Buyer's Guide 📅 September 6, 2026 ⏱ 12 min read

UHF RFID Tender Specification: A Clause-by-Clause Skeleton for Buyers

A UHF RFID tender specification document open beside a fixed reader, antennas and a printed acceptance test log sheet

Whoever writes the specification chooses the shortlist. If the document says “supply and install an RFID system for 12 dock doors,” every bidder will answer “compliant” and the prices will spread widely — because each one is free to assume its own antenna count, its own tag grade, its own cable and its own meaning for the word “working.” Naming those things up front moves the whole conversation to the drafting table, where changing your mind is free.

Below is a 22-clause skeleton you can paste into a tender document today. Fields in square brackets are yours to fill. Clauses that do not apply get deleted, but the numbering stays, so every bid can be read clause against clause. The acceptance section at the end is arithmetic rather than a percentage, because a test is a tag count, a pass count and a pass mark — state all three and the clause can be run.

What a tender specification adds to a shopping list

A shopping list is a bill of materials with a delivery date. A specification is a set of obligations that survive into commissioning. The difference shows up in three places — get these three right and the rest of the document largely writes itself, because they are what the rollouts we have been called in to rescue turned on.

Scope named in full. Coaxial cable, N-type to RP-TNC adapters, antenna brackets, surge protection at the mast, cable trays, the region licence paperwork, and the two days of on-site tuning are all real costs. Name them in the document and the cheapest bid is the one that is genuinely cheapest, because every bidder is pricing the same scope.
Acceptance you can measure. “99.5% read rate” is the industry’s favourite number — the RFID News UK guidance calls it a reasonable supply-chain benchmark, and it is. Pair it with four more facts: the population it is measured over, the number of passes, the tag orientation, and who calls the test complete. Name those four numbers and the clause becomes testable, which is what makes it enforceable.
A clean exit at year six. Fixed readers are long-lived infrastructure, so write the contract against the horizon you actually plan to run them for. Firmware support windows, end-of-life notice periods and spares horizons are what keep year six a maintenance job rather than a re-tender.

How to use this skeleton. Fill every [bracketed field]. Delete clauses that do not apply rather than marking them “N/A” — deletion keeps every surviving clause a question with a single answer. Keep the numbering, and require bidders to respond clause by clause in the same order with one of three words: Comply, Comply with deviation (deviation described), or Alternative offered (alternative described and separately priced). That single instruction does more for comparability than any other sentence in the document.

Clauses 1–3: scope, destination markets and band configuration

1. Sites, read points and phasing. “The Supplier shall furnish, install, configure and commission UHF RFID read points at the locations listed below. Quantities are stated per site and per phase. Prices shall be given per read point, not as a site lump sum.” Then a table: site name, read point type (dock door portal, conveyor tunnel, handheld zone, desktop station), quantity, phase, and target go-live month. A count per site is what lets you cut phase 2 while phase 1 stands as priced.

2. Destination market and radio approval regime. “Equipment shall be supplied against the radio approval regime of the destination country stated per line item. The Supplier shall state, per model number offered, the approval held and the certificate reference.” This is the clause that identifies a manufacturer, and it belongs early. For the United States that means an FCC grant; for the EU and UK, the EU or UKCA declaration of conformity against the harmonised standard; for other destinations, the national radio type-approval reference plus any national product registration reference the market requires. Ask for the certificate, not the assurance — clause 17 makes that formal.

3. Band and channel-plan configuration. Name the destination country and let the supplier name the band, power and channel plan it will supply against. Each plan is country-specific, and a correctly specified band is what lets the shipment be switched on the day it lands.

The RAIN Alliance System Design Guidelines give each national plan a short region code, which is the cleanest vocabulary to use in a tender:

Destination	RAIN code	Band	Max reader radiated power	Channel width / spacing	Channel access
India	IN8A	865–868 MHz	2 W e.r.p.	200 kHz / 600 kHz	AFA
EU / UK, lower band	EU8A	865–868 MHz	2 W e.r.p.	200 kHz / 600 kHz	AFA
EU, upper band	EU9B	915–921 MHz	4 W e.r.p.	400 kHz / 1200 kHz	AFA
United States	US9A	902–928 MHz	4 W EIRP	250 kHz / 500 kHz	FHSS

India’s allocation is 865–868 MHz under the Use of Low Power Equipment in the Frequency Band 865–868 MHz for Short Range Devices (Exemption from Licence) Rules, 2021, notified as G.S.R. 853(E) of 10 December 2021, expressly in supersession of the 2005 rules. Interrogators run at 2 W e.r.p. on channels of at most 200 kHz, permitted only on the four channels centred at 865.7, 866.3, 866.9 and 867.5 MHz, with continuous transmission limited to 4 seconds and at least 100 ms between transmissions on a channel; tags reply at −20 dBm e.r.p. In practical terms India now runs the same plan as the European lower band, which is why the RAIN codes IN8A and EU8A carry identical parameters. Equipment type-approved under the 2005 rules remains valid for its life.

