RFID for Inventory Management: Step-by-Step Implementation

What RFID changes in inventory

With barcodes, every item has to be found, turned so its label faces the scanner, and scanned one at a time. That is why full stock counts happen once or twice a year and why the system quantity and the shelf quantity drift apart in between. Passive UHF RFID tags can be read without line of sight and many at a time, so a person walking an aisle with a handheld reader can capture hundreds of items per second. Counts that were annual can become weekly or even daily, and fixed readers can report movements as they happen.

The same approach works across very different inventories: apparel and footwear in retail stores, stockrooms and back-of-store areas, spare parts in maintenance stores, IT assets such as laptops and monitors, and medical consumables and equipment in hospitals. The details differ, but the phases below are the same.

Phase 1: Define what is counted

Before talking about tags, write down exactly what you want to know and at what level:

• Item level - every unit has its own tag and unique serial. Best for apparel, footwear, electronics, IT assets, instruments and high-value spares. Gives exact counts by SKU, size and colour.
• Case or carton level - one tag per carton, bin or tote, with the contents known from the packing list. Suitable for low-value, fast-moving consumables.
• Asset level - fixed and movable assets that are counted for audit and location rather than sold.

Also decide where stock lives (sales floor, stockroom, store, ward, warehouse location), how often you need a count, and which decisions the data will drive - replenishment, audit, shrink investigation, expiry control. Projects that skip this step tend to tag everything and use the data for nothing.

Phase 2: Choose tags and decide who tags

Tag choice

Tag selection depends on the material being tagged, the read distance needed, memory requirements and cost. Paper or fabric hang tags and labels suit apparel; on-metal tags suit tools, laptops and metal parts; liquid-tolerant designs suit bottles and medical items; rugged hard tags suit reusable assets. The chapter on which RFID chip to use explains how to match chips and memory to applications. Always test shortlisted tags on real products in the real environment before ordering in volume.

Source tagging vs in-house tagging

Source tagging means the manufacturer or supplier applies and encodes the tag before the goods reach you. It is the cheapest and most consistent option at scale, but it needs supplier agreements, a common encoding standard (usually GS1 SGTIN for retail goods) and incoming checks.

In-house tagging means you print, encode and apply tags yourself - at receiving, in a tagging room, or for existing stock during the initial rollout. You will need RFID printers for labels and encoders for pre-made tags; see RFID printers and RFID encoders explained. Write a simple tagging rule book: which tag on which product, exactly where it goes, how the EPC is built, and how a failed or damaged tag is replaced.

Phase 3: Plan how you will read

Handheld cycle counts

For most sites the first and most valuable step is periodic counting with a handheld reader - a sled attached to a smartphone or an all-in-one device. The operator walks a defined route, sweeping shelves and racks, and the app compares what it found with what the system expects. Technique matters: a steady pace, consistent distance from the shelves, sweeping both high and low, and a fixed route so that areas are not missed or counted twice. The handheld RFID readers guide covers device choice and counting technique.

Fixed and overhead reading

Where you need near real-time visibility - stock moving between stockroom and sales floor, items leaving a store, assets leaving a ward - fixed readers at doorways or ceiling-mounted readers covering an area keep the picture updated without anyone counting. Overhead systems can also estimate location within a zone; see overhead RFID readers. Fixed infrastructure costs more, so most projects add it after handheld counting has proved the data.

Software and ERP sync

Raw reads are not inventory. The software must filter duplicates, map EPCs to SKUs or asset records, compare counted quantities with expected quantities, and send adjustments or reports to the ERP, POS or asset register. Decide early whether RFID counts will automatically update the system quantity or only generate a discrepancy report for approval. Integration patterns are covered in RFID software integration.

Phase 4: Baseline, pilot and accuracy measurement

Take a baseline

Before RFID, measure the current state in the pilot area: how long a full count takes, how many people it needs, and how accurate the system quantity is. Without a baseline you cannot prove improvement.

