Home › Topics › Experience Over Metrics › Designing for It

Designing the Desk for Experience, Not Just Speed

Friday, 15:30, at Acme. Beatriz runs a desk of four for a manufacturer with three hundred staff. She is standing at the whiteboard with the results of her first ten-ticket audit beside her. Six of the ten tickets had met their service levels. Seven had a moment that went badly, and in five of those the moment was the same. The requester wrote to ask what was happening before anyone had written to them. Beatriz has already given the talk. She gave it in March, about updates, and in June, about closing properly. Both times the tickets improved for three weeks and drifted back. A talk is not a design. Kwame and Sol are good technicians who mean to write the update and then take the next call. The desk is designed, by accident, so that the next call wins.

Designing the desk for experience means choosing the habits, templates, settings and rhythms so that the five moments in a ticket's life go well by default, when nobody is trying. They do not depend on heroics when somebody remembers. Good technicians on a desk designed for speed produce fast tickets that feel bad. That is not because they are careless but because every default points at the next ticket and nothing points at the requester. This article is part of our Experience Over Metrics series; the moments themselves were laid out two articles ago. And this one is about building the desk so that they happen without a talk.

It covers design rather than effort, then the update cadence, the named technician, and the plain-language reply. Next come no reopen surprises, the kind redirect, and the close that asks. We cover making the defaults stick, with a table, then a short war story. We finish with the experience design checklist to copy, in that order.

Design, Not Effort

A default is what happens when nobody decides, and a desk's experience is made of its defaults. When a ticket arrives and the technician is busy, the default is either an automatic acknowledgement or a named reply from a template that takes thirty seconds. The desk chose one of those, whether or not anyone remembers choosing. When a ticket has sat two days, the default is either silence or a reminder on the technician's list that says "update due". The desk chose that too. When a fix lands, the default close is either a click or a message with a question in it. Every one of these was decided once, usually by whoever set the tool up, and has been running ever since.

Effort is what a technician spends overriding a default, and effort runs out on busy days, which are the days requesters most need the override. This is why talks fail: a talk asks for effort, the effort is given for three weeks, and then a bad Monday arrives and the defaults win. Design asks for the opposite. It changes the default so that the good moment is the path of least resistance and the bad moment requires a decision. A technician who has to delete a template sentence to send a reply without a time will leave the sentence in. (Technicians are not lazy. They are efficient, which on a badly designed desk looks the same.)

Six design choices cover the five moments, and each is a change to a template, a setting, a rhythm or a rule rather than a request for anyone to try harder. They are not expensive. Most of them took Beatriz an afternoon, and the one that took longest was the argument about wording.

The Update Cadence

The update cadence is a promised rhythm, per priority, at which the desk writes to the requester of every open ticket. The desk writes whether or not anything has changed. And the cadence is the design choice that fixes the moment most tickets are judged on. The rhythm is written down: daily for the top priority, every two working days for the next, and a stated interval for the rest. It is measured from the last message the requester could see, so internal notes do not reset it. And it is enforced by the desk's list rather than by memory. Every morning, before the queue, each technician looks at the tickets whose last public message is older than their interval, and writes to those first. The order matters. Updates before new tickets, because the new ticket has not yet been let down and the old one is about to be.

The message when nothing has changed is the one technicians resist, and it is the one that does the most work. "No change since Tuesday; I am still waiting on the part, which the supplier now says is Thursday; I will write to you Thursday afternoon either way." The requester learns three things: the ticket is alive, the reason is real, and the next time is fixed. They can now put it down. The craft of the cadence, including what to say at each interval and how to keep it from becoming noise, is owned by Follow-Up Cadences That Keep Tickets Moving. The design decision here is only that the cadence is a rule with a morning ritual behind it, not an intention.

The cadence has a template, because a technician writing from a blank box at 08:50 with eleven tickets due will write "still looking into it". That message fails the requester while passing the filter. The template has three blanks: what changed, what is next, the next time. A message that leaves the third blank empty is not sent. That is the whole design. On Beatriz's desk it took the chase count from five of ten to one of ten in the first month, with no talk.

