Dashboards That Motivate Instead of Surveil

Thursday, 16:10, at Bluewater Bank. The branch-support desk has four people and, since Monday, a new wall display that somebody in operations was very pleased with. It reads "Tickets closed this week", by name, largest first, in green. Sofia is at the top with sixty-one. Nikhil is at the bottom with nine, because he has spent the week on one ticket. The ticket is about a branch whose network dropped every afternoon and whose forty staff could not serve customers while it did. Dana, the desk lead, has just noticed something else. In the last ten minutes Nikhil has claimed four password resets from the unassigned view, one after another. The branch ticket has not been touched since lunch. He has not been told to do this. The screen told him.

A leaderboard is a ranked list of people by a number. And a surveillance dashboard is any shared display whose main effect is to let people be compared, whether or not that was the intent. A motivating dashboard is one that shows the team its shared work and its shared progress and leaves each person's figures to that person. The difference is not in the numbers, which are often the same, but in who sees whose. The difference in behaviour is large and arrives within a fortnight. This article is part of our Team Dashboards series. This article is about why the leaderboard damages the desk it was meant to improve, and what to put up instead.

It covers what a leaderboard teaches and the number that gets gamed the moment it is displayed, with a table. Next come the private-versus-public rule, team totals over individual ranks, and per-person figures as coaching and not ranking. Then come what motivates a desk and what to do when someone asks for the leaderboard. A short war story and the dashboard ethics checklist to copy come last, in that order.

What a Leaderboard Teaches

A shared screen is a statement of what the organisation values. The people under it adjust to the statement, not to the speech that accompanied it. Dana's speech on Monday was "this is just so we can see how busy we are". The screen's statement was "closed tickets, by name, largest wins", and the desk, being made of intelligent people who could read, obliged. Within a week the password resets, which take four minutes, were being claimed within seconds of arriving. The branch fault, which was one ticket however many hours it took, was becoming the thing nobody wanted to be seen holding. Nobody had decided this. The screen had, and the screen is more persuasive than any lead, because it is there all day and never softens its message.

The lesson a leaderboard teaches is specific: do what moves the number, and avoid what does not. Four behaviours follow, in roughly this order. Easy tickets are claimed first and hard ones left, because a hard ticket is one closed and an easy one is also one closed. Tickets are split, because a request for a laptop, a monitor and a login is three closed instead of one. Tickets are resolved on the first reply rather than on the fix, because "resolved" is when the count moves. And tickets are avoided altogether where they can be, because a ticket that is never claimed never counts against anyone. The general mechanics of how a measured number gets bent, and the cures, are the subject of Metric Gaming and Its Cures. The leaderboard is the mechanism at its most efficient, because it bends the number in public, in real time, with a name attached.

None of this makes the technicians cynical, and the mistake most leads make is to think it does. Sofia at the top of the board was not cheating; she was doing the work the screen asked for and doing it well. Nikhil at the bottom was not slacking; he was doing the work the branch needed and watching the screen tell him it did not count. The people were fine. The screen was teaching the wrong lesson, loudly. And the only honest response is to change the screen, not to lecture the people about ignoring it.

The Number That Gets Gamed the Moment It Is Displayed

There is an old observation that a measure which becomes a target stops being a good measure. And a leaderboard is the fastest way to make a measure a target. The number did nothing wrong while it sat in a report, read monthly by one person who knew what it meant. Displayed by name, in public, updating live, it stops describing the work and starts directing it. From that moment it describes the directing instead. Closed-ticket counts on a leaderboard measure who is best at closing tickets. That is not the same as who is best at support, and the gap between the two is where the requesters live.

The table below takes the figures most often put on a leaderboard. It shows how each gets bent, what the bending costs the requester, and what to show instead. The "instead" column is nearly always a team figure or a private one. In every row the replacement is a number the desk can be proud of without any individual being shamed by it. Read the middle column carefully, because every item in it is a thing I have watched a good technician do under a bad screen. In two cases the technician was me.

Notice what the "instead" column does to the leaderboard's original purpose, which was usually to show that the desk is busy and that people are working. A team figure for tickets resolved today shows the desk is busy. An unassigned count that stays near zero shows people are working. Neither needs a name. The information the manager wanted survives; only the ranking is gone, and the ranking was the part doing the damage.

Private Versus Public Numbers

The rule that resolves nearly every dashboard argument is this: team figures are public and individual figures are private. And no figure moves from the second column to the first without a decision written down. Public means on the wall, on the shared view, in the team meeting, in the monthly report. Private means visible to the technician whose figure it is and to their lead. It is in the technician's own view and in the one-to-one, and nowhere else. The personal view that carries the private half is the subject of Personal Workload Views. This article is about holding the line between the two.

The line is held by the lead, and it is tested from above more often than from below. Technicians rarely ask to see each other's numbers. Managers, project sponsors and the person who owns the wall screen ask regularly, usually in the form "can we just see who is doing what?" The request is reasonable on its face. The answer is that the lead can see who is doing what, in private, and acts on it. The manager sees the team's figures and the lead's judgement. A desk whose individual figures are visible to everyone has a lead whose judgement has been replaced by a sort order.

