Measuring the Cracks: The Monthly Check for Handoff Gaps

Thursday, 15:20, in the quarterly review at Dunmore Freight. Ines runs the five-person desk and has brought the usual page: tickets in, tickets out, first-response rate, resolution time, reopens, all of them fine. The operations director reads it, nods, and asks a question that is not on the page. "How often do tickets get dropped when someone's off, or when they get passed to someone else?" Ines says "rarely". The director says "how rarely?" Ines has a number for everything the desk does. She has no number for the thing the desk stops doing, because nothing on the page measures a stop.

Measuring the cracks means counting, each month, the tickets that went quiet after a handoff, the tickets that changed hands too often, and the time tickets spent with nobody. Those three numbers, and a fourth for tickets held by people who were away, are the only way to see a crack from inside the desk. That is because a crack is an interval in which nothing happens. Every ordinary report counts things that did. This article is part of our Handoffs & Coverage Transitions series. And it is the last one, because the measures are what tell you whether the other five articles are working.

It covers why the queue cannot show you a crack. Next come the four measures in turn, what each counts and how to define it. Then we cover getting the numbers out of an ordinary tool, with a worked month. We include a table of the measures and what each points to. Then we cover reading the numbers by handoff moment and the monthly check, in fifteen minutes. We cover what not to do with the numbers, a short war story, and the monthly cracks check to copy, in that order.

You Cannot See a Crack from Inside the Queue

The queue shows the state of each ticket now, and a crack is a transition that went wrong in the past. Consider a ticket that sat unassigned for two days after a transfer, and was then claimed and fixed. It looks in the queue exactly like a ticket that was claimed in an hour. It is closed, it has an owner, and its resolution time is a single number that contains the two days without pointing at them. The resolution report will say "four days". It will not say "two of them were nobody's".

The ordinary reports are built from the ticket's fields, and the cracks live in the ticket's history. Most tools keep a history, sometimes called an audit log or an activity trail, that records each change of owner, department and status with a time. That trail is where the handoffs are, and the measures below are all built from it. If your tool shows the history on each ticket but cannot export it, the measures can still be taken by hand on a sample. The section on getting the numbers says how. If it does not keep a history at all, the measures can be approximated from the last-update time, which is cruder and still better than "rarely".

The point of measuring is not to find who dropped what. It is to find which of the five handoff moments is dropping tickets on your desk. That way, the fix goes to the moment and not to a person. The five moments are the shift end, leave, the transfer, the escalation and the reassignment. Those moments, and the anatomy of the gap each one opens, are in Where Tickets Fall Between the Cracks. The numbers here are the moments made countable.

Measure One: Silent After a Handoff

A ticket is silent after a handoff when its owner or department changed and the requester heard nothing from a person within one working day of the change. The definition has three parts, and each is chosen on purpose. The owner or department changed, because that is what a handoff is in the tool. The requester heard nothing, meaning no public reply, because an internal note is the desk talking to itself. One working day, because a handoff that has not produced a word to the requester by then has usually not produced any work either.

Count it per handoff, not per ticket. A ticket with three handoffs can be silent after one of them and fine after the others. And the moment that went silent is the information. For each change of owner or department in the month's history, look for the next public reply from the desk and read its time. If it is more than one working day after the change, or there is none, that handoff was silent. The first assignment at triage does not count; it is not a handoff, it is the start.

Where you cannot read the history, approximate. Take every ticket that changed owner or department in the month, which most tools can filter or at least sort. Read each ticket’s last-update time against the change. It will overcount tickets where the new owner worked quietly and undercount tickets where the requester's own reply moved the time. It is still a number, and a number that moves month to month, which is all a measure needs to be.

Measure Two: Reassignment Count

The reassignment count is the number of times a ticket's owner or department changed after triage. Most tickets have none. Some have one, which is the transfer or escalation they needed. A ticket with three or more has been passed around, and the question is whether the three were a sensible route or a ping-pong. Count the distribution for the month: how many tickets had zero, one, two, three or more. The last bucket is the one to read ticket by ticket, because there are never many and each one has a story.

