When to Add Heads: Reading Capacity Honestly

Friday, 17:40, at Bluewater Bank. Kofi runs the four-person desk. The queue on his screen holds 212 open tickets. The oldest is forty-eight days old, raised by someone in lending who has stopped replying. Twenty minutes ago the operations director stopped by his desk on the way out and said, kindly, "I read somewhere that one IT person per hundred staff is about normal. You've got four for three hundred, so you've got one spare, haven't you?" Kofi said something agreeable. He is now looking at a queue that has grown every month since the spring. He is also looking at two technicians who have not taken a full week off this year, and a sentence he had no answer to. This article is the answer.

Capacity is the number of hours a desk can spend on its work in a month, set against the number of hours the work needs. Adding heads means hiring another technician to raise the first number. It is the lever people reach for first and should pull last, because it is the slowest to act and the hardest to undo. The other two levers, process and tools, are cheaper and often enough. Adding heads is also, when the arithmetic says the gap is a person's size, the only lever that works. In that case, a desk that delays it loses the people it already has. This article is part of our Scaling Support Through Growth series. It is about telling those two situations apart with numbers rather than with how tired everyone feels.

It covers why heads are the slowest lever; and the capacity arithmetic in brief, with the article that owns it. It covers the four signals that say the gap is real, backlog age, after-hours activity, leave untaken and the work that stopped. It covers the technician-to-staff question, answered the only honest way; and the false signals that look like a capacity gap and are not. It ends with sizing and timing the ask when the answer is yes; a short war story; and the capacity check to copy, in that order.

Heads Are the Slowest Lever

A hire takes longer than any other fix a desk can make, which is the first reason to be sure before asking. Between the decision and a productive technician sit the approval, the advert, the interviews, the notice period and the first month or two of learning the desk. The usual total is somewhere between four and eight months. The practice of each step is in Hiring Support Staff. A gap that is a person's size today was a person's size six months ago, when the signals were first visible. It will be a person and a half by the time the hire lands if nothing else is done in between. The lever is slow, so it has to be pulled on the signals, not on the crisis.

The second reason is that two cheaper levers exist and the hire is often asked for when one of them was the real fix. A desk that loses tickets because nobody triages does not need a fifth person. It needs a triage rhythm, which costs a week of attention and is the subject of When to Add Process. A desk that spends a day a month tallying a spreadsheet does not need a fifth person. It needs the reports to come out of a system, which is When to Add Tools. A head added to a desk with a broken intake is a faster way to lose tickets. Check the cheap levers first, and say in the ask that you did.

The third reason is that the hire, once made, is permanent in a way the other levers are not. The person who approves it knows that. A process can be relaxed and a tool can be replaced. A technician is a colleague with a mortgage. The finance director is not being obstructive when they ask whether the gap is real. They are asking the right question. The desk's job is to have an answer built from its own numbers, in hours, that a sceptic could check. The rest of this article is that answer.

The Arithmetic in Brief

The sum is hours needed minus hours available, month by month, for at least six months. The article that owns it in full is The Headcount Case. Hours available start from the nominal month and lose, from the real calendar, leave, sickness, training, meetings, projects and daily administration. That leaves most technicians with far fewer ticket hours than they are paid for. Hours needed are tickets opened in the month times the median hands-on time of one ticket, from a sample. The gap is the difference. The headcount case article builds the one-page version a finance director will accept. This article uses the sum only to show what a real gap looks like beside a false one.

At Bluewater, the sum ran like this, in round figures. Each technician's nominal month was about a hundred and sixty hours. After leave averaged across the year, standing meetings, the two projects the desk carried and the daily administration, each had about a hundred and fifteen hours of ticket time. So the four of them had somewhere between four hundred and fifty and four hundred and eighty in an ordinary month and about four hundred in a month with one on leave. Tickets ran between six hundred and forty and seven hundred a month over the six months. The sampled median was forty-five minutes. So the hours needed ran from about four hundred and eighty to five hundred and twenty-five. The gap was somewhere between ten and seventy hours in an ordinary month and over a hundred in a leave month. It was there in every one of the six. The director's spare technician did not exist. The director was not wrong to ask; he had been given a number about somebody else's bank.

Notice what the sum does not say. It does not say Bluewater needs a fifth person, because a fifth person brings about a hundred and fifteen hours and the ordinary gap is less than that. It says the desk is short by something between half a person and a person. It says that the shortfall is permanent rather than seasonal, and that it is worst in leave months. That is a different shape of problem from a desk that is a person short every month. The signals below are how the shape is read, and the shape is what decides the ask.

