Field note · 18 November 2026
Sortation throughput: why the items-per-hour number lies
Every sorter datasheet has a rate on it, and almost no installation reaches that rate. The gap is not marketing: it is the difference between a machine's peak and a line's sustained output.

01Four numbers that get confused
The peak rate is the sorter's mechanical maximum, calculated from pitch and speed with no gaps. It is a real number and it is almost never reached.
The design rate is the peak after derating for gaps, induction losses and discharges. It is the number an integrator should quote, and it is usually 55 to 70 per cent of peak.
The sustained rate is what the line delivers over a shift, including stops, jams, breaks and the ramp at the start. It is the number the operation actually feels.
The peak hour is the busiest hour of the day, and it is usually 1.4 to 1.8 times the average. A line designed to the average will fail in the peak; a line designed to the peak will be idle most of the day.
02The induction derate
A sorter's rate is set by how many gaps its induction can fill, and every induction has a cycle time that includes the scan, the gap decision and the transfer.
If one induction can place a parcel every 1.2 seconds, it delivers 3,000 parcels per hour - and that is before the parcel stream starves for ten seconds in every minute.
Four inductions at 3,000 each is 12,000 theoretical, and a real line with four inductions and a good buffer will sustain 7,000 to 8,000 across a shift. The derating is not a failure; it is what a line is.
03The discharge derate
Every discharge has a probability of failure, and the failures come back as recirculations. A two per cent recirculation rate means two parcels in every hundred pass the sorter twice, which adds load to the induction and the chutes.
The reason recirculation matters more than it looks: the parcel that comes back occupies a cell on the second pass, so the line pays for it twice in the throughput budget.
Chute fullness is the second discharge loss. A chute that is 90 per cent full rejects parcels, and a rejected parcel recirculates. The fill-level logic that closes a chute early is a throughput decision disguised as a safety rule.
04Designing to the peak without buying the peak
The usual compromise is a line sized above the average and below the peak, with a manual or a buffer area that absorbs the difference. That is cheaper than a line sized to the peak and it survives most days.
The buffer is the honest answer to the peak-hour problem. A pre-sort buffer that holds an hour of peak volume lets the line run at its sustained rate while the peak passes through it.
The alternative is overtime or a second shift, and both are more expensive than the buffer in any operation that has the floor space.
05What to ask for in a rate guarantee
A rate guarantee should name the parcel mix, the induction count, the chute configuration and the recirculation limit. Without those four, the number is a claim rather than a commitment.
It should also name the measurement window. A fifteen-minute test proves the machine; a four-hour test proves the line, and the second is the one worth paying for.
And it should name the ramp. A new line takes weeks to reach its sustained rate while the operators learn the exceptions, and a guarantee that starts on day one is a guarantee that will be argued about.
06The measurement window
A rate measured over fifteen minutes measures the machine. A rate measured over a shift measures the line. The difference is stops, jams, breaks and the ramp.
The honest measurement window is a full shift at the design rate, with the shift's own team and the shift's own mix. Anything shorter is a demonstration.
The window should be reported with its stop list. A line that ran at nine thousand items per hour and stopped four times for six minutes each has a sustained rate of about eight thousand, and the two numbers describe different things.
The report should separate planned stops from unplanned ones. A lunch break is not a fault, and a report that mixes them makes every improvement look smaller than it is.
07Ramp and the learning curve
A new line does not start at its sustained rate. It starts at sixty to seventy per cent of it and climbs as the team learns the exceptions.
The ramp is usually four to eight weeks for a sorter, and the curve is steeper if the exceptions are written down and trained than if they are learned by watching.
The contract should recognise the ramp. A rate guarantee that starts on day one is a guarantee that the integrator will fail, and a guarantee that starts at week six is one both sides can plan around.
The ramp is also the period when the parcel mix is still being discovered. A line designed for cartons that starts receiving bags will show a rate drop that no amount of training fixes.
08The second shift economics
The alternative to a faster line is a longer day. A second shift on the same line costs labour and supervision, and it costs nothing in capital.
The comparison is worth doing honestly: a line at eight thousand items per hour for one shift delivers sixty-four thousand items a day. The same line on two shifts delivers a hundred and twenty-eight thousand for the price of the second crew.
Most operations choose the second shift first, because the capital is already spent and the marginal cost is visible. The faster line is chosen when the building is full or the labour market is tight.
The decision belongs in the design brief, not in the third year of operation. A line built for one shift and run for two will be maintained at half the interval it was designed for.
09What the control system should report
Parcels inducted, parcels discharged, parcels recirculated, stops and jam durations. Five counters that, reported per hour, describe the line without any interpretation.
Most control systems report the first two and none of the rest. The recirculation count is the one that explains a rate shortfall, and it is usually available from the chute sensors if anyone asks for it.
The report should be hourly rather than per shift. A shift average hides the peak hour, and the peak hour is where the design is tested.
The counters should be exported, not photographed off a screen. A CSV is a report that survives the person who produced it.
10The cost of a percentage point
One percentage point of recirculation on a line running eight thousand parcels an hour is eighty parcels an hour, which is roughly seven hundred a shift, each handled twice.
At a handling cost of a few pence per touch and a labour rate per hour, the number is small per parcel and large per year. It is also a number that can be reduced by fixing two chutes.
The arithmetic is the argument for measuring recirculation at all. A line that does not measure it has no way to know whether the percentage point is a chute problem or a parcel-mix problem.
The same arithmetic applies to stops. A line that stops twelve times a shift for four minutes loses forty-eight minutes, which is eight per cent of a ten-hour shift and a visible number in the weekly report.
11The rate review that changes nothing
A rate review that reports a number and no cause is a review that changes nothing. The output should be three causes, ranked by parcels lost, and one action for each.
The causes are almost always the same three: induction gaps, chute closure, and jam recovery. The numbers change, the list does not.
Reporting the same three causes every month is not a failure of imagination. It is evidence that nobody has fixed them, and that is a management finding rather than an engineering one.
Where the rate calculation needs a second line rather than a bigger one, the cross-belt sorter rated for the higher pitch is the unit that carries the extra induction without changing the pitch.
What to check
- Peak, design, sustained and peak-hour rates written separately
- Induction count and cycle time verified by observation
- Recirculation limit named in the acceptance criteria
- Rate test run for at least four hours on the real mix