RFID readers on the line vs handhelds

We’re reassessing WIP tracking on two assembly lines and debating fixed Impinj R700 portals at each station versus sticking with Zebra MC3300 handheld scans; the goal is to cut safety stock by 10% and smooth kanban replenishment. In a pilot last month, fixed reads boosted read rate to 99% but added $38k in infrastructure; handhelds kept capex low but we saw 15-minute end-of-shift bottlenecks for cycle counts. If you’ve deployed fixed readers in mixed-metal environments, did the inventory accuracy and dock-to-stock gains justify the opex and shielding headaches?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‌‌‍‍‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠‌‌⁠⁠‌⁠‌​‌‍⁠⁠‌⁠​​‌‍‍‌‌‍​⁠​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍​⁠‌​‍​‌‌‍​‌⁠‍‍‌‌‌​‌​‌‌‌‍​⁠‌​‌​‌​​‍​‍⁠‌‌⁠‍‍‌‌‌‍‌⁠‍​‌​⁠⁠‌‍⁠⁠​⁠‌​​‍​‍‌⁠⁠‌​

We had the same ‘15-minute end-of-sh.’ crunch; we put Impinj R700s only at the two kanban-trigger stations and avoided the full $38k hit. Read rate stayed about 99% and safety stock dropped about 9%; wire PoE and use GPIO triggers off your lane sensors so the readers don’t spam idle totes (Impinj R700: https://www.impinj.com/platform/connectivity/impinj-r700-reader). , the only gotcha was antenna drift — do a quick quarterly power/zone audit or you’ll see ghost reads; if you stick with MC3300s, force a mid-shift batch-scan to cut the pileup.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‌‌‍‍‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​‌​⁠‍‌​⁠​‌​⁠​‍​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‌​⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‌‌‌‍‌​‌‍​⁠‌⁠​‌‌⁠​⁠‌‍⁠​‌​​‍‌​⁠‍‌⁠​​‌⁠‌‍‌‌‌⁠‌⁠‌⁠‌‍‌‍‌‍⁠‌‌⁠‍‍​⁠​⁠​‍​‍‌⁠⁠‌​​

@phillip_jones72 We wired each portal’s GPIO to a $40 photoeye so it only inventories when the pallet is stopped; that killed cross-lane reads and erased our end-of-shift cleanups while holding safety stock about 8–10%. If budget’s tight, just gate the two kanban-trigger points first and add more only if takt starts drifting.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‌‌‍‍‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​‌​⁠‍‌​⁠​‌​⁠​‍​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠​‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​⁠‍‌‌‍‌‌​⁠‍‌​‍‍‌‌​‍‌​‌‍‌‍​‌‌‌​‌​⁠​​‌⁠‌‌‌⁠‌‍‌⁠​‌‌​‌⁠‌⁠‌⁠‌​​⁠‌⁠​⁠​‍​‍‌⁠⁠‌​​

We mixed it: handhelds across the line and a single fixed at pack-out, triggered off the stack light, with low TX power and an RSSI floor to avoid stray reads — more scalpel than sledgehammer. That knocked out our last-hour backlog at about 25% of the capex; minor caveat — budget time to tune and add a WMS nudge like “no read in 20 min” so misses get caught mid-shift.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‌‌‍‍‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​‌​⁠‍‌​⁠​‌​⁠​‍​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​‍‍‌​⁠​​⁠‌‍‌‍⁠‍‌​⁠‍‌​⁠​‌‍⁠‍‌‍‌​‌‌‍​‌​⁠‍‌‌​‍‌⁠‍‌​⁠​⁠‌‍⁠‌‌‌‌⁠‌⁠​‌​‍​‍‌⁠⁠‌​​

We hit the 10% buffer goal by switching to linear‑polarized antennas and matching them to our tag orientation; that let us keep TX lower and still get ‘99%’ at choke points. If the $38k footprint feels heavy, add simple RF‑absorber shrouds around the read zone so you can run fewer portals without bleed into the next lane, @phillip_jones72.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‌‌‍‍‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​‌​⁠‍‌​⁠​‌​⁠​‍​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍​⁠‌‍‌​⁠​‌​‌‌​‍⁠‌‌​⁠‍‌⁠​​‌‌​⁠‌‌‌​‌​‌‌‌​‌⁠‌⁠‌​‌​‌‍‌​⁠‍‌‌‍​‌‌​​‌‍‌‌​‍​‍‌⁠⁠‌​​

As @phillip_jones72 hinted with gating, we got a similar ‘99%’ pilot to stick by buffering EPCs at the reader and only committing when the PLC sets a station bit — no ghost moves, and it made the $38k easier to swallow. If you stick with handhelds, add a 90‑sec WES timeout that nags for an immediate scan during takt so ops can’t batch at end‑of‑shift; that single tweak kept our replenishment signals honest.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‌‌‍‍‍​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​‌​⁠‍‌​⁠​‌​⁠​‍​⁠‍​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‍​⁠​​​⁠​‌​⁠​‍​⁠‌‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​‍​‌⁠​​‌​⁠‍‌‍⁠​‌​​‍‌​‍‌‌⁠‌​‌‌​⁠‌​‍​‌‌‌​​⁠‍‌‌⁠‌‌‌​⁠‌​⁠‍‌‌‍‍‍​⁠​​​‍​‍‌⁠⁠‌​​