Case Study: The South African Retailer That Couldn’t Make RFID Stick
Not every RFID deployment succeeds, and the failures are often more instructive than the successes. This case study examines a mid-sized South African fashion retailer - operating around 120 stores across three provinces - that invested significantly in RFID between 2018 and 2020, achieved poor results, and ultimately mothballed the system without completing the rollout. The lessons from that failure shaped a more successful second attempt that began in 2023.
The details have been reconstructed from interviews with former project team members. The company is not named at their request.
The Original Ambition
The retailer’s initial business case was built around three outcomes: improving inventory accuracy from a measured baseline of 64 percent to above 90 percent, reducing stocktake labour costs by automating the annual count process, and enabling store-level replenishment driven by real RFID reads rather than estimated stock levels. The projected payback period was 28 months. On paper, it was a credible proposal for a fashion retailer of that size.
The hardware procurement was handled competently. Fixed reader infrastructure was installed at receiving doors and point-of-sale exits in the pilot stores. Handheld readers were sourced for floor scanning. The tag specification was developed with input from the main inlay supplier and tested on representative garment types in a controlled environment. The middleware platform was a recognised enterprise product with existing integrations to the retailer’s ERP system.
Where the project ran into trouble had nothing to do with any of those things.
The Tagging Problem
The retailer’s garments were sourced from a mix of local manufacturers and offshore suppliers, primarily in China, Bangladesh, and Mauritius. The plan was for tags to be applied at source - either as woven labels with integrated inlays or as hang tags encoded and applied in the factory. Source tagging is the right approach: it distributes the labour cost, ensures tags are present from the beginning of the supply chain, and allows the RFID data to track product from the point of manufacture.
In practice, source tagging compliance was inconsistent from the first shipment. Some suppliers delivered correctly tagged product. Others delivered product with tags applied to the wrong location on the garment - the inside hem rather than the care label, or the outer hang tag rather than a permanent label - resulting in high tag removal rates when customers discarded the hang tag at purchase. A third group delivered product with tags that had been encoded with incorrect data: wrong SKU, wrong size, or EPC values that did not match the retailer’s format specification.
The compliance monitoring programme was inadequate. The project team had issued a tagging specification to all suppliers but had not built a testing and acceptance protocol into the receiving process. When tagged product arrived at the distribution centre, it was not systematically sampled against the specification before being released to stores. Problems that should have been caught at goods-in reached the shop floor, where they appeared as mysteriously low read rates and inexplicable inventory discrepancies.
The Read Rate Reality
In the pilot stores, read rates averaged 71 percent against a target of 98 percent. The shortfall was traced to three overlapping causes: tags applied to the wrong location, tags damaged during transport and handling (the woven label inlays from one supplier had a chip bond failure rate of around 8 percent on arrival), and a subset of garment types - heavily embellished items with metallic thread and sequins - that consistently degraded read performance regardless of tag placement.
A 71 percent read rate does not mean a 71 percent inventory accuracy improvement. It means the system has a structural blind spot covering nearly a third of all items. When the RFID count says there are 47 units of a particular SKU on the floor and the true count is 65, the replenishment system makes decisions based on a fiction. In some cases, replenishment orders were suppressed for lines that were actually running low because the RFID count showed adequate stock. In others, stock was ordered for items that the system said were missing but which were present and simply unread.
The inventory accuracy figure at the end of the pilot period was 73 percent - an improvement of 9 percentage points over the pre-RFID baseline, but far below the 90 percent target that had justified the investment.
The Organisational Response
When the pilot results came in below target, the project team’s response was to adjust the technology rather than address the root causes. Reader placement was changed. Antenna configurations were revised. A different middleware configuration was tested. Each change produced marginal improvements that eroded again within a few weeks as underlying compliance issues continued to feed bad data into the system.
The change management programme had been minimal. Store staff had received a half-day training session on the handheld readers and a briefing on the new stocktake process. What they had not received was any explanation of why the new process mattered, what the system was supposed to do for them, or what to do when the read rates seemed wrong. When the handheld scans produced results that did not match staff’s intuitive sense of what was on the floor, the most common response was to complete the RFID scan and then do a manual check anyway - defeating the purpose of the system and adding to workload rather than reducing it.
Store managers, measured on sales and shrink metrics that had not been updated to reflect the RFID programme, had no formal accountability for data quality. The connection between accurate RFID reads and the business outcomes the system was meant to deliver was never made visible to the people whose cooperation determined whether it would work.
The Decision to Stop
After 18 months and a rollout that had reached 34 of the planned 120 stores, the programme was suspended. The capital already invested was written off. The remaining stores were not equipped. The system continued to operate in the 34 stores where it had been installed, but without active management or ongoing investment in compliance, read rates drifted further downward over the following year.
The formal post-mortem identified supplier compliance and change management as the primary failure causes - the same two factors that appear in the majority of RFID failure analyses. What it also noted, more candidly than most such documents, was that the project governance had created the conditions for failure by approving a rollout timeline that left no room for learning from the pilot before expanding, and by not establishing accountability for data quality at any level of the organisation.
The Second Attempt
When the retailer returned to RFID in 2023 with a new technology director and a rebuilt business case, the approach was structurally different. The supplier compliance programme was built before any store infrastructure was installed: six months of supplier onboarding, factory audits, sample testing against acceptance criteria, and a receiving-dock validation process that flagged non-compliant shipments before they entered the DC. Suppliers that could not meet the specification within a defined remediation period were moved to a different fulfilment process that did not depend on source tagging.
The pilot covered a single store format - compact urban stores where the garment range was narrowest and the tag performance most predictable - rather than attempting to prove the technology across all store types simultaneously. Success criteria were defined in advance: 95 percent read rate in goods-in scanning, 92 percent inventory accuracy after 90 days, stocktake time reduction of at least 60 percent. The pilot ran for four months before any decision on expansion was made.
Read rates in the second pilot averaged 96 percent. Inventory accuracy reached 93 percent by day 60. The stocktake that had previously taken a team of four people a full day was completed in under two hours with two people. The expansion decision was made on the basis of those numbers, and the rollout is ongoing.
What the Failure Taught
The retailer’s first RFID programme failed not because RFID does not work in South African fashion retail. It works, as the second programme is demonstrating. It failed because the organisation treated it as a technology deployment rather than a business change programme, skipped the supplier compliance work that determines data quality, and moved to scale before understanding what was wrong with the pilot results.
The cost of that failure - in capital written off, in organisational credibility lost, and in the time it took to return to a technology that was always capable of delivering the intended outcomes - was entirely avoidable. The second attempt is succeeding because it treated the lessons of the first as design constraints rather than unfortunate history.