How to measure accuracy properly

1. Choose a sample of locations or SKUs in the pilot area (for example 30-50 SKUs or a set of shelves).
2. Do a careful physical audit of that sample - counting by hand, slowly, with two people if possible. This is your "truth".
3. Compare the system quantity (before RFID) with the truth: this gives baseline inventory accuracy.
4. Run the RFID count on the same sample immediately afterwards, with no stock movement in between, and compare it with the truth. This gives the RFID read accuracy.
5. Investigate every difference: untagged items, damaged tags, tags hidden behind metal, items counted from a neighbouring area, or genuine audit errors.

Inventory accuracy is often defined as the percentage of SKUs (or locations) whose system quantity exactly matches the physical quantity. State which definition you use and keep it the same before and after.

Run the pilot

Tag the pilot area's existing stock, train two or three counters, and run counts at a fixed frequency (for example weekly) for four to eight weeks. Track count time, read accuracy, inventory accuracy, tagging compliance on new receipts and the number of exceptions.

Phase 5: Go-live and routine counting

After the pilot, roll out area by area. For each area: tag existing stock, switch incoming goods to the tagging rule (source or in-house), train staff, and set a count schedule. A typical pattern is:

• Daily or several times a week - fast-moving, high-value or high-shrink categories.
• Weekly - the main sales floor and stockroom.
• Monthly or quarterly - slow movers and fixed assets.

Roles and responsibilities

Assign clear owners before go-live. A store or stores manager owns the count schedule and signs off adjustments; a small group of trained counters performs counts and searches; the receiving team owns tagging compliance for new goods; and IT or the RFID partner owns readers, software, the EPC-to-SKU mapping and the ERP interface. Without named owners, counts slip, exceptions pile up and confidence in the data falls within a few months.

Write the counting procedure as a one-page standard: route, reader settings, how to handle fitting rooms, carts and returns areas, when to stop stock movement during a count, and how to upload results. New staff should be able to follow it after a short demonstration.

Retire the old full count only when RFID accuracy has been stable for several cycles and auditors or finance agree.

Illustrative timeline

Illustrative only, for one store or stockroom with a single product category; larger estates run several areas in parallel.

Equipment checklist

Exception handling

Every RFID count produces exceptions. Decide in advance who handles each type and how:

• Unknown tags - EPCs that are not in the item master. Usually untagged-at-receiving goods, tags from another store or supplier, or encoding mistakes. Route them to a review list rather than ignoring them.
• Missing items - expected but not read. Search with the handheld's "find" mode before writing off; check damaged tags, metal shelves, items in carts or fitting rooms, and transfers not yet booked.
• Extra items - read but not expected in this location. Often misplaced stock or a booking delay; correct the location or transaction.
• Duplicate EPCs - two items with the same code. Re-tag one and fix the encoding process.
• Untagged items - spot-check new receipts for tagging compliance and feed results back to suppliers or the tagging team.

KPIs and pitfalls

KPIs

• Inventory accuracy % - SKUs or locations where system quantity equals physical quantity.
• Count time - person-hours per count and per area.
• Count frequency - how often each area is actually counted.
• Read accuracy - share of tagged items found by an RFID count versus a controlled audit.
• Shrink - unexplained loss as a share of sales or inventory value, measured the same way before and after.
• Tagging compliance - share of received items correctly tagged and encoded.
• Out-of-stocks and replenishment accuracy where counts drive replenishment.

Pitfalls

• Tagging before deciding what decisions the data will support.
• Choosing tags by price alone and discovering poor reads on metal or liquid items.
• No baseline, so improvement cannot be shown.
• Letting new, untagged stock in after go-live, which slowly destroys accuracy.
• Counting with inconsistent routes and technique.
• Pushing RFID counts straight into the ERP without an approval step during the early weeks.

Indicative budgeting: tag cost per item, the number of handhelds and any fixed infrastructure dominate the budget, and all of them vary with volume and product mix - treat any figure as indicative until you have tested tags and agreed a scope. To learn to run projects like this end to end, see the hands-on RFID courses in India.