Days to Minutes

Rosters Are Written in Days, Forecasts Are Written in Minutes

Ask a resource planner how much cover a queue needs next month and the answer starts as a number of days: seven-day opening, a four-week rotation, a fortnight of extra weekend support. Ask the forecasting model the same question and every input it wants is a minute, because that is the only unit where demand and supply can be compared at all.

Contacts do not arrive in days. They arrive in fifteen or thirty-minute intervals, each with its own volume and its own average handle time, and the roster has to answer that shape rather than a daily total. Converting the coverage window into minutes is the step that lets a shift pattern be checked against a workload figure instead of against intuition.

Conversion factor: 1 d = 1 440 min, so multiply days by 1 440. Round-the-clock cover for a full week is 10 080 min per staffed seat, and a four-week rotation is 40 320 min. If a forecast calls for 60 480 agent-minutes in that week, you are being told to keep six seats occupied without a break.

Why Capacity Planning Drops Down to the Minute

Demand arrives interval by interval

Volume is forecast per fifteen or thirty minutes, not per day. A queue with 4 000 daily contacts can be comfortable at 06:00 and two agents short at 11:15, and only the interval view shows it.

Supply is contracted in shifts and days

Employment terms, rotation patterns and holiday entitlement are all expressed in days and weeks. Nothing in a contract is written per interval, so the roster side of the equation always needs converting first.

Handle time is recorded in seconds

Average handle time comes off the platform in seconds: 280 s of talk plus 60 s of wrap. Multiplied by forecast volume it becomes a workload, and the natural unit for that product is minutes.

The two sides meet nowhere else

Workload in minutes against coverage in minutes is a comparison you can defend in a budget meeting. Days against seconds is not, which is why the conversion happens before any staffing model is opened.

Sizing a Coverage Window Before Anyone Is Rostered

Start from the window the business has committed to, express it in minutes, and only then argue about headcount. The order matters, because the window is usually fixed and the headcount is not.

Enter the length of the window you have promised

Seven for a full week of opening, fourteen or twenty-eight for a rotation cycle, ninety-one for a peak trading block. The minute total updates as each digit lands, and spaces typed inside a long number are ignored rather than rejected.

Multiply by the seats the queue has to hold open

One seat covering the window continuously supplies that minute total. Two seats supply twice it. Compare the result against the workload your forecast predicts and the gap, in minutes, is the thing a hiring case has to close.

Reverse it when the minutes are already known

A quarterly forecast handed over as 131 040 agent-minutes means nothing until it is scaled. The swap control (↔) puts minutes on the left and gives 91 d, which is one seat held open for the whole quarter.

Move the figure into the staffing model unformatted

Copying returns digits with nothing attached, so a cell in a capacity workbook or an input field in an Erlang tool accepts it without the thousands spacing having to be deleted first. Ctrl + C over a selection inside the box strips the spacing too.

Coverage minutes are not paid minutes: 1 440 minutes of cover in a day is what the queue needs, not what one employee delivers. Contracted hours, breaks, training and absence all sit between the two, and treating the converted figure as a headcount answer will understaff the queue badly.

Roster Cycles Expressed as Agent-Minutes

Common planning windows with the minutes a single continuously staffed seat has to supply across each one. Multiply the third column by the number of seats the interval forecast demands to get the workload a roster is really being asked to cover.

Coverage window	Length (days)	Minutes per staffed seat	Planning note
One operating day, queue open 24 hours	1 d	1 440 min	Split into 96 half-hour intervals for forecasting
Weekend cover, Saturday and Sunday	2 d	2 880 min	Usually the first window to be outsourced
Monday to Friday span	5 d	7 200 min	The baseline most legacy rosters were built on
Full week, round the clock	7 d	10 080 min	Needs night differentials and a follow-the-sun option
Two-week roster cycle	14 d	20 160 min	Short enough to reforecast, long enough to rotate shifts
Three-week rotation	21 d	30 240 min	Fits three shift bands without repeating a pattern
Four-week roster cycle	28 d	40 320 min	Aligns with monthly reporting but drifts off calendar months
Peak trading block, thirteen weeks	91 d	131 040 min	Recruit and train well before the block opens

The step worth staring at is the one from five days to seven. Extending a weekday queue to full weekend opening adds 2 880 minutes per seat per week on top of 7 200, a 40 per cent increase in coverage before a single extra contact has been forecast. That is the number a request for "just open Saturdays" is actually asking the budget to absorb.

What the Converter Adds to a Capacity Plan

Test a rotation pattern in seconds

Try 14, 21 and 28 in the left box one after another and the minute totals for each candidate cycle appear without a workbook being touched.

Read a forecast total back as cover

Flip the direction and a demand figure handed over in minutes becomes the number of days a single seat would have to be held open to absorb it.

Part days convert cleanly

A window of 0.5 or 0.375 of a day, the fraction a single shift represents, keeps its precision instead of collapsing to a whole-day figure.

Model inputs paste without cleanup

The copied value carries no unit and no separators, which matters when the destination is a numeric field that would reject a formatted string.

Questions From Workforce Planning

Why does an Erlang calculator want minutes rather than a daily total?

Because the model answers one interval at a time. Erlang C takes the number of contacts arriving in a short window, the average handle time and a target answer speed, then returns the agents needed for that window alone. Feed it a day and the arrival pattern disappears: a queue that peaks at three times its overnight rate would be sized for an average that occurs at almost no point in the day. Convert the day into 96 half-hour slots, forecast each one, and the roster you build actually matches the shape of the demand.

How many productive minutes does one full-time agent really supply?

Far fewer than the contract suggests. A 40-hour week is 2 400 paid minutes, and a 37.5-hour week is 2 250. Strip out breaks, briefings, coaching, system time and absence, and what reaches the queue is commonly somewhere near 1 700 minutes a week. Across a year with five weeks of leave, that is roughly 47 working weeks, so one head delivers on the order of 80 000 productive minutes rather than the 112 800 the contract implies. Planning against the contracted number is the single most reliable way to build a roster that fails in week one.

Where exactly does shrinkage come out of the minute total?

Shrinkage is applied after the model has told you how many agents must be on the phones, not before. If the interval requires 30 agents available and total shrinkage runs at 30 per cent, you roster 30 divided by 0.70, which is 43 people. Applying the reduction to the demand side instead produces a smaller number and a queue that misses its service level every day. Most operations split the figure into planned shrinkage such as leave and training, and unplanned shrinkage such as sickness and late starts, because the two behave very differently across a roster cycle.

Is occupancy simply the share of rostered minutes spent on contacts?

Close, but it is measured against time logged in and available, not against the full shift. If an agent is available for 400 minutes in an interval group and spends 340 of them handling contacts and wrapping them up, occupancy is 85 per cent. The remaining minutes are waiting time, and they are not waste: a queue with no idle time cannot absorb a spike, so it answers slowly and abandons. Sustained occupancy above the mid-eighties is a well-known driver of attrition, which is why planners treat it as a ceiling rather than an efficiency target to be maximised.

How do I turn an average handle time into the minutes a forecast needs?

Multiply forecast volume by handle time and keep everything in the same unit. Take 8 000 contacts across a day at 360 s each: that is 6 minutes per contact and 48 000 minutes of workload. A day gives one seat 1 440 minutes, so 48 000 divided by 1 440 is 33.3 seats occupied continuously, and at an 85 per cent occupancy ceiling the figure rises to about 39. Shrinkage is then applied on top to get rostered heads. Every step in that chain is a minute figure, which is exactly why the day count is converted at the start.