The Named Technician

The named technician is the design choice that one person's name is on every ticket from the first reply onward. The requester knows it, and that person is who the requester hears from until an introduction says otherwise. It fixes the first reply, by replacing "Support Team" with a person. It fixes the handoff, by making every change of name an event that has to be announced. A ticket owned by a group is owned by nobody in the requester's mind. And a reply signed by a group is a reply from a building.

The mechanism is claim-based ownership, in which a technician takes a ticket by claiming it rather than being assigned it by a dispatcher. And the queue shows who holds what. In QuerySolv a ticket is claimed by a person and carries that person's name from then on. That is what lets the first reply say "I have this" and mean it. Most tools have something similar. Where yours assigns by group, the design fix is a rule that nobody replies from a group and the first human reply claims the ticket. The name goes in the signature and the first sentence. "Hi Priya, Kwame here, I have this" is a sentence that does more for the first moment than any amount of speed.

The named technician also needs a design for absence, or the name becomes a trap. A requester who was told "Kwame has this" and then hears nothing for three days because Kwame is on leave has been let down by the name. The rule is that nobody goes on leave holding a ticket without an introduction to the person covering, written to the requester. The rule also requires that the cover reads the history before writing. That is a handoff, and it is designed the same way as any other: the old name introduces the new one before the new one speaks. A desk that keeps names on tickets without keeping this rule has designed a better first moment and a worse third one.

The Plain-Language Reply

The plain-language reply is the design choice that every message to a requester is written for the requester, in words they would use, at the depth they need. It fixes the fix moment and most of the understood-the-reply target. The failure it replaces is the message written for another technician that happened to be sent to a housing officer, the "perms corrected on share" message. And that failure is not a writing problem. It is a defaults problem: the resolving reply box is the same box the technician writes notes in. So the note voice is what comes out.

The design fix is a resolving template with three labelled blanks: what was wrong, what I did, and what to try, in that order. It is written as if to someone who will read it on a phone in a corridor. "The two new folders had been set up without your team's group on them, so you could see them but not open them. I have added the group. Try opening either folder now; if one still refuses, reply here and it comes back to me." The template does not make anyone a better writer. It makes the plain version the shape the box is already in, and a technician filling blanks writes differently from a technician writing a note. The voice itself, how to sound like a person and not a policy and how to keep five people sounding like one desk, is the subject of the tone series and in particular Writing Like a Person, Not a Policy. This article only puts the template where the note used to be.

There is a second half to this choice, which is reading. Ten resolving replies a month, read by the desk lead against one question: could the requester say what was wrong in their own words from this? A reply that fails is not a reprimand. It is a template that needs a blank renamed, or a technical term that needs a standing plain translation on the team's list. Beatriz keeps a list of eleven such translations on the whiteboard. "Permissions" is on it. So is "DNS", the domain name system, with the honest note that the plain translation is "the thing that turns names into addresses, and it was wrong".

No Reopen Surprises

No reopen surprises is the design choice that a requester who replies to a resolved or closed ticket lands back on the same ticket, with the same history, with the same named person. They never land in a fresh ticket at the back of the queue or in a message saying "this ticket is closed, please raise a new one". This choice fixes the close moment from the far side: a close only feels safe to accept if the way back is open. A desk whose closes are one-way teaches requesters to argue about closing, to leave tickets open as insurance, or to stop confirming anything at all.

The mechanism is a tool setting in most desks, and it is worth checking rather than assuming. A requester's reply to a resolved or closed ticket reopens it, with its history, in QuerySolv and in many tools. Some tools instead create a new ticket from the reply, and some do nothing with it. A desk on one of those needs a rule that any such reply is linked to the original by hand. The rule also requires that it goes to the original owner the same day. The requester should not be able to tell which of those happened. From their side, they replied and the same person answered, knowing everything.

