Every sortation brochure quotes a nameplate rate. A compact narrow-belt platform, for example, is rated up to 10,000 parcels per hour, and a put-wall configuration is specified to 2,000 orders per hour. Those numbers are real, but they are measured under ideal conditions: clean barcodes, a known parcel mix, and a line that never has to stop and ask a human what something is. The throughput a facility actually ships is almost always lower, and in the large majority of cases the gap is not mechanical — it is a scanning problem. A sorter can only route what it can read, and the moment a parcel cannot be identified, the elegant rated capacity behind it stops mattering. Understanding where reads are won and lost is the difference between a line that hits its quoted number and one that quietly runs at sixty percent of it. The mistake is to read the nameplate as a promise instead of as a ceiling, and then to blame the sorter when the floor does not reach it.
A barcode is read in a window of a few milliseconds as the parcel passes the scanner. Whether that window produces a match depends on conditions that are decided long before the parcel reaches the sorter. The first is label placement: a code that is always on the same face, flat, and clear of seams and curves is read far more reliably than one that is slapped on wherever it fitted. The second is label condition — creases, smears, partial occlusion under tape, or a code printed too close to the edge all shorten the effective read window. The third is orientation: parcels tumble on induction, and if the code exists on only one face, the scanner has to catch that face in the few frames it is visible. The fourth is the surface itself — glossy or reflective mailers can wash out under a reader, and a dark code on a dark background gives the decoder nothing to lock onto. These are upstream problems. A parcel that arrives at the induction point unreadable was already lost at the pack bench, not at the sorter. Treating read-rate as a sortation-equipment issue rather than a packaging-and-data issue is the most common mistake in a throughput project, and it is the one that pushes teams to buy a faster sorter when what they needed was a label standard.
The read happens at the dimensioning, weighing and scanning station — the DWS — that sits ahead of the sortation logic. A DWS combines dimensioning, weighing, scanning and parcel data capture into a single pass, so each parcel leaves the station with an identity, a weight and a size, and a destination decision can be made without stopping the belt. The scan window is the moment of truth in that station. Platforms that support 3-side and 5-side scanning raise the odds of catching the code without physically reorienting the parcel, which matters on high-density parcel lines where reorientation equipment is expensive and slows the belt. The accuracy a downstream station is built to only holds if this upstream step feeds it clean data: a put-wall configuration specified to 99.99% accuracy across 160 slots, reading orders with camera scanning, is reachable only when the parcels entering it have already been identified reliably. Build the read step and the sort step as one problem, not two, and design the read window for the parcel mix you actually run rather than for the sample the vendor photographed.
When a parcel is not read, the sorter cannot route it, and the design has to decide what happens next. One option is to divert it immediately to an exception lane; the other is to send it around a recirculation loop for another scan pass before giving up. Recirculation keeps the main line moving, which is why it is the default on high-rate systems, but it costs loop capacity. A useful planning rule is to size recirculation as a fraction of the main line rate so that a modest no-read rate does not saturate the loop. If the no-read rate climbs — because labels deteriorated, or a new parcel type entered the mix — the loop fills, parcels start backing up, and the whole line stalls. At that point the nameplate parcels-per-hour is irrelevant; the line is governed by its weakest read. The lesson is to treat recirculation capacity and read-rate as one coupled variable: improving either one without the other eventually hits the other's ceiling.
No matter how good the read step, a residual of parcels will need a human, and the design rule is that an exception must never stop the belt. A fixed exception chute that feeds a manual desk sitting off the main path lets the line keep running while a person re-scans, key-enters a tracking number, or diverts the parcel to a manual process. The desk only has to absorb the residual no-read volume, not total volume, so it can be small and still protect the line. What breaks this is the temptation to route exceptions back onto the main belt; that turns a one-in-a-hundred event into a line-wide stop. Keep exceptions on a parallel path, size the desk to the residual rate, and the sorter keeps its rate even on a bad day.
