Stray RFID reads: keeping neighbouring shelves out of your count

RFID has an unusual way of going wrong: it can report more than is there. A reader pointed down aisle 4 may also pick up tags from aisle 5, from a pallet waiting by the door and from a carton on a passing cart. The rack holds 400 items, the screen shows 480, and because nobody can see radio waves, the extra 80 look like a perfectly good result. Controlling stray reads comes down to three adjustments and one test.

Why a reader reads too much

• Radio does not stop at an aisle boundary. It passes through cartons and plastic and bounces off metal racks, so a tag several metres away may still answer.
• Higher power means a longer range. On some devices the default is higher than one aisle needs.
• Some tags do not belong to the count: stock in the next zone, goods in transit, tags left on empty cartons or scrap, and returned items.
• If the system counts reads instead of unique tags, one tag that is read several times looks like several items.

Three adjustments: power, aim and position

1. Power: start low, raise it in steps until the tags in the target zone read reliably, and stop there.
2. Aim: point the antenna at the shelf being counted, not down the aisle. Where possible, stand in the aisle facing the rack and read from close range.
3. Position and zone: fix a standing point for each rack section, work section by section and mark the zone boundaries on the floor. Move tagged cartons of the neighbouring bay out of range, and pause reading when a cart goes by.

Height and distance matter too. If the height and the distance from the rack face change from one pass to the next, the result moves as much as it does with the power setting.

A metal sheet set up between aisles can help in one layout and make things worse in another by reflecting the signal, so try it rather than assume.

Filtering and removing duplicates

• Count each tag once, however many times it is read, using its unique number.
• Filter by your own tag range. If your tag numbers share a recognisable beginning, ignore everything outside that range. This removes other companies' goods and items outside the scope.
• Keep unknown tags on a list and check them one by one when the zone is finished. Each is either a stray from outside or a real item missing from the records.
• Read a location label first at the start of each section, so that the tags that follow are linked to a place.
• Record the filter and de-duplication settings used, so that the result can be reproduced later.

The known-sample test

Before the real run, test as follows.

1. Put 50 tagged items with a verified quantity on the racks of the target aisle.
2. Put another 50 tagged items in the next aisle, and 20 more on a cart at the end of the aisle.
3. Read the target aisle with the planned settings. The expected result is 50.
4. Record the number read inside the aisle (it should be 50), the number from outside (it should be 0) and the number missed.
5. Adjust power and aim and repeat three times. Three runs of 50 with no strays make a pass.

For illustration, a first read at high power returns 50 plus 31 from outside, 81 in all. At medium power the result is 50 plus 6. With lower power and the antenna turned to the rack, it is 50 plus 0, three times in a row. Those settings go straight into the work instruction. Lowering power can make the reader miss real tags, so the test must look at strays and misses together.

Summary

An RFID figure should be a number of confirmed items, not a number of reads. Adjusting power, aim and position, counting unique tags only and testing with a sample keeps the figure from inflating. Smart Inventory uses UHF readers where they suit the goods, tests tag positions and reader settings before the full count, and switches to barcode scanning in areas where stray reads cannot be brought down.