The design has a second part, which is what the same person says. A reopened ticket is a requester who was told it was fixed and found that it was not. And the first reply has to say so: "You are right, it has come back; I am sorry about that. I can see the earlier fix, and I am going to look at why it did not hold." Not a fresh acknowledgement, not a request to describe the problem, not a hint that the requester did something to break it. The craft of that reply, and what a reopen is telling the desk, is in Handling a Reopened Ticket Well. The design decision here is that the reopen goes to the name on the ticket. That name reads the history before writing. And the template for the first reply on a reopen is different from the template for a new ticket. A desk with one template for both will greet a returning requester as a stranger.

The Kind Redirect

The kind redirect is the design choice for the moment that happens outside the ticket. That might be the tap on the shoulder, the message to one technician, or the person at the desk with a laptop under their arm. A desk designed for speed answers "can you raise a ticket?" and turns away, which is efficient, measurable and remembered for years. A desk designed for experience logs the request on the spot, hands back the number, and says when someone will be in touch. That takes a minute longer and is remembered as being helped. The walk-up is a channel, whether or not the desk wants it to be. And the design decision is what the channel feels like to walk through.

The mechanism is that every technician can open a ticket in under a minute from wherever they are standing. The ticket has a source that says "walk-up" so the channel can be counted. And the sentence is agreed and the same for everyone: "Let me log that now so it does not get lost, and I will send you the number; someone will be back to you by this afternoon." The request has been redirected into the queue without the requester feeling redirected. And the queue now holds the work, which is the only way the desk will ever get credit for it. The full craft of the redirect, including the second and third time the same person walks up, is owned by Redirecting Without Alienating People. The design here is only that the sentence exists, the ticket takes a minute, and nobody is ever sent away to type.

The kind redirect is also where a desk's experience is most visible to people who never raise tickets. Colleagues watch how the person at the desk is treated. A desk that is warm in the queue and cold at the counter has a reputation made at the counter. And no cadence of updates will fix it, because the people forming the reputation never see an update. Design the counter the way you design the first reply: a name, a time, and a ticket number before they walk away.

The Close That Asks

The close that asks is the design choice that no ticket is closed by the desk on the day it is fixed. The resolving reply contains a question. And the requester's answer, or a stated number of silent days, is what closes it. This choice fixes the last moment, which is the one the requester carries out of the room. The close that asks is the cheapest of the six: a template change and a setting. The resolving message ends "If that has sorted it, I will close this; if not, reply and it comes straight back to me." The ticket sits resolved, not closed, until the requester says yes or the days run out, and then it closes with a note saying which.

The default this replaces is the click. On a desk designed for speed, resolve and close are one action, taken in the moment of the fix, because closed tickets are what the clock counts. The requester finds out from a notice that says "closed". The silent close, how it happens and how to stop it, is the subject of Avoiding the Silent Close. This article adds only the design view. The click is a default, and the question is a template. A desk that changes the template and separates resolved from closed has removed the silent close without asking anyone to remember anything.

The question in the close has a second job, which is to catch the add-on. "Anything else while I am here?" at a desk-side visit, or "is there anything else in this ticket I have not answered?" in a message, is how the desk finds the missed question. It is the one the requester asked in their second paragraph that nobody read. The audit in the first article of this series found unanswered questions in a third of closed tickets on one desk. Each one was a small lesson to the requester that the desk reads the first line only. The close that asks reads the whole ticket once more before closing it. It is the moment at which the desk proves it was listening, which is the thing the requester was never sure of.

Remember: a desk's experience is made of its defaults, and effort runs out on busy days. Change the template, the setting, the rhythm and the rule so that the good moment is the easiest thing to do. And read ten tickets a month to see whether the defaults are holding.

Making the Defaults Stick

Each of the six choices is a default, and a default sticks only if something enforces it and something notices when it slips. The table below gives, for each moment, the design choice and the thing that enforces it. It also gives the sign that it has slipped, which is the sign the monthly reading looks for. The enforcement is never a person remembering. It is a template that has the blank in it, a setting that separates resolved from closed, a list the morning starts with, or a rule everyone can recite. The slip signs are all counts from the ticket reading, so that a drift shows within a month rather than at the next complaint.