Signal One: Backlog Age

The first signal is the age of the backlog, and specifically whether it is growing. A desk that is keeping up has a backlog that is young. Most open tickets are days old, a few are waiting on a supplier or a requester, and the oldest one has a reason. A desk that is short of hours has a backlog that ages, because every day the technicians close what arrived today and the older tickets slip another day down the list. Pull the count of open tickets older than thirty days at each month end and put the six figures in a row. If the row climbs, the desk is falling behind at roughly the rate it climbs. The reading of the full report, bands and all, is in Reading an Aging Report.

Backlog age is the signal requesters feel, which is why it is first. Kofi's forty-eight-day ticket from lending is one person's experience of a capacity gap. They asked, nothing happened, they asked again, and they gave up. The count of those is the count of people who have learned the desk is slower than finding a workaround. The desk will not see most of them again. Backlog by age is a standard report in most help desk tools, QuerySolv among them, so this signal costs a few minutes a month to read and nothing to collect. It is the cheapest number in this article and the one most desks do not look at, because it only ever gets worse while you are watching it.

Signal Two: After-Hours Activity

The second signal is work done outside the desk's hours. Every ticket update carries a time. A count of updates stamped after the desk closes and at weekends, taken month by month, shows whether the desk's hours are enough for the desk's work. A technician who answers three tickets at nine in the evening is lending the desk three hours it does not have. The queue looks healthier for it, which is exactly the problem. The gap is being paid by a person and hidden from the number. This signal and the two beside it are owned by Measuring Load Without Surveillance, which also says how to read them without turning them on individuals. That rule applies here in full.

Count the desk's after-hours updates as one figure, never by person. The question is whether the desk is short of hours. The answer is the same whether the evening work is spread across four people or done by one. If it is done by one, that is a conversation about distribution and pacing, which belongs to the burnout series. A per-person figure read for any other purpose is ranking, which the library forbids. A rising after-hours count beside a rising backlog age is the strongest pair of signals in this article. It means the desk is falling behind even while borrowing time it was not given.

Signal Three: Leave Untaken

The third signal is leave that was not taken, and the leave that was taken showing up as a gap. Pull the desk's leave balance at the end of the year, or whenever the organisation resets it. Look at how much was carried over, lost or paid out. Technicians who do not take their leave are doing it for one of two reasons: nobody can cover them, or they know what the queue will look like when they return. Both are capacity statements. The second one is a technician forecasting the backlog signal for you.

The other half of this signal is what happens when leave is taken. If a technician's fortnight away produces a pile that takes the rest of the desk a month to clear, the desk has no slack. A desk with no slack is a desk whose capacity is exactly its work on a good month and less on every other kind. The handover and coverage practice that limits the pile is in Vacation Coverage Without the Pile; what this article takes from it is the diagnosis. A desk that cannot absorb one person's leave is one person short for the leave months. The sum should be run on those months separately, because that is where Bluewater's hundred-hour gaps lived.

Signal Four: The Work That Stopped

The fourth signal is invisible in the queue, which is why it is last and why it is missed. A desk short of hours stops doing everything that is not a ticket. The knowledge base is not updated. The problem record that would remove a category of tickets is never opened. The quarterly review of the spares cupboard does not happen. The prevention hour, the one the proactive series in Buying the Time spends a whole article on protecting, is the first thing cancelled and the last thing restored. The queue, meanwhile, looks much as it did, because the queue only shows tickets, and the desk is still closing tickets. It has stopped doing anything that would reduce them.

Make a list of the non-ticket work the desk is supposed to do and tick, each month, what actually happened. Most desks can name eight or ten items: the knowledge base, the problem review, the spares check, the report, the onboarding improvements, the monitoring tidy-up, the training afternoon. A month where two happened is a signal. A quarter where none happened is a desk that has become a queue with people attached. The desk will stay one until the hours come from somewhere. This signal matters for the ask, too. The fit between the gap and one person, which the headcount case requires, is often made up of half a person of ticket gap and half a person of work that stopped. The second half is the part the finance director has never been told about.

