The Numbers to Show and the Numbers to Keep

Tuesday, 11:40, at Harbourview Clinic. Imran, the practice manager, has walked through the desk's office on his way to somewhere else and stopped at the wall screen. Nadia, the desk lead, put it up last month so the team could see the queue. The screen has a column she added as an afterthought: tickets closed this month, by name. Aisha, 91. Tom, 74. Ben, 38. Imran reads it, nods, and says, "Should I be worried about Ben?" Ben is not in the room. Ben covers the two satellite surgeries and drives to them twice a week. He takes every ticket that involves a clinical device, which is to say every ticket that takes a day. His thirty-eight were the hardest thirty-eight on the desk. The number on the screen was true, the question it produced was wrong, and the number produced it all by itself.

The audience matrix is a table of every number the desk produces, who sees it, in what form, and what it is for. It exists because a help desk makes far more numbers than any one reader needs. Also, the same number can be a tool in one person's hands and a verdict in another's. Tickets closed per technician is a coaching figure for the desk lead and a ranking for anyone else. Nothing about the number changes between the two readers except the damage it does. This article is part of our Reporting to Leadership series. It is about deciding, once, which numbers go up, which go sideways to department heads, which stay on the desk, and which belong to one person alone.

It starts with the four audiences and the question each is asking. It then covers what leadership needs, what department heads need, and what the desk needs. Next are the per-technician numbers and the rules for them, followed by numbers that change audience. After that comes what to say when leadership asks for a kept number. It ends with a short war story and the audience matrix to copy.

Four Audiences, Four Questions

Every number the desk produces has an audience, and there are four of them, each asking a different question. Leadership, the people who fund the desk and can change what it has, are asking "is anything going wrong that I need to decide about?". Department heads, the people whose staff raise the tickets, are asking "how are my people being looked after?". The desk itself, meaning the lead and the team together, is asking "are we running well, and where do we need to change something?". And each technician, alone, is asking "how am I doing, and what is on my plate today?". The same month of tickets answers all four questions, and no single report can. Each audience needs a different cut, at a different level of detail, with a different amount of context attached.

The matrix starts from the questions, not the numbers. Take each number the tool can produce and ask which of the four questions it answers. If it answers none, it is a number the tool produces because it can, and it goes nowhere. Most numbers answer exactly one question. A few answer two, at different levels. The median time to resolve is a leadership number for the whole desk and a coaching number per person. The matrix records both uses with the different rules each carries. The rare number that seems to answer all four is usually a number that answers none of them well. The usual example is tickets closed. It looks like productivity, service, capacity and effort at once and measures only how many times someone pressed a button.

The matrix is written down because the decision about audience is made once and then forgotten, and the forgetting is where the harm happens. Nadia did not decide to show Imran a ranking. She decided the team should see its own counts, which was reasonable, and the screen was on a wall, which was convenient, and Imran walks through that office. Three reasonable decisions produced a column that nobody had chosen to show him. A matrix with "per-technician counts: desk only, never on a shared screen" in it would have caught the third decision at the first. Numbers do not stay where you put them. They stay where you wrote down that they go.

What Leadership Needs

Leadership gets whole-desk numbers, a handful of them, each with last month beside it, and nothing cut by person. The five that fit a page start with tickets in and closed, then the backlog and how much of it is older than thirty days. Next are the median time to resolve and satisfaction as good ratings over replies. Last is the month's biggest interruption as people times hours. The page itself, and the reasons for each of the five, are in The Monthly Report They Will Actually Read. The reason the list is short is not that leaders cannot cope with more. It is that each number on the page costs a question. A leader has about two minutes' worth of questions for the desk each month. The desk should spend them on the numbers that can lead to a decision.

Whole-desk means exactly that: the number for everyone the desk serves, for the whole month, with no split by technician. There is no split by department unless the split is the point of the sentence. "Backlog grew to 138, almost all of it in tickets waiting on the housing supplier" is a whole-desk number with a cause attached. "Backlog by department: finance 31, operations 44, housing 63" is a table that invites the reader to find a villain. They will, and it will be the department with the most staff. The cut by department belongs with the department head, in a conversation, as the next section describes.

What leadership never gets is any number with a technician's name beside it. Not tickets closed, not time to resolve, not satisfaction, not even in a flattering month, not even when the technician would be pleased. The reason is in the opening scene. A per-person number read by someone who does not know the person's work becomes a judgement of the person, instantly and without malice. The desk lead then spends a month defending a colleague against a column. Leadership gets a sentence instead, when one is needed: "everyone on the team is carrying a normal load, and Ben's tickets are the long ones because he covers the clinical devices". That is the whole of what they need to know about the people, and it is said in words, by the person who knows them.

What Department Heads Need

A department head is asking a narrower question than leadership and a more personal one: how are my people being served, and is there anything in my department that the desk keeps seeing? They get their own slice of the whole-desk numbers. That includes requests from their staff this quarter, and how many of their people are waiting on the desk right now and for how long. It also includes the biggest interruption their department suffered. Where the desk has service-level targets, they also get "kept the promise on all but N of your requests", because the promise was made to them. If you run QuerySolv, the department is already a durable thing on the ticket, with its own mailbox and form and service levels. So the slice is a filter rather than a reconstruction. In any tool, the slice needs a department field on every ticket that is filled in at creation. The matrix should say where that field comes from.