MomentDesign choiceWhat enforces itThe sign it has slipped
First replyThe named technicianClaim before replying; a first-reply template with name and next-time blanks; no replies from a groupReplies signed "Support"; first replies with no time in them
Waiting updateThe update cadenceA written interval per priority; the morning list of tickets past their interval, worked before the queue; a template whose next-time blank cannot be left emptyTickets with a gap past the interval; any chase from a requester
HandoffThe named technician, continuedThe rule that the old name introduces the new one before reassigning; leave cover written to the requesterA new owner's first message asking what the problem is; a reassignment with no public message beside it
FixThe plain-language replyA resolving template with three labelled blanks; the team's list of plain translations; ten replies read a monthResolving messages under fifteen words; replies that fail the "could they repeat it" test
CloseThe close that asks; no reopen surprisesResolved and closed kept separate; the close question in the template; replies to closed tickets routed to the same owner with the historyTickets closed on the day of the fix by the desk; a second ticket from the same requester about the same fault
Outside the ticketThe kind redirectA ticket openable in a minute from anywhere; a "walk-up" source; one agreed sentenceWalk-up tickets logged after the fact with no number handed back; a busy desk and a thin queue

Beatriz's version of this table lives on the wall above the desk. The right-hand column is what she reads ten tickets against on the last Friday of the month. In the first month the slip signs were everywhere, which was expected. By the third, the cadence and the close were holding without her. The named technician was holding except on leave weeks. The plain-language reply was the one still drifting, because it is the one that depends on a person's judgement rather than a blank. She reads twenty of those now instead of ten. (Templates can be enforced. Clarity has to be practised.)

A Short War Story: The Fast Desk at Larkspur Hotels

Nadia ran the desk for Larkspur Hotels, a group of nine properties, and ran it fast. There was a first response inside twenty minutes on most tickets and a resolution inside the day on most of the rest. The service-level table had been green for a year. The replies came from "IT Support", and the closes were a click on the day of the fix. The front-desk staff at the properties, who raised most of the tickets, had learned that a ticket was something that got closed at you. They stopped raising them for anything they could live with. And the thing they lived with for two days, one spring, was the booking terminal at the largest property printing confirmations with the wrong rate. They corrected the rate by hand at check-in because "IT would just close it". The error reached the finance team through a guest's complaint. The cost was about two hundred bookings re-invoiced and a weekend of two people's time. There was also a general manager who asked Nadia, in front of her director, why the desk had not known. Nadia's audit afterwards found nothing slow and nothing wrong; it found a desk whose every default pointed away from the person who had asked. She changed four templates and one setting over the following month. The lesson she gave her team was the one on this page. The front desk did not stop raising tickets because the desk was slow. They stopped because it was designed to be finished with them.

The Experience Design Checklist

The checklist below is the six design choices as things a desk either has or does not. The enforcement is beside each so that "we do that" has to mean "here is the template" rather than "we try to". Go through it once with the team and tick only what is genuinely in place, which on most desks is fewer than half on the first pass. Then pick the two unticked items with the most red in your monthly reading and build those. The cadence and the close are the usual first pair, because each is an afternoon's work and each moves the chase count visibly. Come back to the list each quarter. A default that was in place in spring can be gone by autumn, usually because a template was "tidied".

THE EXPERIENCE DESIGN CHECKLIST — SIX DEFAULTS, CHECKED QUARTERLY
Desk: ______________________   Checked by: ______________   Date: ____________

1. THE UPDATE CADENCE
  [ ] An interval per priority is written down:  P1 ____  P2 ____  others ____ working days
  [ ] The interval counts from the last message the requester could see, not from a note
  [ ] Each morning starts with the tickets past their interval, before new tickets
  [ ] The update template has three blanks: what changed / what is next / next time
  [ ] An update with the next-time blank empty is not sent

2. THE NAMED TECHNICIAN
  [ ] Every ticket is claimed by one person before the first human reply
  [ ] No reply is ever signed by a group or a team name
  [ ] The first-reply template has blanks for the name and the time to come back by
  [ ] Nobody reassigns a ticket without a public introduction written first
  [ ] Leave cover is introduced to the requester before the leave starts