If you run QuerySolv, the split is already the shape of the tool. Each technician sees their own numbers in their own view. The shared dashboard shows what is waiting and what is overdue for the whole desk. On any tool the test is simple. Could a technician, without asking anyone, see a colleague's closed count or reopen rate? If yes, the figure is public whatever the policy says, and the behaviour in the table above has a head start.

Team Totals Over Individual Ranks

What goes on the wall is the team's work, and the team's work has a shape that motivates without ranking anyone. The unassigned count, with the age of the oldest, shows whether the desk is keeping its first promise. Watching it drop to zero at eleven after a heavy Monday morning is a shared satisfaction nobody has to be told to feel. The overdue count at zero is the same satisfaction about a harder promise. Tickets arrived against tickets resolved since midnight shows the day's weight and the day's progress, as information, never coloured as judgement. Those three are the live heart of a shared view. And the five counts they come from, with the order to read them, are in The Real-Time Queue View.

Team totals have a second virtue, which is that they are honest about how support actually works. A desk is not four people racing. It is four people covering one queue. The person on the long branch fault is making the three others' days possible by taking the thing that would otherwise have stopped all four. A team figure credits that. An individual figure punishes it. The branch ticket at Bluewater, counted in the team's resolved total on Friday, was one ticket among seventy and had been the most important of the week. The team's number said so by including it without comment. Sofia's sixty-one and Nikhil's nine were both true. Neither was the point.

Where a desk wants a figure that improves over time rather than a live one, make it a team figure in the monthly report. Pick one that cannot be bent by splitting or premature closing. That could be the median time to resolution beside the reopen rate, or the share of tickets resolved without a handoff beside satisfaction. Pairs resist bending because the trick that improves one worsens the other, which is the idea behind Reading Metrics Together. A single number improves. A pair improves or tells you why not.

Per-Person Figures Are for Coaching, Not Ranking

Individual figures exist and are useful, and the use is coaching. That happens in private between a technician and their lead with the figures in front of both. The lead opens the technician's reopens and they read three of them together. Either the fixes were sound and the requesters had second questions or the fixes were not and there is something to learn. The lead opens the technician's resolution times for one ticket type and asks what made the slow ones slow. None of this works if the figure has already been displayed. A displayed figure has already been defended, and a defended figure cannot be learned from.

No per-person figure is read at all until there are enough tickets behind it to mean anything, and the floor is higher than most leads expect. A technician's resolution time over nine tickets is nine stories, and the median of nine is the fifth story. A reasonable floor, and it is a floor rather than a standard, is thirty to fifty resolved tickets and at least a month. That comes before a per-person figure is looked at. Even then it is read as a median with the slowest few listed, never as an average, for the reasons set out in Median and Percentiles, Not Averages. Satisfaction by person needs the most care of all, because the responses per person are few and the requesters who answer are not a fair sample. The careful method is in Satisfaction by Category, and Carefully by Agent.

Every per-person figure is read beside a quality figure, because a speed figure on its own rewards the behaviour in the table. Closed count beside reopens. Resolution time beside satisfaction. First-response time beside whether the first response said anything. The pairing is what turns a number into a conversation rather than a verdict, and the conversation is the whole value of the figure. A lead who cannot say what the paired figure is has not got a coaching tool. They have got a ranking with one row.

What Actually Motivates a Desk

What motivates a support desk is the same thing that motivates most people who do skilled work. It is visible progress on something shared, recognition from people who understand what the work took, and the absence of things that make the work pointless. The shared view provides the first, when the counts fall through the day because the desk made them fall. The second comes from the lead and from colleagues, in words, about specific tickets. "The branch thing was a nightmare and you stuck with it" is worth more than a month at the top of any board. That is partly because everyone in the room knows it is true. The third is the lead's job and has no dashboard.

Celebrate tickets, not counts, and tell them as stories. Suppose a desk has a five-minute slot in its weekly meeting for "the ticket of the week", chosen by the team and told by the person who worked it. That desk has a recognition system that rewards exactly the hard, unglamorous, single-ticket work that a leaderboard punishes. The branch fault would have been Bluewater's ticket of the week by acclamation. On the leaderboard it was nine. The gap between those two verdicts is the whole argument of this article. And the desk knew which one was right before anyone wrote it down.

Honest opinion, since the reader will want one: a leaderboard does not motivate the people at the bottom. It motivates the people at the top to stay there by the means the table describes. The people in the middle learn to ignore it, which is the best outcome available and still means the screen is teaching the desk to ignore screens. I have never seen a help desk leaderboard improve the service requesters received. I have seen several improve the number on the leaderboard.

When Someone Asks for the Leaderboard

The request will come, and it will come from someone senior, in good faith, usually phrased as transparency. "Why can't we just see who is closing what?" The answer is not "no". It is "here is what you actually want, and here is why the ranked version gives you the opposite". What they want is to know that the desk is busy, that people are pulling their weight, and that problems are being caught. The team figures show the first. The lead's judgement, backed by private figures and one-to-ones, provides the second. The overdue and unassigned counts at zero show the third. Offer all three, in the one-page monthly report whose format is in The Monthly Report They Will Actually Read. Offer them before the request arrives if you can.

