Traceable Quality
Talk to us
Menu

Insight

A 30-day RFID pilot, week by week

One read point, real stock, a hand count beside the reader, and pass marks written down before the first read. What each week is for, and what to change when a mark is missed.

The last part of the RFID series ended on one line. Run the pilot on real stock, through real doors, for a month. Count the misses.

That’s easy to say. Here’s the month.

One read point. Tag real stock and count by hand. Run beside what you have and log every miss. Then the sums, against pass marks written down before any of it.

One read point

Not the whole plant. One dispatch door, or one station on the line, where you already know what’s supposed to go past.

Pick the one you’ll learn most from, which usually isn’t the easiest. A conveyor at the packing station with one carton on it at a time will read beautifully and teach you nothing about a dispatch door with fifty mixed cases on a pallet and a forklift doing 6 km an hour.

Week 1: tag real stock, count by hand

Buy the tag you’d buy in volume, from the supplier you’d buy it from. A paper UHF label is ₹5 to ₹8. An on-metal tag is about ₹60. The vendor’s samples are usually a better tag than the one in your quote, and they always arrive in a clean box.

Stick it where it’ll sit in production, on the face and at the height your packers will actually use, and have your packers do it. Time them. If tagging adds twenty seconds a carton, that’s a finding, and it’s cheaper to find it now.

Then count by hand what passes the read point, every shift, against what the ERP says. That’s the baseline. Without it you’re comparing the reader to a number nobody checked.

Weeks 2 and 3: run it beside what you have

Switch nothing off. The hand count carries on, and so does the ERP.

Every time the reader and the count disagree, write it down. Which case. Which load. Which shift. Where the tag sat and what was next to it. A miss log that only says “missed” is useless, because the fix depends on where the misses cluster, and they always cluster.

Same forklifts, same speed, same drivers. If the reads only work when the forklift slows down, you’ve learned that the reads don’t work.

And change nothing mid-run. If something has to change, note the time and start the count again.

Week 4: the sums

Read rate per tag first. Tags read over tags that passed.

Then whole pallets, because that’s what you’re actually buying. A pallet only counts as read if every case on it reads. At 99% a tag, a fifty-case pallet reads complete 60% of the time. For 95% of pallets you need 99.9% a tag. That’s part 14’s sum, and it’s the one a quote leaves out.

Then check you’ve counted enough to say it.

To claim 99.9% you need about three thousand passes with no misses.

Fewer than that, and a clean run could be luck. A tag that really reads 99.8% gets through a thousand passes without a miss about one time in seven. So if your read point sees 100 items a day, two weeks isn’t a pilot. Run it longer or pick a busier door.

Last, stray reads: tags this door heard that left through another. Part 11 covered why it happens. A reader at full legal power in India, 2 W ERP on four channels between 865.7 and 867.5 MHz, hears a long way past its own door frame. The pilot question isn’t whether there are stray reads. There will be. It’s whether any of them got past the rules and became an event your ERP believed.

Write the pass marks down first

Before week 1. On paper, signed by whoever owns the decision.

A pilot fails quietly like this. The numbers come in a bit short, and the bar moves to meet them. 99.9% becomes “99.7, which is close”. On a fifty-case pallet, 99.7% a tag is 86% of pallets complete instead of 95%. Nine more pallets in every hundred leave with a case nobody accounted for.

Pass, or change one thing

Hit the marks and add the next door. That’s the whole decision.

Miss, and the log tells you what to change. Misses in the middle of the pallet are a tag position problem, or a tag that isn’t sensitive enough. Misses against metal or liquids want an on-metal tag on those items. Misses on one shift are about how that shift builds and drives loads, not about the radio. Stray events want a second antenna for direction and a beam break at the door.

Turning the power up is the change people reach for, and it’s usually the wrong one. A typical dock door is already close to the limit. And a reader that’s rated for 15 m on a spec sheet reads more like 5 m on a real floor, for reasons the range calculator walks through.

Change one thing. Run weeks 2 and 3 again. Change two and you won’t know which one worked.

The plan on one page

The RFID pilot planner puts all of this on one sheet: the read point, the tag, the dates for each week, the checklist, and pass marks worked out from your own traffic. It tells you how many misses you can afford, and when your pilot is too small to prove anything. Print it, sign it, and pin it next to the door you’re testing.

Next step

Tell us what you are trying to trace.

Describe the flow: what arrives, what happens to it and what leaves. We will tell you what it takes to capture it.