3. THE PLAIN-LANGUAGE REPLY
  [ ] The resolving template has three labelled blanks: what was wrong / what I did / what to try
  [ ] The team keeps a list of plain translations for its common technical terms
  [ ] Ten resolving replies are read each month against "could they repeat it?"
  [ ] A failed reply changes a template or the translation list, never only the person

4. NO REOPEN SURPRISES
  [ ] A reply to a resolved or closed ticket lands on that ticket, with its history (tested, not assumed)
  [ ] The reopen goes to the named owner, or to cover with an introduction
  [ ] The first reply on a reopen uses its own template and never re-asks the problem
  [ ] No message to a requester ever says "please raise a new ticket" for the same fault

5. THE KIND REDIRECT
  [ ] Any technician can open a ticket in under a minute from wherever they are
  [ ] Walk-up and direct-message tickets carry a source that says so
  [ ] One agreed sentence: log it now, hand back the number, say when
  [ ] Nobody is sent away to type

6. THE CLOSE THAT ASKS
  [ ] Resolved and closed are separate states, and the desk does not close on the day of the fix
  [ ] The resolving template ends with a question and the way back
  [ ] A set number of silent working days closes the ticket, with a note saying so: ____ days
  [ ] The close re-reads the ticket for any question not yet answered

Ticked on this pass: ____ of 25     The two to build next quarter: ______________  ______________

Beatriz ticked nine of twenty-five on her first pass, which she described to Kwame and Sol as "better than I feared, which is a low bar". She built the cadence and the close in the first month and the named technician in the second. The talk she had given in March was never given again. The tickets did not drift back this time, because there was nothing to drift back to. The templates had the blanks in them, and the blanks were where the moments lived.

Frequently Asked Questions

Will all this messaging slow the desk down?
Each of the six choices adds between thirty seconds and two minutes to a ticket. And most of that comes back as fewer chases to answer, fewer reopens to handle and fewer requesters walking over to ask. On desks that have measured it, the chase count falls within a month and the time spent answering chases falls with it. The service level stays in place and is still reported; design for experience does not loosen it.
Which of the six should we build first?
Build the update cadence and the close that asks first, because each is an afternoon's work. Each is enforced by a template and a setting rather than by memory, and each moves the monthly reading visibly. Then the named technician, which needs a rule about introductions and leave cover. Leave the plain-language reply until the others hold; it depends on practice rather than a blank, and it takes the longest to settle.
Our tool assigns tickets to a group, not a person. Can we still have a named technician?
Yes, by rule rather than by setting. The first human to reply claims the ticket, and puts their name in the first sentence and the signature. That person stays the name on it until an introduction says otherwise. Replies from the group name are not allowed. Most tools can show who last replied, which is enough to make the rule checkable in the monthly reading.
What do we write when nothing has changed?
Say that nothing has changed, say why in one sentence, and give the next time: "No change since Tuesday; still waiting on the supplier, who now say Thursday; I will write Thursday afternoon either way." The requester learns the ticket is alive, the reason is real and the next time is fixed. An update with no next time is noise; an update with one is the moment most tickets are judged on, handled.
How do we know our reopen setting works the way we think?
Test it. Resolve a test ticket, close it, and reply to the closing message from the requester's address. Watch where the reply lands: on the same ticket with its history, in a new ticket, or nowhere. Do this for each channel the desk accepts, because e-mail and the web form can behave differently. If the reply does not land on the original, write the rule that links it by hand the same day and routes it to the original owner.
Is the kind redirect not just encouraging walk-ups?
It logs them, which is the first step in reducing them. A walk-up that becomes a ticket with a "walk-up" source is counted. So the desk can see the channel's size and later design a window or a chat route for it. A walk-up turned away with "raise a ticket" is neither counted nor reduced; it becomes a direct message to a technician next time. Redirect into the queue, kindly, and measure the shift over the following months.

From the QuerySolv team: we build self-hosted help desk software for Windows. QuerySolv runs as one Windows service on your own network. It takes requests by web form, e-mail, live chat, social messages and REST API. It is licensed once per installation rather than per agent per month. The free trial runs the full product for 30 days.