A barcode on a box that never passes a reader is not a barcode, it is a decoration. In an automated environment the symbol has to pass through the reader’s field, in an orientation the reader can decode, at a speed it can track. Those are geometric requirements, and they are satisfied or not satisfied by the placement decision. When a scan rate is poor, orientation is the first thing to check and the cheapest thing to fix.
What the reader actually sees
A fixed reader has a field of view of finite width and height at a finite distance. Anything passing outside that field is invisible regardless of how good the label is. Where boxes vary in height, or are presented on a conveyor at inconsistent positions, the symbol moves in and out of the field and the read rate becomes a function of the box rather than of the reader. A wider field or a second reader solves it; so does constraining where the label can be.
Bars across the scan line
A linear barcode is read by a scan line crossing the bars. If the bars are parallel to the scan line, nothing is decoded. On a box travelling along a conveyor, the bars should be vertical if the scan is horizontal, and the practical rule is that the bars should be perpendicular to the direction the reader sweeps. Rotating a label ninety degrees is the single most common fix for a symbol that never reads, and it costs nothing.
Two orientations means two labels
Where a box can arrive at the reader in either of two orientations, one label reads half the time. The options are a second label on the opposite face, a label wrapping a corner, or an orienting mechanism on the line. Two labels is cheapest and works; wrapping a corner works if the surface allows it and the quiet zones survive; orienting the box is a mechanical project. Most sites end up with two labels, and most wish they had started there. The surface underneath matters just as much – see barcodes on a curved box surface.
Presentation angle and specular reflection
A label angled away from the reader presents a foreshortened symbol and, if the surface is glossy, a reflection of the light source. Both reduce decode reliability. Where a label has to be read at an oblique angle, a matte label stock and a slightly larger symbol help, and moving the reader closer to perpendicular helps more. The geometry is usually constrained by the line, which makes the label position the adjustable variable.
Speed and the exposure problem
At speed, a camera-based reader has a limited exposure time and a moving symbol blurs. The blur is worse the faster the line and the smaller the symbol. Where a line has been sped up and the read rate has dropped, increasing the symbol size usually recovers it, because a larger symbol tolerates more blur. Increasing the label size is cheaper than slowing the line and cheaper than upgrading the reader.
Where labels get covered
A label on the side of a box is covered when boxes are stacked side by side. A label on the end is covered when boxes are staged end to end. A label near the base is covered by the conveyor rail. A label near the top is covered by a lid or a band. Each of those is obvious once stated and each one accounts for a share of real scan failures. The only reliable test is to watch boxes pass the reader in the configuration they will actually be in. See the discussion of placement in where to put an RFID tag, which shares most of the same constraints.
A short checklist
Is the symbol inside the reader’s field for every box size? Are the bars perpendicular to the scan? Does the box arrive in more than one orientation? Is anything covering the label in the real configuration? Is the symbol large enough for the line speed? Five questions, and answering them honestly usually accounts for the entire gap between the achieved read rate and the expected one.