How RFID stock counting works (and how accurate it really is)

An RFID stock count is a person walking a zone with a UHF handheld that reads every tag in range — hundreds per second, without line of sight — and an app that compares what was read with what the system expects. It is far faster than scanning barcodes, and in well-run stores it is far more accurate. But “accurate” hides two different numbers, and most disappointing projects confuse them.

1. 1
   Load the expected stock

   The counting app pulls the expected items for the zone — by EPC, or by SKU and quantity — from the store system or WMS.

2. 2
   Walk the zone

   The counter sweeps every fixture high and low at a steady pace. The reader runs continuous inventory rounds; each tag in the field replies with its EPC.

3. 3
   Singulate and de-duplicate

   The EPC Gen2 protocol (ISO/IEC 18000-63) lets the reader separate hundreds of tags. Using session S2 or S3 with a single target, each tag answers once and then stays quiet, so the reader can reach the weak tags.

4. 4
   Compare and reconcile

   The app marks each item found, missing or unexpected (read here, booked elsewhere). Missing items in a small zone trigger a targeted recount before the result is posted.

5. 5
   Post the result

   Counts go back to the system of record; adjustments are approved there, not in the handheld.

Vendors quote read rates; retailers care about inventory accuracy. They move together but are not the same:

A 99 % read rate can still leave inventory accuracy at 90 % if 10 % of items were never tagged, were booked to the wrong location or left without being sold. Conversely, a slightly lower read rate with frequent counts usually beats a perfect count done once a year.

Published retail studies — the most cited come from Auburn University’s RFID Lab — typically report item-level inventory accuracy somewhere around 60–70 % before RFID and 95 % or better after it, with counts going from a few times a year to weekly or more. Treat those as directional: they come from apparel-heavy pilots with good tagging discipline.

For a single count, disciplined UHF handheld counts commonly read the high 90s per cent of tagged soft goods. Expect lower figures, until the tags and the walk are adjusted, for:

• Metal — cans, tools, electronics in metal housings, products on metal shelving. Standard inlays detune; use on-metal tags.
• Liquids — bottles and pouches absorb UHF energy; place the tag on the cap, neck or an air gap.
• Dense stacks — folded denim in the middle of a pile, cartons deep in a pallet.
• Small tags — tiny inlays for jewellery or cosmetics have short range by design.

Tag choice for each of these is covered in RFID tags for metal, liquids and apparel.

Read-rate problems have their own guide: RFID read rate: why tags get missed.

• Count zones in a fixed order, and one zone at a time, so a missed shelf shows up as a gap in that zone.
• Keep the antenna within arm’s reach of the fixture and sweep high, middle and low; tags face every way.
• Walk at a steady, slow pace; rushing is the most common cause of missed tags.
• Use the reader’s power setting: full power for big open floors, reduced power in tight stockrooms to avoid reading next door.
• Recount small zones with missing items before closing the count — it is minutes, not hours.

1. 1
   Seed a test zone

   Put a known set of tagged items, including hard ones (metal, liquids, dense piles), in one fixture.

2. 2
   Count it ten times

   Different people, different times of day. Read rate = average found ÷ known.

3. 3
   Audit a SKU sample

   Hand-count 50–100 SKUs across the store and compare with the system after an RFID count.

4. 4
   Track it by zone

   Accuracy by department or rack row tells you where to fix tags or process.

A count report says “97.2 %”. A map says where the other 2.8 % are: one aisle nobody walked, one metal fixture, one stockroom booked as a separate location. Spatial Map colours each zone, rack or department by found, missing and unexpected stock from your counting app’s results — see RFID stock counting and the wider RFID inventory management system.

Rack locations coloured by stock status

Choosing handhelds or fixed readers? See RFID handheld vs fixed reader vs gate and the hardware we work with — e.g. readers from Zebra, Chainway, SATO or Sanray (trademarks of their owners). Need a counting app or middleware built for your readers? That is custom development.