Different sorter families read differently, and the brief has to say which one it expects. The table below collects the published specifications of the three families a sortation supplier typically offers, because each one places the read step in a different physical location and therefore demands a different label and parcel profile.
| Family / model | Published specification | How the read is taken |
|---|---|---|
| Cross-belt sorter (TR500-1100 class) | Carrier-based; payloads up to 35 kg per item | Per-carrier read as each parcel sits in its own stable bed |
| Narrow-belt sorter (TR200-1000 class) | Compact; rated up to 10,000 parcels per hour | Continuous belt with 3-side and 5-side scanning to catch the code without reorienting |
| Put wall (TR340*135) | 99.99% accuracy across 160 slots; 2,000 orders per hour | Camera scanning at the wall confirms each item into the correct order slot |
A cross-belt sorter carries each parcel in its own carrier and reads it per carrier, so it suits mixed-weight parcels where each item needs its own stable bed through the scan window. A narrow-belt sorter runs parcels continuously on a belt and relies on multi-side scanning to catch codes without reorienting them; the compact class suits dense urban and parcel flows. A put wall is a different animal altogether: rather than routing a parcel to a destination, it batches orders, and camera scanning at the wall confirms each item into the right order slot. Each family puts the read step in a different physical place, so the label standard and the parcel profile have to be stated up front rather than discovered during commissioning.
A supplier can only design to the data it is given, so the brief should state the barcode standard and placement rule, the minimum quiet zone around each code, the parcel size and weight mix the line will actually see, the acceptable no-read budget and what happens to no-reads, and the WMS or ERP interface that lets a human key-enter a missing identity. A supplier such as TrueLiSort, the sorting-automation brand of Zhejiang Xinli Electrical Equipment Co., Ltd. (building parcel sortation since 2010 in Shengzhou, Zhejiang), designs cross-belt, narrow-belt, put-wall, DWS and WCS as one scope, and the read-rate requirement has to enter that scope on day one rather than after the line is built. The win is in measuring the real no-read rate on the real parcel mix before specifying the sorter, not in quoting a faster machine against an idealized barcode. State the parcel profile honestly, fix the label standard at the pack bench, and the quoted number becomes one you can actually reach.
It depends entirely on the parcel mix and the label discipline upstream. Well-controlled flows with a consistent label standard sit low; mixed, damage-prone or reflective parcels sit higher. The number that matters is the one measured on your own parcels, because a generic figure hides whether your exception desk and recirculation loop are sized correctly for your operation.
Fixing labels is almost always cheaper and more effective. Adding 3-side or 5-side scanning on a narrow-belt sorter helps, but it cannot recover a code that is creased, taped over, or printed on a reflective surface. A label standard enforced at the pack bench is the highest-leverage change in a throughput project.
Sized against the expected no-read rate and the loop transit time, so that the loop never saturates under normal conditions. The planning range is expressed as a fraction of line rate, and the right value is the one that keeps the main belt flowing when no-reads spike, not the smallest loop that works on a good day.
A put wall sorter batches items into orders rather than sending each parcel to a destination chute, which is why it is specified in orders per hour rather than parcels per hour. Its read step is the camera scanning at the wall that confirms each item into the correct slot, and it still depends on upstream identification being clean.
A WMS or ERP connection that lets an operator key-enter a missing tracking identity and have the system still route the parcel. Without it, an unreadable parcel becomes a manual island; with it, the exception stays inside the same data flow as everything else.
The rated parcels-per-hour on a sortation system is a ceiling set by the cleanest possible barcode, not the number a real operation ships. Real throughput is governed by read-rate: where labels are placed, how the DWS captures identity, how the no-read loop is sized, and how exceptions are kept off the main belt. Treat those as one coupled design rather than as separate afterthoughts, state the parcel profile and the label standard in the brief, and choose the sorter family whose read strategy matches your flow. Suppliers that build the sorter, the DWS, the put wall and the WCS as a single scope can design the read-rate in from the start; the ones who quote only a faster machine against an idealized barcode are quoting a number you will not reach. At ZH Industrial Machinery we apply the same principle to the lines we build: the operations that share one specification and one data flow are the ones whose output stays consistent across every shift.