiting, and the last of it was a manager's twenty minutes rather than her own. Count waiting in the requester's working hours only, and only for the hours in which the open ticket was costing them something. The honest number is smaller than the frightening one and large enough on its own.

Waiting comes in three kinds, and the classification is the whole method. The first is blocked: the requester cannot do their job, or the central part of it, until the ticket is resolved. A laptop that will not start, an account that is locked, a shared system that is down for one department, a new starter without a login. Every working hour the ticket is open is an hour lost, less whatever they manage to do by other means. The second is degraded: they can work, but slower, or around the problem, or with somebody else's help. A printer out on one floor, a missing shared drive they can reach by another path, a slow machine. The cost is a fraction of each open hour plus the time spent on the workaround. The third is merely waiting: nothing is lost but the ticket is open. A request for software they do not need until next month, a question that was curiosity, an upgrade they asked for in passing. The cost is reading the acknowledgement and perhaps one chase.

Classify from the ticket's content and the requester's situation, not from its priority or its type. A password reset is blocking for a person who cannot log in and merely waiting for a person who asked for a change while already logged in. A printer ticket is degraded for an office with a second printer and blocking for a dispatch bay that prints every label. The priority field tells you what the desk decided. The classification tells you what the requester lost. The two should agree more often than they do. That is a finding in itself and the subject of Making Business Impact Visible. For the waiting count, read the ticket, picture the requester's morning, and pick one of the three.

The Blocked Process: When Several People Wait at Once

A blocked process is a ticket whose waiting is paid by more than the person who raised it. And it is where the waiting cost stops being a requester's morning and becomes a department's week. The new starter is the plainest case. A login that takes two days to appear costs the starter two days. It costs their manager the hours spent finding them something to do. It costs the colleague who lends a machine and a password they should not lend. And it costs whatever the starter was hired to do, which nobody had thought to put on the ticket. One ticket, one requester, four people waiting. The ticket system shows one open ticket.

The multiplier is people affected times hours each, and the arithmetic is short once you remember to do it. At Redbrook, a new starter in engineering waited two working days for an account that needed three systems. That meant fifteen hours for the starter, about two for the manager, and about two for the colleague who improvised. That was roughly nineteen hours of waiting from one ticket whose handling time was ninety minutes spread over two days of chasing approvals. A dispatch bay scanner that failed at 10:00 and was replaced at 14:00 had three pickers keying by hand at perhaps two thirds of their normal pace. That is four hours times a third of three people, about four hours lost, plus the supervisor's half hour on the phone. Neither ticket was dramatic. Both were worth a technician's afternoon in somebody else's time.

Flag blocked-process tickets in the sample and count them separately, because they distort the per-ticket figure. They are also the tickets the desk should learn to recognise on arrival. A desk that knows which ticket types routinely block several people has found the tickets to work first, before any discussion of priority grids. It has also found the ticket types where a runbook or a pre-built account template pays back fastest. New starters, shared systems with one department depending on them, and anything in a bay, a ward or a till lane are the usual members of the list. Your own list is in your last quarter's tickets, waiting to be noticed.

The Arithmetic in Hours

The method is a sample, a classification, an estimate per class, and a scaling to the month, with a low and a high carried throughout. Pull forty resolved tickets from one ordinary month, chosen at random rather than by interest. The interesting tickets are the blocked ones and picking them inflates the result. For each ticket, read it and classify it as blocked, degraded or merely waiting. For each, estimate the requester's lost working hours as a low and a high figure. Use the open time in working hours as the ceiling for a blocked ticket and the table above as the guide for the others. Mark any ticket that blocked more than one person and note how many. Then add up the sample, scale to the month's ticket count, and report both ends.

Count only working hours, on the requester's schedule, and say so. A ticket raised at 16:50 on Friday and resolved at 08:40 on Monday was open for sixty-four hours and blocked its requester for about two. That is because the requester went home. If your tool, as QuerySolv does, runs its service-level clock on a business-hours schedule, the same schedule is the right one for the waiting count. If it does not, subtract the nights and weekends by hand for the blocked tickets. That is tedious for forty tickets and impossible for four hundred, which is why the method is a sample. The event history on a ticket also shows when it was waiting on the requester rather than on the desk. Those stretches come off too. A requester who took a day to answer a question was not waiting for the desk during it.

Use the sample to estimate, never to prove. Forty tickets give a range with honest width. They do not give a figure to two decimal places, and a report that quotes one has mistaken its arithmetic for its evidence. Redbrook's sample came out at twelve blocked, sixteen degraded and twelve merely waiting. The ranges Priya used were three and a half to six hours for a blocked ticket. For a degraded one, she used half an hour to an hour and a half. For the rest, she used nothing to a quarter of an hour. Those are her ranges for her desk, from reading her tickets. Yours come from yours, and will differ. The one thing that will not differ is the shape of the result, which is the subject of the next section.