Read the three-plus tickets for a pattern of return. A ticket that went from the desk to Facilities and back to the desk is a bounce. And a bounce means a transfer that was declined without being settled. A ticket that went desk, Facilities, payments, Facilities is a dispute, and a dispute means nobody was asked to decide. Both point at the same fix, the acceptance and the two-bounces rule, which are in Transferring a Ticket Between People or Departments. A ticket that went desk, lead, specialist, done is three changes and no fault; it is the escalation path working. The count flags; the reading decides.

Do not count the triager's first assignment, and do not count a claim from the unassigned queue as a reassignment. The number is meant to show tickets that moved after someone owned them. A desk where every ticket is assigned by a lead and then claimed will otherwise show every ticket with a count of two. In that case, the measure becomes noise. Define it once, write the definition on the check sheet, and count the same way every month.

Measure Three: Time Spent Unassigned

Time unassigned is the total working time a ticket spent with no owner after it was first triaged. A ticket is unassigned when it is sitting in a department's queue waiting for someone to claim it. That is the normal state of a ticket immediately after a transfer and the abnormal state of one two days later. Measure it per ticket from the history: each interval between an owner being cleared and an owner being set, in working hours, summed. Then take the median and the worst for the month, and the count of tickets over one working day.

The median says how long a transfer normally waits for an acceptor. The worst says what happens when the acceptor is away. The count over a day says how many requesters waited for nobody for longer than a day. That is the number the operations director was asking about, more or less. Break it down by the department the ticket was unassigned in. The number is almost always a property of one or two departments and not of the desk. In those cases, the department has no named acceptor, or an acceptor who reads the queue on Tuesdays.

If the history cannot be exported, the unassigned view itself gives a living version of the same number. Each morning, note the age of the oldest unassigned ticket in each department before the sweep. The month's list of those ages is a rough but honest picture of time unassigned. And the department whose number is always biggest is the department to visit.

Measure Four: Held by Someone Away

The fourth measure is the count of tickets that were owned by an absent person for more than one working day of their absence. It needs the absence list for the month beside the ticket export. It also needs a look at each absence: who was away, which days, and which active tickets carried their name on those days. Most months the number is small and the tickets are the same few. They are the ones left behind by a pre-leave sweep that did not quite reach zero, or by an unplanned absence on a day the lead did not open the list.

This is the measure that tests the routine in Vacation Coverage Without the Pile. And it is the one with the clearest target, which is zero. A ticket owned by an absent person for a day is a ticket doing nothing visibly. Three of them in a month is a sweep that is being skipped. Three of them belonging to the same person is a conversation. Three of them in a month when nobody was on planned leave is the unplanned-absence check not happening by its fixed time. Each of those is a different fix, and the measure says which.

Getting the Numbers from an Ordinary Tool

Export the month's tickets with their history, or the nearest thing your tool offers. The columns needed are the ticket number, the time of each change, what changed (owner, department or status), the old and new values, and the time of each public reply. Many tools export the history as one row per event. Some export only the current fields. In that case, you get the reassignment count from a "number of assignments" column if there is one, and the rest from the sample method below. Whatever comes out goes into a spreadsheet. The spreadsheet does the counting with sorts and a working-days function. Most spreadsheet tools have one that counts working days between two times.

Here is Dunmore's first month. The desk opened 312 tickets. Seventy-one had at least one handoff after triage. Of those handoffs, nineteen were silent, with no public reply within a working day. Fourteen of the nineteen were transfers to the warehouse-systems team. Nine tickets had three or more changes; five of them were bounces between the desk and warehouse-systems. Median time unassigned was four working hours. The worst was three working days, in warehouse-systems. Eleven tickets waited unassigned for more than a day, nine of them in warehouse-systems. Six tickets were held by someone away for more than a day. Four of them were held by one technician who had been off sick for a week in which nobody opened his list. Ines wrote the whole thing on half a page, and every number pointed at the same two places.

