SLA compliance is the share of the promises you made about time which you then went and kept, worked out as a percentage and handed over to whoever it was that asked you for it. Straightforward enough as an idea. The awkward part of it arrives later, on the morning the figure gets pulled for the first time and comes back a good deal healthier than the desk has felt from the inside all year, and you are stood there holding two things which are both true and do not agree with one another at all. Both of them stay true as well, mind you.
You will most likely go looking because somebody made you. A client puts a line about response times into a renewal, or the finance side wants a slide for the board, or a new person arrives and asks how the queue is doing, and off you go into the reporting after a figure that has never once been pulled in the whole of it. It comes back at ninety seven percent. And you know, reading it, about the woman who wrote in on a Tuesday in March and heard nothing that answered her until the following week, and she is sat inside that ninety seven percent, counted as a promise kept.
Nobody built the thing to flatter you. The figure is doing precisely the job it was set up to do, and the job is a narrower one than the words on the front of it suggest. That is most of it, truth be told.
So how a desk goes and passes its own test
Take a normal target apart and you will find two clocks in the inside of it. The top one runs from the moment the mail lands until somebody says anything at all back, which is the bit called first response and the bit every report in the world leads with. Underneath that sits resolution, a longer and much less obliging clock, running on until the thing the person actually wrote in about has been sorted out and finished. Both of them go into the policy on the very same afternoon, written by the same people, meant with the same seriousness.
Only the top one is inside your own control on an ordinary Tuesday. The one sitting under it never was.
Resolution waits on a supplier who has not rung back, or on a refund sitting in whatever queue refunds sit in over at the bank, or on the one person who knows that particular product being off with a chest infection. So the first response target is the one that gets watched and quoted and put up on the wall, on account of it being the one a desk can hold in a bad week, and the resolution target quietly goes and turns into a thing you would like to happen rather than a thing you are promising anybody.
And a first response is a low bar by its nature. A line saying the mail has been received and that somebody will look into it stops that clock every bit as dead as a full answer does, and it should, since a person who knows they have been heard waits far better than a person shouting into nothing. Nobody set out to game any of that. The one answering at half eight with forty things in front of them is doing the sensible thing by clearing the easy acknowledgements first, and they are being measured on speed, and they are hitting the target honestly and fairly. Then the mail goes back into the pile and the second wait starts. The second wait has no clock on it at all.
That is a fair share of the gap between the number and the desk, right there. It is not fraud and it is not carelessness either, it is only a measurement that gives out earlier than the customer does, which is a duller explanation than anybody wants and is generally the true one.
Where the clock and the customer stop agreeing
Business hours are the first place the two of them part company, and there is nobody to be blaming for it. You set the clock to run Monday to Friday, nine until six, which is sensible enough and is how the whole trade does it, since it would be daft to be judged on the eleven hours when one and all were asleep. The man who wrote in at ten past six on the Friday is not asleep, though. He is at home waiting. He spends the whole weekend waiting on a thing your reporting has parked. By Monday morning your clock has run about forty minutes and his has run two and a half days, and the pair of those are honest measurements of two entirely different things.
Reopening does something odd as well. Once a ticket has been answered and closed the clock on it has stopped for good, so when the same person writes back underneath saying that is grand but the part you sent is still the wrong one, that second wait belongs to nothing whatever. Plenty of desks go and open a fresh ticket for it and start a fresh clock, which is tidier, and reads better, and means one customer’s fortnight goes into the record as two separate quick jobs. Nobody decided that either.
The Friday tidy up works the other end of the very same problem. A queue that gets tidied on a Friday afternoon, with everything not currently moving marked resolved so that Monday starts clean, is a queue that has quietly reset a heap of clocks, and not one person doing it thinks of it as anything beyond housekeeping. Which it is, mind you. It also decides a fair bit of what the report will say.
Then the priority field, which settles a great deal more than anybody lets on, since the target hanging off urgent is a different target from the one hanging off normal. Urgent tends to get set by tone. Damage comes into it later. The mail with three exclamation marks in the subject line gets marked urgent by whoever opens it, and the one written politely by somebody who assumed you were busy sits at normal and waits away there, and the polite one is quite often the bigger problem of the two. We went into what happens after that over in escalation, where the same thing turns up again a rung further up.
None of these are cheats. Every one of them is a reasonable decision taken by a reasonable person, and the whole of it together is how a desk arrives at a figure it cannot quite recognise its self.
Why the breach that costs you never lands on the report
An SLA report can only count what got into the queue in the first place. A mail that landed on somebody’s own address instead of the shared one is simply not in there. Neither is the question that got asked over the phone and written down on the back of a docket, nor the thing half promised in a corridor on the way past by a person who fully meant it at the time. None of that carries a clock and none of it will ever breach a target of yours. It is not in the sum at all, top nor bottom, and the customer waiting on it has no notion whatever that there is a difference.
The arithmetic is worth doing slowly as well, mind you, since a percentage is built to flatter. Say a thousand conversations in a quarter and ninety seven of every hundred inside target. That leaves thirty people who were not, and thirty people is a room full. Thirty is more than enough to keep a review page ticking over for a year, and the ones who breach are by definition the ones whose problem was awkward, which is to say they are the ones with the most to be saying about it afterwards.
There is one more thing about a breach that nobody warns you of, and it is the part worth sitting with a while. On most desks a breach is an event with no consequence attached to it at all. Something goes red on a screen somewhere. Then nothing much happens. The person who sees it go red is generally the very same person already up to their eyes, and a thing going red does not go and create an hour that was never in the day to begin with. Then it turns up on a monthly report a fortnight later, read by somebody who cannot do one single thing about a Tuesday in the middle of last month.