The Technician-to-Staff Question

Everyone asks for the ratio, and the honest answer is a set of worked examples with the spread left in. No industry ratio fits every organisation, because the number of staff is a poor measure of the work. The devices per person, the number of systems, the number of sites, the shift pattern, the maturity of self-service and the amount of non-ticket work the desk carries all move the ratio. They move it by more than a factor of three between organisations of the same size. The fictional desks of this library, when their arithmetic was run, came out between about one technician for every forty staff and one for every hundred and fifty. The spread is the finding. A desk that quotes a ratio from somewhere else is quoting a number about a different company.

Meridian Parts has about a hundred and twenty staff, most of them on a shop floor sharing a dozen terminals, one building, three business systems and no home working. Its arithmetic said one technician with a little slack: roughly one per hundred and twenty. Larkspur Media has a hundred and forty staff, each with a laptop, two monitors and a home office, a stack of creative applications with their own licensing, and a remote half. Its two technicians are stretched. The ratio is roughly one per seventy, and heading for one per sixty as the headcount grows. Harbourview Clinic has about three hundred staff across three sites, a clinical system with its own supplier, and shifts seven days a week. Its four technicians cover the hours only with a rota that counts as a fifth. That is roughly one per seventy-five on paper and nearer one per sixty in practice. Bluewater, as above, has three hundred staff, one building, a regulated environment and four technicians. It is between half a person and a person short.

Two things drive the spread more than any other. The first is devices and systems per person: a laptop, a phone, a dock and ten applications generate several times the tickets of a shared terminal and two applications. The desk's volume per hundred staff, which is the base the forecasting article in Forecasting Volume Simply teaches you to carry, is the number that captures it. The second is everything the desk does that is not a ticket: onboarding, projects, the second site, the monitoring, the vendor calls. A desk that carries two projects a quarter has a technician's worth of hours fewer than its headcount suggests. The ratio does not know that. (The ratio does not know anything. It is a division sum that has been promoted.)

Use your own ratio for one thing only: tracking yourself over time. Bluewater's one per seventy-five, written down with the month and the method, is a number Kofi can compare with Bluewater's one per ninety next year and ask what changed. It is not a number to compare with Harbourview's. It is not a number to carry into a budget meeting, where it will be met with a different number from a different source. Then the meeting will be about which number is right rather than about the desk. Take hours to the meeting. Hours are yours.

False Signals

Several things look like a capacity gap and are not. A desk that asks for a head on one of them will be found out, usually by the person who approved it. The first is the bad month: a rollout, a migration, an outage that generated three hundred tickets in a week. The backlog age climbs, the after-hours count climbs, and then both fall back, which is why the sum runs over six months and not one. The second is the repeat ticket: a category that has doubled because one thing is broken and has not been fixed. That is a problem record, not a hire. The hunt for it is in Finding the Repeat Tickets. A desk that hires to absorb a problem it could have fixed will absorb it forever.

The third false signal is the distribution problem. The desk as a whole has the hours, but one person is drowning while another is quiet. That is because tickets are routed by habit or by who answered the phone. The desk-level numbers look fine and one technician's evenings do not. The fix is in how work is claimed and shared, in Round-Robin vs Skill-Based vs Claiming, not in a fifth chair. The fourth is the broken intake: tickets are old because nobody triaged them, not because nobody had time to fix them. The test is whether the old tickets were ever assigned to a name. Old and unowned is a process gap. Old and owned, with notes showing the owner was working on something else, is capacity.

The fifth false signal is the loudest, and it is the feeling. Everyone is tired, the queue feels endless, and the obvious answer is more people. The feeling is real and it is also what a broken process feels like from inside. The four signals exist so that the feeling can be checked. Where they agree with it, the ask is strong. Where they do not, the feeling is telling you something else is wrong, and the something else is usually cheaper than a hire and a great deal faster.

When the Answer Is Yes

When the signals and the sum agree, size the ask to the shape of the gap rather than to a whole number of people. A gap that is permanent and a person's size is a hire. A gap that is half a person every month is a hire only if the work that stopped fills the other half. In that case, the ask should say so, with the list. A gap that appears only in leave months is cover, a shared technician with another team, or a change to how leave is scheduled. Asking for a head to fix it will be refused by anyone who reads the six months. The sizes of ask, from the whole person down to the shift of hours, are set out in the headcount case. The smallest ask that closes the gap is the one that gets approved.