When there is no export, count a sample by hand. Take twenty tickets that changed owner or department in the month, chosen from across the month rather than from the top of the list. Open each one's history, and tally the four measures with a pen. It takes about thirty minutes. Twenty is enough to find a department whose transfers go silent. Such a department shows up in a sample of twenty as four or five tickets in a row. Do it the same way each month and the trend is real even if the total is a guess. A measured guess beats "rarely" on every page it is ever written on.

Reading the Numbers by Moment

Each number points at one of the five moments, and the moment points at the fix. Silent handoffs that cluster at the end of a shift or the start of the next day are the handover: promises made at 14:00 and nobody told. Silent handoffs that cluster after transfers to one department are that department's missing acceptor. Silent handoffs after escalation are the lead's acceptance time. The measures for that path, including how long an escalation waits before it is picked up, are in Measuring Escalations. Silent handoffs after a reassignment made by a lead are reassignments made without the sentence to the new owner.

Time unassigned is always about the transfer moment, and always about a department, and the breakdown by department says which. Held by someone away is always about leave. The breakdown by person and week says whether it was planned leave with a skipped sweep or unplanned absence with a skipped check. The reassignment count is about transfers when the pattern is a bounce and about escalation when the pattern is a ladder. The ticket-by-ticket read of the three-plus bucket tells them apart in ten minutes.

The numbers are at their most useful read against the crack map and the coverage matrix. Dunmore's fourteen silent transfers to warehouse-systems were a blank cell on the crack map: no acceptor named for that department. The four tickets held by one sick technician were a blank on the leave rows: nobody assigned to open absent people's lists. Both blanks had been visible on the map for two months. The numbers did not find anything new. They made the blank cells cost something, which is what finally got them filled in.

The Monthly Check in Fifteen Minutes

The check is fifteen minutes on the first working day of the month, with the export open and the sheet beside it. Minutes one to five: the four numbers, written next to the previous two months'. Six to nine: the three-plus tickets and the silent handoffs, read by moment and by department. Ten to twelve: the worst single case, read in full, because the worst case is where the lesson is plainest. Thirteen to fourteen: one change, with an owner and a date, and the number it is expected to move. Fifteen: file the sheet.

One change, and one is the right number. Dunmore's first change was a named acceptor in warehouse-systems, reading the unassigned view by 09:30 every day. The following month, silent transfers to that department went from fourteen to three. Time unassigned over a day went from eleven tickets to two. Nothing else had changed, so Ines knew why. The second month's change was the absent-list check at 09:30, owned by whoever was first in. Two changes in two months, each measured, and the director's question had an answer with a trend on it.

Write one sentence for the director at the bottom of each sheet, in plain words, with the numbers in it. "In March, nineteen handoffs went silent for over a day, mostly transfers to warehouse-systems; a named acceptor there starts on the first, and April should show under ten." The sentence is the whole value of the measure to anyone outside the desk. And it is the sentence Ines did not have on the Thursday. The aging report, which reads the same tickets by how long they have waited regardless of handoffs, is the companion to this check. The report is explained in Reading an Aging Report. The two together say both how long tickets wait and where in their lives the waiting happens.

What Not to Do with the Numbers

Do not rank people by them. A technician with a high reassignment count is usually doing triage, which is the job of assigning tickets to other people. A technician with many silent handoffs on their tickets is usually the one who transfers to the department with no acceptor, which is not their doing. The numbers describe transitions, and transitions have two sides and a system around them. A leaderboard of cracks will produce a desk where nobody transfers anything and every ticket is held by whoever opened it. That is the one arrangement guaranteed to make the cracks wider and invisible at the same time.

Do not discourage handoffs to improve the count. A desk with a low reassignment count can be a desk that routes well or a desk where people hold tickets they cannot fix. And the numbers alone cannot tell which. Read them with the resolution time and the reopen rate beside them. A falling reassignment count and a rising resolution time is a desk that has stopped transferring. And that is a worse month than the one before, however the cracks sheet looks.