Why It Dwarfs the Desk's Own Time

The waiting line is larger than the desk's handling line at every desk I have ever counted it for. And the reason is arithmetic rather than failure. A ticket's handling time is minutes: twenty-five for Farah's laptop, ten for a reset, forty for a drive. Its open time is hours or days, because the desk has a queue and a ticket waits its turn. For a blocked requester every one of those hours is lost. A desk that resolves a blocking ticket in twenty-five minutes of work after five hours in the queue has spent twenty-five minutes and cost five hours. Multiply that ratio by every blocked ticket in a month and the waiting line is several times the handling line before anyone has done anything wrong.

There are also more requesters than technicians, which is the second half of the arithmetic. Four technicians at Redbrook support about three hundred and fifty people. When a shared system is down for one department of thirty for an hour, the desk spends one technician-hour and the organisation loses up to thirty. The desk's time is the narrow end of the funnel and the requesters' time is the wide end. And the cost of a ticket is mostly at the wide end. This is why a faster desk is worth more than its own hours would suggest. It is also why the argument for the second technician at a one-person desk is almost never the technician's own workload. It is the hours at the wide end, and only the waiting count shows them. The time to resolution figure and what it does and does not measure has its own article in What Time to Resolution Actually Measures. Waiting cost is that figure weighted by what each hour of it cost someone.

The Worked Example at a Fictional Desk

Here is Redbrook's month, from the sample of forty tickets described above, scaled to the four hundred and twenty tickets the desk handled. Every number in it is a worked example rather than anything to copy. Twelve blocked tickets at three and a half to six hours each is forty-two to seventy-two hours in the sample. Sixteen degraded tickets at half an hour to an hour and a half is eight to twenty-four. Twelve merely waiting at nothing to a quarter of an hour is nothing to three. The sample total is fifty to ninety-nine hours over forty tickets. The month is ten and a half times the sample. So the waiting line is roughly five hundred and twenty-five to a thousand and forty hours. Priya wrote that as "roughly five hundred to a thousand requester hours a month" on the cost map and in every sentence that followed.

Set the result beside the desk's own line and the shape appears. The desk spent about three hundred and ten hours handling tickets that month. The requesters spent between five hundred and a thousand waiting for them. At the cautious end the waiting line is more than one and a half times the handling line. At the other end it is more than three. Three of the blocked tickets in the sample affected more than one person. Those three carried about a third of the blocked hours between them. That told Priya more about where to look than the total did. One of the three was a new starter. All three, when she went back through the quarter, were types the desk could have recognised on arrival.

What Shrinks It

Order of work shrinks waiting more than speed of work does, and this is the finding that changes how a desk runs its morning. A blocked ticket waiting five hours in a queue behind four degraded ones costs five hours of someone's day. The same ticket worked first costs twenty-five minutes of waiting. The four degraded tickets each wait a little longer, at a fraction of the rate. The handling hours are identical. The waiting hours are not close. A desk that picks up blocked tickets before degraded ones, and degraded before merely waiting, has cut its waiting line without working any faster. That is the whole case for triage by impact rather than by arrival. The age of a ticket is part of the picture too, and what the age figure actually measures is in What Ticket Age Actually Measures.

The honest estimate shrinks it next, because a requester who knows the fix is tomorrow stops waiting and starts working around. Farah lost two hours on Friday afternoon partly because the acknowledgement said nothing about when anyone would look at her ticket. So she kept trying, kept checking, and kept the quote open on a machine that could not produce it. A reply at 14:30 could have said "we will be with you first thing Monday. If you need the quoting tool today, the loan laptop on the third floor has it". That would have cost the desk three minutes and saved her most of the afternoon. The waiting hours that come from not knowing are the cheapest to remove. And the craft of removing them is the subject of Honest ETAs. A blocked person who has a workaround is a degraded person, and the difference between the two classes is most of the waiting line.

The loan pool, the pre-built account and the runbook for the blocked types shrink it last and most permanently. A spare laptop with the common tools on it turns a blocked laptop ticket into a degraded one inside ten minutes. A new-starter account created from a template the day before the start date turns a nineteen-hour blocked process into none. A runbook for the shared system that one department depends on turns an hour's outage into twenty minutes. None of these make the desk faster at the ticket. All of them make the requester wait less, which was the point. The desk's hours barely move. The waiting line falls by hundreds of hours, and nobody notices, because waiting is the number nobody counts, until you start.

What Never to Do With the Number

Never add it to the desk's hours, never convert it to money, and never aim it at a person. The first produces a figure nobody can check. The second hands the argument to whoever owns the hourly rate, and that is never the desk. The third is a misreading of what the number is. The waiting line is a property of the queue, the order of work, the information given to requesters and the desk's capacity. It is not a property of Lena, who fixed the laptop in twenty-five minutes on the first working morning she had it. If a waiting figure is ever cut by technician, it is for coaching on order of work, never for ranking. In that case, it comes from at least a hundred of that technician's tickets, and it stays inside the desk.