The slice is delivered in a conversation, quarterly, not as a report. The numbers in it only mean something with the context the department head can supply and the desk cannot. "Your department raised 120 requests this quarter, forty of them about the expenses system" is a fact. The head of finance knows that the expenses system was replaced in the second month and that the forty are people learning it. Or they know it was not and the forty are a fault. The desk learns which in the conversation, and that is usually the most useful thing the desk learns all quarter. The shape of that conversation, and the manager round that goes with it, is in Beyond the Score.

Two numbers are kept back from department heads, for the same reason that per-technician numbers are kept back from leadership. The first is the list of their heaviest requesters by name. "Three people in your team raised half of your tickets" is a sentence about the system, usually a system those three people depend on more than their colleagues do. It is read by their manager as a sentence about three people. The desk reads the per-user pattern to find the system; the method is in Per-Department and Per-User Patterns, and the result goes to the head as "the expenses system generates most of your tickets", not as names. The second is satisfaction by individual requester, which is a private reply from a person and stays that way.

What the Desk Needs

The desk, read together as a team, needs everything the tool can produce about the whole desk and none of it about individuals. It reads that set once a month as a scorecard. The scorecard has all the whole-desk numbers with their honest pairs: speed with reopens, volume with backlog, satisfaction with its reply count. It also has the cuts that find patterns: tickets by category, by channel, by day of the week, by hour. Add the aging list, the oldest ten tickets by name, and the reopen list. Include the escalation count and where the escalations went. The pairs, and the fifteen-minute monthly read of them, are the subject of Reading Metrics Together, and the scorecard is the document the leadership page is cut down from.

The desk also has daily numbers that are not reports at all: the queue as it stands, what is unassigned, what is overdue, what is waiting on someone outside. These are a dashboard's job, and a dashboard shows now rather than a month. That is why none of them go up to leadership and why a screenshot of one is not a report. The team reads the queue together each morning and the scorecard together each month. The two readings have different purposes. The queue is for deciding what to do today, the scorecard for deciding what to change this quarter. A number that appears on both is being used for two things and the matrix should say which.

The scorecard stays on the desk, and it is not a secret. If a leader asks to see it, they can; the point of keeping it on the desk is not concealment but audience. A scorecard with twenty numbers and six cuts, handed to someone with two minutes, is not read. A report that is not read teaches its reader that the desk's documents can be skipped. Show the page, keep the scorecard, and offer the scorecard to anyone who asks what is behind the page. Nobody asks twice. (The scorecard is forty rows long. It is not a document that invites a second visit.)

The Per-Technician Numbers

Per-technician numbers exist for one purpose, which is coaching. The rule that follows is that they are read by the desk lead with the technician, in a conversation, and by nobody else. Coaching means looking at one person's tickets to find what would make their work easier or better. Look for the category they are slow on because nobody showed them the runbook. Look for the reopen pattern that says they close before confirming, or the satisfaction comment that says they are brilliant and brusque. Every one of those findings is useful to the person and useless to a ranking. A ranking is what the same numbers become the moment they are shown side by side to anyone who was not in the conversation.

Before any per-person number is read at all, there is a minimum sample, and the matrix states it. A technician's median time to resolve over twelve tickets is twelve tickets' worth of luck; over fifty it begins to say something about the technician. Harbourview's rule is a worked example from a fictional desk and not a standard. It requires thirty resolved tickets in the window before a per-person time figure is looked at. It requires twenty survey replies before a per-person satisfaction figure is looked at. The window is widened to a quarter where a month does not reach those counts. A smaller desk waits longer rather than lowering the bar. Below the sample, the number is not read, not "read with caution", because a number read with caution is still a number that has been read. The small-numbers problem and the technician who scores low because they get the hard tickets are worked through in Satisfaction by Category and, Carefully, by Agent.

The number is always read with the mix of work beside it, because the mix explains most of the difference between people. Ben's thirty-eight tickets against Aisha's ninety-one is a fact about who takes the clinical devices and who takes the password resets. A per-person figure without the category split beside it is a measurement of the routing, not the person. The coaching conversation starts with "what did you get this month?" and only then "how did it go?". A desk that shows per-person numbers without the mix has built a leaderboard, whether or not it calls it one. The damage leaderboards do, and the public-versus-private rule for every number on a screen, is the subject of Dashboards That Motivate Instead of Surveil.

Each technician's own numbers are, first of all, their own. They should be able to see them, privately, whenever they want, in the same way they see their own queue. Then the monthly conversation holds no surprises. A technician who first learns their reopen rate from their manager has been ambushed by a report. Most tools give each technician a view of their own work and their own figures, which is the right default. The desk lead's rule is simply never to put two people's views on one screen. What leadership gets instead of per-person numbers is one sentence from the lead about the team, and a team total. That is the deal: the lead carries the individual numbers so that nobody above has to. In return the lead is believed when they say the team is fine.