Do not chase zero on everything. Silent handoffs and time unassigned will never be zero on a desk that does real work, because some handoffs happen at 16:55 and some acceptors are ill. Held by someone away can be zero, and should be. For the other three the target is a number that is small, stable and explainable. A month in which it jumps is a month with a story. The ticket-by-ticket read will find it. The sheet is a diagnostic, and diagnostics are for finding the next fix, not for passing.

A Short War Story: The First Count at Westbrook Dental Group

Westbrook Dental Group's three-person desk supported eleven practices. Its lead, Callum, ran the cracks check for the first time after a practice manager complained. Her appointment-book integration had been broken for eleven days with "no one answering". The ticket had been transferred to the practice-systems team on day one. There it sat unassigned for nine days. That was because the team's only acceptor was a part-time contractor who read the queue on the days he was in. That fortnight, he was in on none of them. The first month's count found twenty-three silent handoffs, fourteen of them transfers into practice-systems. It found a median time unassigned there of two working days against four hours everywhere else. Callum had suspected the department for a year and had never had a number. With one, he got a named acceptor in practice-systems inside a week. The desk's own early-shift technician agreed to read that unassigned view on the contractor's days off. By the third month silent handoffs were down to four and the practice manager had stopped phoning. The crack had been open for a year. And it took one afternoon of counting to close, because a number is harder to argue with than a feeling.

The Monthly Cracks Check Sheet

The sheet below is the four measures, the breakdowns, the one change and the sentence for the director, on one page. Fill it in from the export or the twenty-ticket sample. Keep the completed sheets together, and read the last three before deciding the change. The definitions are printed on the sheet so that the count is the same every month, whoever does it.

MONTHLY CRACKS CHECK — ONE PAGE
Desk: ______________   Month: ______________   Counted by: ____________
Source:  [ ] full history export   [ ] sample of 20 handed-off tickets

DEFINITIONS (count the same way every month)
  Handoff = owner or department change after triage; first assignment and claims do not count
  Silent  = no public reply to the requester within one working day of a handoff
  Unassigned = no owner, after triage, measured in working hours
  Held by someone away = owned by an absent person for more than one working day

THE FOUR NUMBERS                    this month   last month   month before
  Handoffs in the month               ______       ______       ______
  Silent after a handoff              ______       ______       ______
  Tickets with 3+ changes             ______       ______       ______
  Time unassigned: median             ______       ______       ______
  Time unassigned: worst              ______       ______       ______
  Unassigned over one working day     ______       ______       ______
  Held by someone away over a day     ______       ______       ______

BY MOMENT (silent handoffs, read by hand)
  After shift end ____   After transfer ____   After escalation ____   After reassignment ____
  Transfers by receiving department:  __________ ____   __________ ____   __________ ____

THE THREE-PLUS TICKETS (read each one)
  #______ route: ______________________  [ ] bounce  [ ] dispute  [ ] ladder, fine
  #______ route: ______________________  [ ] bounce  [ ] dispute  [ ] ladder, fine
  #______ route: ______________________  [ ] bounce  [ ] dispute  [ ] ladder, fine

HELD BY SOMEONE AWAY
  #______ owner ________ absent ______ to ______   [ ] planned, sweep missed   [ ] unplanned, check missed
  #______ owner ________ absent ______ to ______   [ ] planned, sweep missed   [ ] unplanned, check missed

WORST SINGLE CASE
  #______  what happened: __________________________________________________
  Which blank on the crack map or coverage matrix: ___________________________

THE ONE CHANGE
  What: ___________________________________________   Owner: ________   By: ________
  Number expected to move: ______________  from ______ to about ______

ONE SENTENCE FOR MANAGEMENT
  ___________________________________________________________________________

Two lines on the sheet earn their keep more than the rest. One is the definitions block, because a measure that is counted differently each month is a measure of the counter. The other is the sentence for management, because it is the line that turns the numbers into a decision somebody else can make. Ines's sentence at the next quarterly review had three numbers and a trend in it. The director read it and moved on to the next page. That is the best thing that can happen to a number in a meeting. The cracks had not gone. They had become ordinary, which on a desk is the same as fixed.

Frequently Asked Questions