Never quote it without its range and its method, and never quote it from calendar time. "Requesters waited roughly five hundred to a thousand hours last month, estimated from a random sample of forty tickets classified as blocked, degraded or merely waiting, in working hours" survives any question a finance reader will ask. "Users lost two thousand hours" does not survive the first, which is "how do you know". The honest answer, that you counted weekends, ends the conversation. The range is not a hedge. It is a description of what a sample can and cannot tell you, and readers who approve budgets have met plenty of people who did not know the difference.

And never let it become a reason to close tickets faster than they can be fixed. A ticket closed before the fault is gone reopens, and the requester waits through both attempts. So the waiting line goes up, not down, when the desk rushes. The way to shrink it is the order of work, the honest reply and the prepared fix. And all three take a little longer at the ticket in exchange for a great deal less time at the wide end of the funnel. The waiting count rewards the desk that does the job properly, in the right order, and says what it is doing. It is the only support figure I know that does.

A Short War Story: The Dispensary Door at Harbourview Clinic

At Harbourview Clinic, the badge reader on the dispensary door failed intermittently for about two months. Each time it did, the desk received a ticket and classified it as low priority. That was because it concerned one door and nobody was locked out of anything they could not reach by asking. The desk fixed it within a day or two by re-seating a connector, for about twenty minutes of handling each time. A ward manager called Grace kept a tally for one week of what the door was costing her nurses. There were eleven of them, each waiting at the door or walking round to be let in several times a shift. It cost roughly twenty-five minutes a nurse a day. That came to about four and a half hours of nursing time a day and something close to ninety hours in a month. Compare that with perhaps two hours of the desk's time in the same month. She took the tally to the desk lead, Oluwaseun, who had the reader replaced that week. Oluwaseun then went back through a quarter of low-priority tickets looking for anything else that was cheap for the desk and expensive for a ward. Oluwaseun found four. A ticket that costs the desk twenty minutes can cost everyone else a working week. And the desk's own figures will never show which ones do.

The Waiting-Cost Worksheet

The sheet below is the method as a form, one copy per month. And it takes an afternoon the first time and about two hours after that. Section A sets up the sample and the rules, including the working-hours schedule, so that two people filling it in get the same number. Section B is the classification and the per-class ranges, filled in by reading the tickets. Section C flags the blocked-process tickets, which are the ones to learn from. Section D is the scaling and the two sentences that leave the desk. Section E is the checklist that keeps the figure on its own line. Keep the sheets. The second month's figure beside the first is the trend, and the trend is what shows whether the order-of-work change worked.

THE WAITING-COST WORKSHEET — ONE MONTH
Desk: ______________________   Month: ____________   Prepared by: ______________
Working-hours schedule used for all counting: ______________________________

A. THE SAMPLE
   A1 Tickets handled in the month ................................ ____
   A2 Sample size (random, not chosen) ............................ ____
   A3 Scaling factor (A1 over A2) .................................. ____
   A4 Stretches waiting on the requester excluded? ................ yes / no

B. CLASSIFICATION AND RANGES (from reading each sampled ticket)
                          tickets     lost hours each      sample hours
                                      low      high        low      high
   B1 Blocked ............. ____      ____     ____        ____     ____
   B2 Degraded ............ ____      ____     ____        ____     ____
   B3 Merely waiting ...... ____      ____     ____        ____     ____
   B4 Sample total ........ ____                           ____     ____

C. BLOCKED-PROCESS TICKETS (more than one person waiting)
   Ticket ________  type ______________  people affected ____  hours ____
   Ticket ________  type ______________  people affected ____  hours ____
   Ticket ________  type ______________  people affected ____  hours ____
   C1 Ticket types that routinely block several people ........... ______________________

D. THE RESULT
   D1 Waiting hours for the month, low (B4 low x A3) .............. ____ h
   D2 Waiting hours for the month, high (B4 high x A3) ............ ____ h
   D3 The desk's own handling hours, for comparison only .......... ____ h
   D4 Sentence one: "Requesters waited roughly ____ to ____ hours this month,
      estimated from a random sample of ____ tickets, in working hours."
   D5 Sentence two: "The ticket types that blocked several people were
      ______________________; these are now worked first."

E. BEFORE THIS LEAVES THE DESK
   [ ] Sample was random; blocked tickets were not picked on purpose
   [ ] Working hours only, on the requester's schedule
   [ ] Both ends of the range are quoted, with the sample size
   [ ] D1 and D2 appear on their own line, never added to D3
   [ ] Nothing in this sheet is cut by technician
   [ ] No currency anywhere on the sheet

Priya's first sheet took the afternoon and produced the "five hundred to a thousand" line that went on the cost map. Her second came a month after the desk started working blocked tickets first and replying with a time and a workaround. It came out between three hundred and six hundred and fifty, with the same handling hours. Nobody on the desk had worked faster. Farah's next laptop ticket, raised on a Thursday, was answered in eleven minutes with the location of the loan machine. She sent that quote on time. That is not a number on any report and is the whole of what the number was for.

Frequently Asked Questions