Then explain the table in one sentence. The person asking has usually not thought about it and will see it at once when it is said: "The moment we put closed tickets by name on the wall, the easy tickets get claimed in seconds and the branch fault gets left, and the requester with the branch fault is the one who phones you." Senior people have seen measures bent before, in budgets and sales targets, and recognise the shape. The conversation about declining a request from above without damage is the subject of Saying No Upward. The leaderboard conversation is one of its easier cases, because the alternative on offer gives the asker more of what they wanted, not less.

If the leaderboard goes up anyway, and sometimes it does, measure its effect and report that. Count reopens, ticket splits and the age of the oldest hard ticket for the month before and the two months after. And put the figures in the monthly report without comment. The numbers will make the argument that the conversation did not. And they will make it in the language the person who asked for the board already speaks. Nobody enjoys being proved wrong by a report. It is, however, considerably more persuasive than being told so in a meeting.

A Short War Story: The Fastest Desk at Meridian Parts

At Meridian Parts, a desk lead called Rafal put a wall display of "average resolution time this week, by technician" above a three-person desk. A manager had asked whether the desk could "be a bit quicker". The desk got quicker within a fortnight: the figure fell from about nine working hours to under five and stayed there. Reopens, which had run at about four a month, ran at nine, then eleven. Tickets were being resolved on the first reply and reopened on the second. A sales director's laptop was "resolved" three times in one week for the same fault. Then he stopped raising tickets at all and started phoning the IT manager directly, which the IT manager mentioned to Rafal with some feeling. The cost, when Rafal counted it, was about twenty hours a month of rework on reopened tickets and a requester the desk had lost to the phone. It also cost a month of the IT manager asking whether the desk was "resolving things or just saying so". The display came down, and the median resolution time went into the monthly report beside the reopen rate. The desk's figure drifted back up to about eight hours, which was what the work took. The screen had never made the desk faster. It had made the desk say it was.

The Dashboard Ethics Checklist

The checklist below is run before any figure goes onto a shared screen or into a team meeting, and again at the quarterly review of whatever is already up. It is deliberately short and deliberately blunt. The arguments for a ranked figure are always reasonable in the moment, and the checklist is there to be less reasonable than the moment. A figure that fails any line in section A does not go up. Section B is the private-figure discipline, section C the replacement list, and section D the questions to ask the person who wants the ranking.

DASHBOARD ETHICS CHECKLIST — BEFORE ANY FIGURE GOES ON A SHARED SCREEN
Desk: ______________________   Figure proposed: ______________________   Asked for by: ______________

A. THE SHARED-SCREEN TEST (all boxes must be ticked, or it does not go up)
   [ ] The figure is about the team or the queue, not about a named person
   [ ] It cannot be improved by splitting tickets, closing early, or avoiding hard tickets
   [ ] It names the action somebody takes in the next hour when it changes
   [ ] It is a count or an age, not an average or a rank
   [ ] If it is a total, it is shown as information only, never coloured as good or bad
   [ ] Its definition is written down and two people get the same number from it

B. THE PRIVATE-FIGURE DISCIPLINE (for any per-person figure)
   [ ] Visible only to the technician and their lead
   [ ] Read only after ____ resolved tickets (thirty to fifty) and at least one month
   [ ] Read as a median with the slowest few listed, never as an average
   [ ] Always read beside a quality figure: ______________________
   [ ] Discussed in the one-to-one, for coaching, and recorded nowhere shared
   [ ] Never read out, displayed, ranked, or used in a pay or grading decision on its own

C. WHAT WE SHOW INSTEAD
   Live, shared:   unassigned count and age of oldest   overdue count   due soon   waiting on us
                   arrived / resolved today (information only)   who is on the desk
   Monthly, team:  median time to resolution beside reopen rate
                   satisfaction with response count, by category
                   tickets resolved without a handoff
   Weekly meeting: the ticket of the week, chosen by the team, told as a story

D. WHEN SOMEONE ASKS FOR A RANKED FIGURE
   What do you want to know? ________________________________________________
   Which team figure or private one-to-one answers that? ______________________
   What behaviour would the ranked version reward? ___________________________
   Offered instead: ____________________________   Accepted: [ ] yes  [ ] no, and we measure the effect

E. QUARTERLY REVIEW OF WHAT IS ALREADY UP
   Figures on shared screens today: ____   Each passes section A: [ ]   Removed this quarter: ____
   Reviewed by: ______________   Date: ____________   Next review: ____________

Dana took the closed-by-name display down at Bluewater on the Friday and put the team's resolved total for the day in its place, beside unassigned and overdue. That took an hour and prompted one e-mail from operations asking where the names had gone. She answered with the branch ticket, which had been resolved on the Friday afternoon by Nikhil after three days. She also gave the number of password resets Sofia had cleared in the same week. That was real and was praised in the team meeting by name, in words, by the people who had watched her do it. Nobody at Bluewater has asked for the names back. The branch has not dropped since.

Frequently Asked Questions