The Service Desk Meets Every Target — and the Business Has Stopped Calling
A slightly noisy measure of the right thing beats a beautifully precise measure of the wrong one, every time.
The review that everyone passes
Picture the monthly service review. The service desk manager walks the room through a wall of green. Incidents resolved within the service level target: ninety-four per cent. First-call resolution: seventy-eight. Average speed to answer: eleven seconds. Abandonment under five. Every threshold in the service level agreement is met or beaten, and the slide says so in a colour meant to close the conversation before it starts.
At the far end of the table the head of a business unit — claims, or lending, or whatever the organisation happens to sell — says nothing. What the room does not know is that she has quietly stopped logging calls. Her team keeps a short list, written on an actual sheet of paper, of the two analysts who understand how the settlement run behaves at month end, and they ring them directly. The incidents that would have been difficult never enter the system at all. That is not a failure of the numbers. That is why the numbers are so good.
This is the quiet disorder underneath service management as it stands in 2006, and it is worth naming plainly, because almost everyone is now busy doing the thing that produces it.
How we got here, and why it looked like progress
Give the framework its due, because the progress was real. Before the discipline of the last few years, the desk was heroics and tribal knowledge. There was no single point of contact, no reliable categorisation, no record you could hand an auditor. When the one person who knew how the overnight batch really worked took another job, the knowledge left with them. Service Support gave us the single point of contact, the working separation of incident from problem from change, the configuration management database, and an audit trail that the Sarbanes-Oxley regime now more or less demands. Only a few months ago BS 15000 became ISO/IEC 20000, and for the first time an organisation can be certified against its service management rather than merely assert it. I have no wish to pretend the pre-framework desk was better. It was not.
But somewhere in the adoption we quietly swapped one question for another. We stopped asking is the business getting what it needs? and started asking is the desk faithful to its own process? — and then we built every metric, every incentive, and every review slide around the second question, as though answering it were the same as answering the first.
A service level agreement measures the desk’s fidelity to its own promises. It says nothing about whether the promise was worth making.
Green outside, red within
Two mechanisms do most of the damage, and both are invisible on the board.
The first is reclassification. An incident that is really a problem gets its workaround applied and is closed as resolved; the underlying fault is untouched, so the same failure returns next week, is logged fresh, and closes fresh again. The recurrence never dents first-call resolution because, on paper, each occurrence is a new and separately triumphant ticket. Priorities get negotiated downward at the point of logging so the clock is generous from the start. Resolved comes to mean the ticket is shut, not that the cause is gone. The number is green on the outside and red within.
The second is deflection. The genuinely hard demand leaves the measured system entirely — the sheet of paper, the direct call, the favour owed. What remains inside the queue is therefore the easy population, and the easy population performs beautifully. We are not measuring the service; we are measuring the residue left after the difficult service has escaped.
The figure that would tell the truth is never the one with a target attached. It is the reopen rate, the share of “resolved” incidents that recur within thirty days, the volume of change the business made without ever raising a request. Take a desk reporting that headline ninety-four per cent: look instead at its top three incident categories and you may well find a thirty-day recurrence rate north of a third. A third of the solved problems were not solved. No one reported it, because no target was attached to it, and in this discipline what carries no target carries no consequence.
And the temptation, seeing this, is to blame the framework. That is too easy, and it is wrong. Nothing in the method instructs anyone to reward the closing of tickets over the solving of problems; we chose that, because tickets are countable and problems are awkward. The separation of incident from problem management is one of the best ideas we have. The trouble is that incident management has a desk, a queue, and a target, while problem management has a part-time analyst and good intentions — so the organisation fattens the part that is watched and starves the part that matters.
The one change worth making
Add a single measure the business owns, and give it teeth. Not another internal threshold, but a measure of the outcome the business actually feels: did the thing stay fixed, and was the demand that mattered served without anyone having to route around us to get it. Put the reopen and recurrence rate on the same page as the service level attainment, in the same colours, every month, and let it be exactly as visible as the green. A desk confident in its service has nothing to fear from that line; a desk that flinches at it has just told you where to look.
The objection is honest and deserves an answer: business-owned measures are softer, more arguable, and unfair to a desk that does not control the engineering fixes it depends upon. All true — and none of it a reason to go on measuring only ourselves. A slightly noisy measure of the right thing beats a beautifully precise measure of the wrong one, every time.
The green board is not a lie. It is a faithful answer to a question we stopped needing to ask. The business has already delivered its verdict on the answer, in the most eloquent form available to it. It has stopped calling.