Lite RFID system for industrial SMBs: traceability without migrating to SAP
How to digitize plant-floor traceability without throwing your ERP overboard. UHF RFID + ESP32 + Python + reactive dashboard stack. Real case: ROI under 3 months.
The real problem, not the brochure one
A Galician industrial SMB — 60 employees, technical-parts manufacturing, a corporate ERP in place since 2015 — loses 3-4 hours a week per supervisor hunting for batches on the plant floor. The Excel stock-control sheets don't reconcile at month-end. When a client requests traceability for a specific batch (a frequent ask in regulated sectors), the answer takes two days.
The SAP rep smells an opportunity and pitches a traceability module for €80,000–200,000 with a 6-month rollout. For an SMB on tight margins, that's using a sledgehammer to crack a nut. The right question isn't "which SAP module do we need?", it's "how do we solve this without rewriting the ERP?".
A lite RFID stack that actually works
The answer fits in five pieces, all open-source or low-cost:
Fixed UHF readers at critical zones (plant entrance, exit, machine A, machine B, warehouse). Cost: €200-400 per reader. Honest brands: Impinj Speedway, Zebra FX9600. For tight budgets: Chinese UHF Gen 2 clones at €80-150.
Passive UHF RFID tags stuck to containers and pallets. €0.15-0.40 per tag. Useful life 3-5 years in an industrial environment. We do NOT use active tags (expensive, batteries that die).
ESP32 microcontrollers at each reader, acting as a bridge between the reader (serial or Ethernet) and the backend (HTTP/MQTT over TLS). Cost: €5-12 per unit.
Python backend (FastAPI + PostgreSQL) that centralizes events, deduplicates reads and exposes REST. Hosted on a modest VPS (€5-15/month) or on the client's existing server. One
eventstable and onetagstable cover 80% of the model.Reactive dashboard (React or Svelte) consuming the backend in real time. Views by batch, by machine, by operator, by end client. Configurable alerts.
Typical hardware investment: €2,500-5,000 for a small-to-medium SMB. Software: the cost of my project (4 weeks) or of the local integrator you choose.
The part the SAP rep won't tell you
The lite RFID system does NOT replace the ERP. It lives on top. The trick is in the sync:
[RFID reader] → [ESP32] → [Python backend] → [Dashboard]
↓
[Sync with ERP]
(REST API or nightly batch)
The ERP stays the source of truth for finance, procurement, invoicing. The RFID system is the source of truth for physical events on the plant floor. Every night (or in real time if the ERP allows it) a job syncs:
- Reads of "batch X left machine Y at 14:32" → entry in the ERP's production module.
- Real stock per SKU counted in the warehouse (via RFID pass on shelving) → inventory adjustment in the ERP.
If your ERP exposes a REST API → clean integration. If it only has CSV exports → a nightly batch works perfectly. If it's a legacy system with nothing → we write directly to its DB (yes, with automatic backup beforehand and transactional review).
Zero-trust architecture (when it applies)
In regulated sectors (pharma, tier-1 automotive, defense) or when the client has strict compliance, we add:
- mTLS between each ESP32 and the backend. Every microcontroller has its own client certificate. Lose one, revoke it, done. Without a valid certificate, the backend rejects it.
- Wireguard or Tailscale tunnel between the OT network (plant) and the IT network (backend). Isolates industrial traffic from corporate traffic.
- Audit logs signed with the backend's key. Any change is traced.
Extra cost: 1-2 additional implementation days. If the client needs it, it's not up for debate.
Proven ROI: the real case
60-employee SMB, machining sector, rollout executed in 4 weeks:
</div>
<div class="stat-callout">
<span class="stat-value">-90%</span>
<span class="stat-label">Month-end stock discrepancy</span>
</div>
<div class="stat-callout">
<span class="stat-value">7.9 months</span>
<span class="stat-label">Total project payback</span>
</div>
<div class="stat-callout">
<span class="stat-value">+52%</span>
<span class="stat-label">First-year ROI</span>
</div></div>
| Metric | Before | After | Delta |
|---|---|---|---|
| Hours/wk hunting for batches | 4 h × 5 supervisors = 20 h | 0.5 h | -19.5 h |
| Client claims requesting traceability | 2-3/month, 2-day response | same frequency, 5-min response | satisfaction ↑↑↑ |
| Month-end stock discrepancy | 3-7% | <0.5% | -90% |
| Annual cost of person-hours saved (€18/h) | — | — | +€18,252 |
Total investment: ~€12,000 (hardware + software + 4 weeks integration). Payback in 7.9 months, ignoring the qualitative upside (happy clients, compliance, less internal friction).
When NOT to use this approach
Honesty before the sale:
- No if your production is discrete and low-volume (<50 batches/month). The ROI doesn't add up.
- No if you work with products too small for an RFID tag (precision jewelry, microelectronics). There are alternatives (laser-etched DataMatrix codes).
- No if your ERP is 100% cloud with no API or DB access. With no integration point, the RFID system stays an island — useful but limited.
- No if you expect the system to work without operator training. Two 1-hour sessions are enough, but you have to invest them.
Next step
If your operation has similar symptoms — supervisors losing time on the floor, month-end close with discrepancies, traceability claims without a fast answer — the first step is a 90-120 minute audit of your process. I'll tell you whether a lite RFID fits, the exact stack and a closed budget range.
Book a free coffee chat 30 min · no commitment · closed budget after the callPublished by Alex Fernandes — AI and industrial-operations specialist, Ourense (Galicia, Spain). I connect the industrial plant floor with AI systems and RFID/IoT traceability in production.