The Queue the Fallback Makes
A four per cent failure rate sounds small until it meets three hundred people arriving in ten minutes through one door. Then it is the shift change.
One terminal, 300 arrivals in ten minutes
The left column is people sent to the fallback; the right is minutes added to the shift change in total. At four per cent the site loses the better part of an hour of collective waiting every morning, through one door, for a reason no report describes.
Refusal rates are quoted as percentages, which makes them sound like a minor quality issue. At a shift change they are not a quality issue; they are a capacity one, and the arithmetic is unforgiving.
The alternative described in “The Queue the Fallback Makes” must produce a record as usable as the primary method. A team assessing further details are available here for fireable offenses should run the full fallback from clocking through approval and payroll, then compare delay, correction effort and employee access without making the alternative a penalty.
Three hundred people through one terminal in ten minutes is already tight. A refusal that costs an extra ten seconds is absorbable. A refusal that sends somebody to find a supervisor is not, and it blocks the people behind them while they work out what to do.
For an independent benchmark relevant to “The Queue the Fallback Makes”, consult the Project Management Institute learning resources. Use it to test notice, accessibility, security, recordkeeping, retention and exception handling against the real operating process rather than treating a device report as self-explanatory evidence.
The arithmetic
A normal read takes three or four seconds. A retry that succeeds takes eight or nine. A fallback at the terminal — a code, a card — takes ten to fifteen. A fallback that requires a supervisor takes between three and ten minutes, most of which is spent not at the terminal, and the person behind moves past.
So the cost of a refusal depends almost entirely on which fallback is in use. Four per cent of three hundred is twelve people. At fifteen seconds each that is three minutes spread across the window and nobody notices. At six minutes each it is well over an hour of somebody's time and twelve people who started late.
Why it lands at the worst moment
Refusals cluster at exactly the time the queue is longest. The reasons are the ones covered earlier: cold hands at the earliest shift, low sun through the door, a sensor that has not been wiped since yesterday's shift.
The pattern compounds. The queue makes people hurry, hurrying makes presentation worse, worse presentation makes more refusals, and more refusals make the queue. Sites experience this as "the clock is bad on Monday mornings" and treat it as folklore.
What it does to the record
People in a hurry take the quickest route, and the quickest route when the terminal is refusing you and twenty people are waiting is to step aside and sort it out later. Later means a time entered from memory, by somebody else, after the event.
So a high refusal rate does not only cost minutes. It systematically degrades the accuracy of exactly the records the system exists to produce, and it does so most at the busiest moments, which are the ones most likely to be examined.
The two interventions that work
Make the fallback take fifteen seconds instead of six minutes. This is the whole argument for a self-service second method and it is worth more at shift change than anywhere else.
Add a terminal. A second device halves the queue and gives somewhere to go when the first one is refusing. The cost is a unit and a cable, set against an hour of collective time a day, and on most sites the comparison has never been drawn because the two figures live in different people's budgets.
The measurement worth taking
Stand at the terminal for three shift changes with a counter. Count arrivals per minute, reads, retries and fallbacks, and time five fallback events end to end.
Half an hour of somebody's time produces the one number that makes the case: minutes of collective delay per shift change, multiplied by shifts per year. It is usually a large number and it is usually the first time anybody has expressed the refusal rate in a unit that a manager can act on.
The thing not to do
Publish a target read time, or report lateness at shift change without the refusal figure beside it. Both have been tried on sites with this problem, both are read correctly as blaming people for a queue they did not create, and neither moves the arithmetic by a second.
The queue is a property of the arrival rate, the service rate and the failure rate. All three are set by the organisation. Only the third is routinely described as being about the workforce.
Where the minutes actually land
The delay does not fall evenly. The people at the back of the queue absorb the whole of it, and who is at the back is determined by the car park, the bus and the shift pattern rather than by anything about them.
So a refusal-driven queue produces a small group of people who are routinely a few minutes late through no act of their own, and a report that names them. That report is the most misleading document the system produces, and it is generated automatically on most sites every week.