Ask early, because the lever is slow. A desk whose signals are visible in March and whose hire starts in November has spent eight months at a gap. The cost of those months is not only the backlog. It is the technician who leaves in September, which turns a one-person gap into two and resets the hiring clock. The right time to ask is when the signals have agreed for a quarter and the forecast says they will go on agreeing. That is what the year-ahead plan in Planning a Year Ahead is for. The plan turns "we need someone" into "we will need someone by September, so we should advertise in April". Finance directors are considerably more generous to a request that arrives before the crisis, because it looks like management rather than panic. It is also, as it happens, management.

Say what the hire will change, in the same numbers. If backlog age over thirty days is the signal, say what it will fall to. If after-hours updates are the signal, say they will stop. If the work that stopped is the signal, name the three items that will restart. Then, six months after the hire starts, report those numbers back to the person who approved it. The desk that does this gets its next ask approved on a single page. The desk that does not gets asked, next time, what happened to the last one.

A Short War Story: The Borrowed Ratio at Pennant Brewing

Pennant Brewing was a regional brewer with about two hundred and twenty staff, two sites and a three-person desk. It asked for a fourth technician in the spring. Its backlog over thirty days had climbed from twenty to ninety tickets in six months, and it had a leave balance nobody had touched. The finance director declined the request, citing a ratio he had read in a supplier's brochure that put the desk comfortably within the normal range for its size. The desk lead, Marta, had no hours to put against the brochure because she had never run the sum. In August the longest-serving technician resigned, citing the evenings. A second handed in notice in October. By that point the backlog was over two hundred and the brewery's new bottling line went live with one technician and a contractor. Pennant hired three people over the following seven months to replace two, and paid agency rates through the winter. The finance director, when he eventually saw the hours, approved a fourth head he could have approved for a quarter of the cost in the spring. A ratio from a brochure is a number about somebody else's company, and Marta's hours were the only number about hers.

The Capacity Check

The check below is one page, done at each month end by whoever runs the desk, in under half an hour once the reports are to hand. Section A is the sum in brief; the full version with the one-pager lives in the headcount case and is where the ask gets written. Section B is the four signals, each a number. Section C is the false-signal check, which the desk completes before anyone outside it sees the page. Section D is the conclusion, and it has four possible answers, not one. Keep six in a row before you ask for anything. The row is the evidence, and a single month is an anecdote with a table.

CAPACITY CHECK — ONE DESK, ONE MONTH
Desk: ______________________   Month: ____________   Prepared by: ______________

A. THE SUM (full method: the headcount case)
   A1 Hours available this month (nominal, less leave, meetings, projects, admin) ... ______
   A2 Tickets opened this month ................................................... ______
   A3 Median hands-on hours per ticket (sampled on ________) ....................... ______
   A4 Hours needed (A2 times A3) .................................................. ______
   A5 Gap this month (A4 minus A1) ................................................ ______
   A6 Gap, previous five months ......................... ____ / ____ / ____ / ____ / ____

B. THE SIGNALS
   B1 Open tickets older than thirty days at month end ............................ ______
      previous five months ................................ ____ / ____ / ____ / ____ / ____
   B2 Ticket updates outside desk hours, whole desk ............................... ______
   B3 Leave: days carried over or lost this year ____   weeks to recover after last leave ____
   B4 Non-ticket duties on the list ____   done this month ____   done last quarter ____

C. FALSE-SIGNAL CHECK (tick when ruled out)
   [ ] Not a single bad month: the gap and B1 are present in at least four of the last six
   [ ] Not a repeat ticket: no single category explains the growth; problem review done
   [ ] Not distribution: the desk-level numbers, not one person's, show the gap
   [ ] Not intake: the old tickets were owned by name and worked, not left unassigned
   [ ] Cheap levers checked: triage rhythm in place; reports come from the system

D. CONCLUSION (tick one)
   [ ] Keeping up: no ask; repeat next month
   [ ] Leave-month gap only: ask for cover or a leave-schedule change, not a head
   [ ] Half-person gap: ask for a head only with the restarted duties from B4 named
   [ ] Person-sized gap for six months: write the headcount case now; lead time ____ months
   Decision taken on: ____________   Reviewed with: ______________

Kofi ran the check for the six months behind him on the Monday. It took the afternoon because the median had never been sampled. The conclusion was the third box: a half-person gap, permanent, worst in leave months, with seven of nine non-ticket duties undone for two quarters. The ask he wrote was for one technician, with the knowledge base, the problem review and the spares check named as what the other half of the person would restart. He took the ask to the operations director with the six-month row on top. The director read the row, asked where the one-per-hundred figure had come from, and did not wait for an answer.

Frequently Asked Questions