So the report gets filed and the number gets quoted and the desk carries on away exactly as it was. That is a fair description of most SLA reporting, truth be told, and it is the reason so many teams end up with a compliance figure they neither trust nor act on and would not much miss if it went.
What the target is actually for
Set against all of it you would be entitled to ask what the point of the exercise is at all. There is a good answer to that one, only it has nothing whatever to do with reporting, which is the awkward bit of it.
Left alone, a queue sorts its self out by volume and by nearness. The customer who rings twice gets seen. So does anybody whose account manager happens to sit beside you. The one who wrote the once, plainly, and then waited the way you would want one and all to wait, goes to the bottom of it, and every desk on earth does this and most of them half know they are doing it. A target is the thing that argues back on the quiet one’s behalf. It says this one has been sitting nineteen hours and yours has been sitting forty minutes, so this one goes first, and it says that without anybody having to have the argument out loud with a colleague they happen to like.
That is also why the useful way to set targets is a good deal smaller than the matrix people picture when they hear the words. A grid with four priorities down the side and three channels across the top and twelve different numbers in it is a thing that gets built the once, in a quiet week, and consulted approximately never after. We have all built one. One target that everybody in the room could recite from memory will do more work in a year than twelve that live in a document nobody opens.
The one worth having is the first response, since that is the one you can hold when the week goes against you, with a shorter one sitting behind it for urgent, and urgent tied to what is broken rather than to how the mail was written. Business hours want setting to the hours somebody is genuinely sat there, and if that is eight until four with nobody at all on a Saturday then that is what goes into it, since a target quietly assuming cover you have not got is a target you will be breaching every weekend of the year for a reason no report of yours will ever explain to you your own self. Internal desks want much the same treatment, and we wrote that side of it up on the IT ticketing system page, where the requester is a colleague and the shoulder tap is the whole difficulty.
Our end of it, and the countdown on the ticket
Fair is fair, so the product part goes in here and the cost of it goes in directly after.
Maxdesk is a help desk built around the one shared mailbox. On this subject what is in the box is a policy with a first response and a resolution target, set by priority, running on business hours you configure your own self, and there is a default sitting there from the start at an hour to first reply and eight hours to resolve, which you will almost certainly change, though it does at least mean the thing is measuring something from the first day rather than from whenever somebody gets round to it. The countdown lives on the ticket rather than in a report, in front of whoever has it open. Which is the point of it. Breaches get counted live on the dashboard beside open tickets and average first reply and how old the backlog has got, which is the same place your customer satisfaction metrics end up, so the numbers sit beside one another instead of in two separate decks. Rules you build by clicking will route and tag and escalate off the priority, and the mechanics of it are set out over on helpdesk automation.
The billing goes on the workspace and never on the number of people sat inside it, and that matters more on this particular subject than it does on most. Every SLA anybody ever missed was missed on a day when there were not enough people reading the mail. A per head price is a standing argument against putting another head on the mail, and never more so than in the week you most need one. Ours has no opinion whatever on how many of you are in there.
The part of this we do not fix
Our breach alert goes off at the breach, which is the moment after the useful moment. What saves a ticket is somebody seeing the countdown while it is still running down, and that is a person looking rather than a machine telling, so if you were hoping to buy your way out of anybody watching the queue then this will not do it for you. No software anywhere takes that part off you, and it would be a poor page that pretended it did.
Reporting carries an honest cost as well. History on the free plan rolls the three months and the reports on it are the plain ones. Then it goes. So a year of SLA performance is not a thing you will produce off that plan at all. Twenty dollars against the workspace opens that window to twelve months, takes the ads away, puts your own name on the outgoing mail and gives you the deeper reports with an export in them. The AI plan is ninety nine and stretches the memory out to twenty four. If a client is going to ask you for four quarters of figures then better you know that today than the week they ask.
And there is a whole other meaning of the phrase which we have nothing at all to do with. If your SLA is a contract with service credits in it, a penalty coming off next quarter’s invoice when uptime drops under some agreed figure, then that is a legal instrument and it belongs with whoever minds the contracts. We do not watch uptime, plainly said. There is no status page in here and nothing at all that would mind an on call rota for you, and a severity ladder is somebody else’s software altogether. We said much the same over on incident management when that subject came round.
Mail is the one road we read, which puts a hard edge on any target set in here. A promise made on a phone call carries no clock, not with us and not anywhere else in your building, so a shop that does most of its talking out loud will find a compliance figure from here covers the one slice of the promise only, and you would want to know which slice that is before you go quoting it at anybody. The mechanics of the mail side are laid out on email management software for anybody weighing it up.
Where to start, if the number looks too good
Leave the reporting alone for the one evening. Go into last month’s mail instead and pull out twenty conversations at random, nothing clever about the picking of it, and find two dates on each. The day the person first wrote to you. Then the day they got the answer that actually ended it, and not the day somebody first said something back.
Lay the twenty gaps beside one another and see whether the shape of it matches the figure you would have quoted at a client. The middle of the twenty generally comes out grand. It is the tail that catches people, and it runs a fair bit longer than anybody had imagined it would, and the leaving all happens out there rather than in the middle.
If the two do not match, and they usually will not, the fix is not a stricter target. It is deciding what is meant to happen to a thread that has been answered the once and is still not finished, on account of that being the gap the clock cannot see, and it is nearly always the whole of the difference between the desk and the number.
Whatever you decide about tooling after that, go and do the twenty first. A figure you have checked by hand against twenty real conversations is worth a good deal more to you than any figure that came out of a dashboard, ours included.