• Per-technician numbers are read for coaching, in a conversation with the technician, and for nothing else.
• No per-person time or quality figure is read below the minimum sample written in the matrix; widen the window rather than lower the bar.
• Every per-person number is read with that person's mix of work beside it.
• Each technician can see their own numbers privately at any time; two people's numbers never appear on one screen.
• Leadership and department heads get a team total and a sentence from the lead, never a name beside a number.

Numbers That Change Audience

A number moves up the matrix when it starts driving a decision, and the commonest example is the thirty-day count. On a desk with four old tickets, the count is a scorecard number the team reads monthly and clears. On the month it reaches thirty-one and the cause is a supplier, it becomes a leadership number. The decision it needs, someone senior phoning the supplier, is above the desk. The rule is that a kept number is promoted to the page for as long as it carries an ask, and comes back down when the ask is answered. It does not stay on the page because it was once interesting. The page is one page, and a number that has stopped moving is occupying a space another number needs.

Numbers also move down. A number that has sat on the leadership page for six months without moving and without prompting a question is a number leadership has stopped reading. It comes off, with a sentence saying so and where to find it. The exception is the five core numbers, which stay even when boring, because their trend is the quarterly review's raw material and because a boring number is a kind of reassurance. "Satisfaction has been fifty-something good of sixty-something replies for six months" is dull and is the sentence a leader most wants to be able to say about the desk without thinking.

The one direction a number never moves is from the technician row to anywhere else. A per-person number does not become a leadership number because it is bad, or because it is good, or because leadership asked. The thing that moves up in those cases is a sentence, written by the lead, about what is being done. "One technician's reopen rate is higher than the others' and we are pairing them on confirmations this month" is a leadership sentence. The name and the rate are not, and the matrix says so in a column of its own.

When Leadership Asks for the Kept Numbers

Sooner or later a leader asks for the per-person figures, usually kindly, usually because they want to understand the team or praise someone. The answer is a team total, a range, and an offer. "The team closed 405 this month; everyone is within the normal spread for their mix of work, and I review each person's numbers with them monthly. If you want to know how a particular person is doing, I would rather tell you in person than send a column, because the column does not know who takes the hard tickets." Most leaders accept that at once, because it is the answer they would give about their own team.

If the request is really about a person, say so and treat it as one. "Should I be worried about Ben?" is a question about Ben. The answer is a conversation about Ben's work, with the mix, the sample and the lead's own view. It is not a report column that will be read again next month without the conversation attached. If the organisation has a process for performance concerns, that is where a real concern goes, with evidence prepared for it. The monthly report is not that process and must not become its back door. A column on a page cannot be cross-examined. A desk lead can, and should want to be.

The one honest exception is the desk lead's own manager reviewing how the desk lead manages, which can legitimately include looking at how the lead reads and acts on individual numbers. Even there, what goes up is the lead's method and decisions, with the numbers as evidence in a conversation, not a league table. The matrix has a row for this too. When the question comes, it has already been answered, calmly, in writing, by a lead who was not being asked it at the time. Deciding audiences in a quiet month is easy. Deciding them with a director standing at the wall screen is not, which is the entire argument for the matrix.

A Short War Story: The Column at Bellweather Logistics

At Bellweather Logistics, a desk lead called Carl added a tickets-closed-per-technician column to the monthly report. He wanted the operations director to see how hard his four people worked. The director, reading it for three months, noticed the lowest name every time. The name was Marta, who handled the warehouse handheld scanners. Each took most of a day to diagnose and reflash, and she was the only person on the desk who could. In the fourth month the director asked Carl to "manage Marta's performance". Carl spent most of that month assembling the category split and the time-per-ticket figures. They showed that Marta's thirty tickets a month were a full load and the hardest one on the desk. The director accepted it, and asked for the column to stay. Marta had seen her name at the bottom of a page every month for a third of a year. She found a job where nobody counted her and left before the summer. The desk covered the handhelds with a contractor three days a week for four months while Carl trained a replacement. The median time to resolve a handheld fault roughly tripled meanwhile, which the depots noticed. A per-person number on a leadership page becomes a verdict the moment it is read, and the person it is about is the last to be consulted.

The Audience Matrix

The matrix below is the template: one row per number the desk produces, four audience columns, and a notes column for the form it takes and the minimum sample. Copy it, add rows for anything your tool produces that is not here, and delete nothing. A number that goes to nobody stays in the matrix with "nobody" in every column, because that is a decision too. The next person to run the desk will otherwise add it back. The sample sizes in the notes are the worked-example figures from this article; set your own, in writing, and err towards larger. Review the matrix twice a year, and whenever a new screen, report or reader appears.

Nadia took the column off the wall screen that afternoon and replaced it with the team's total. She wrote the matrix the following week. It took her two hours and settled, in advance, nine conversations she has not since had to have. Imran got a sentence about Ben, in person, the next morning: he covers the clinical devices and the satellite surgeries, his tickets are the long ones, and he is the person the clinic can least afford to lose. Imran said he had wondered why the other two never drove anywhere. The number had told him something. It had simply not told him anything true.

Frequently Asked Questions