The dwell rules differ enough to matter for throughput. RAIN records the US limit as 400 ms per channel with all 50 channels used evenly inside 20 seconds, against 4000 ms for the European plan — so a conveyor timing tuned in Chicago earns its own tuning pass in Rotterdam. Add a sub-clause for multi-country rollouts: “The Supplier shall state whether region profiles are selectable by firmware or fixed at manufacture, and shall state the process and lead time for reconfiguring a unit to a different region profile after delivery.” If you may ever move stock between sites, a field-selectable profile is worth paying for, and it is the single clause that most often decides whether an export order lands ready to switch on.

Clauses 4–7: reader hardware

4. Antenna ports and multiplexer depth. “Each fixed reader shall provide not fewer than [4] monostatic antenna ports. Where the Supplier proposes an external multiplexer to reach the required port count, the multiplexer insertion loss shall be declared in dB and included in the link budget under clause 10. Multiplexer depth shall not exceed [1] stage.” Multiplexers are legitimate engineering, and each stage spends insertion loss and cycle time, so you want it declared in the bid rather than discovered at commissioning. A four-port reader per dock door with two antennas per side is a common starting point in our own portal designs — specify the geometry you need and let the bidder tell you the port count that delivers it.

5. PoE class and negotiation behaviour. This is the clause that keeps a reader performing on the wall exactly as it did on the bench. A reader transmitting at full rated power draws more than a switch port hands out by default, so write the requirement against the class you need, and against the behaviour when a lower class is granted.

“The reader shall operate at its full rated conducted transmit power on all antenna ports simultaneously when powered from an IEEE 802.3[at/bt] Class [4/6] source. The Supplier shall state the reader behaviour when the source grants a lower class, including whether transmit power is reduced, ports are disabled, or the unit holds at boot, and shall state whether that behaviour is reported over the host interface.”

Type	Standard	Class	PSE output	Power at the device	Pairs
Type 1	802.3af	3	15.4 W	13 W	2-pair
Type 2	802.3at	4	30 W	25.5 W	2-pair
Type 3	802.3bt	5	45 W	40 W	4-pair
Type 3	802.3bt	6	60 W	51 W	4-pair
Type 4	802.3bt	8	90 W	71.3 W	4-pair

Figures are from the Ethernet Alliance overview of 802.3bt. Two details deserve their own sub-clause. Class 5 and above are delivered over all four pairs, so the switch, the patch panel and the cable all have to cooperate — on a 2-pair run, a 4-pair-capable reader operates as a Class 4 reader, which is worth designing for deliberately rather than discovering. And 802.3bt defines power demotion, where a source with less available power than the device requested grants a lower class and the device runs in a reduced mode. Require that demotion be logged and alarmed, so a portal announces the day it drops below its commissioned setting rather than at the next audit.

6. Ingress protection and temperature, per location. State these per mounting location, not once for the project, and state jet and immersion exposure separately — each rating is earned by its own test, so ask for both by name. “Readers and antennas at locations [list] shall be rated not less than IP[65] and shall operate over [−10] to [+55] °C ambient. Where wash-down at [location] involves directed jets, the Supplier shall state the rating against jet ingress. Where equipment may be temporarily submerged, the Supplier shall state the rating against immersion and the depth and duration tested.” Add the humidity and condensation case if you have cold rooms, so the rating covers the door-opening cycle as well as the steady-state ambient.

7. GPIO, mounting and cable entry. “Each reader shall provide not fewer than [2] opto-isolated general purpose inputs and [2] outputs, of which at least [1] shall be a dry-contact relay rated [30 V DC, 1 A]. The Supplier shall state the connector type, supply the mating connectors, and state whether GPIO state changes are reported as asynchronous events over the host interface.” Inputs are how a photo-eye tells the reader a pallet has arrived; outputs are how the reader drives a stack light or a barrier. The asynchronous reporting requirement is what buys you event-driven integration rather than a polling loop. Finish with cable entry direction, gland sizes, and whether the mounting bracket is in scope — brackets belong on the scope list from the first draft.

Clauses 8–10: antennas, cabling and the dB budget

8. Antenna gain, polarisation and beamwidth, per read point. Specify the read zone you want and the beam that produces it, and let the part number follow. “For read point type [dock door portal], the Supplier shall offer antennas of stated gain, polarisation and 3 dB beamwidth in both planes, and shall submit a plan and elevation drawing showing the resulting read zone footprint at the tag height range [0.2 m to 2.4 m] and the aiming angle of each antenna.” Circular polarisation forgives tag orientation, at an approximate theoretical polarisation-mismatch cost of about 3 dB against a matched linear pair; linear buys range where you control orientation. The beamwidth number is the one that predicts stray reads, so it belongs in the bid, alongside the drawing rather than inside the commissioning report.

9. Cable grade, run length and adapters. “The Supplier shall state the cable type, the run length per antenna, the attenuation in dB/m at [865/915] MHz from the manufacturer’s datasheet, and the count and type of every connector and adapter in each run. Total loss per run shall not exceed [x] dB. Adapters between connector families shall be declared and counted.” Cable is where quiet money is saved, and every dB of it comes straight out of the link budget: an extra 3 dB of cable loss halves the power reaching the antenna, which is why declaring the run makes the price comparable.

10. The calculated radiated power at each read point, submitted with the bid. This is the clause that makes clause 3 enforceable. “For each read point, the Supplier shall submit the calculation: conducted power at the reader port, minus total cable and connector loss, plus antenna gain, giving EIRP in dBm; and the corresponding e.r.p. where the destination market is regulated in e.r.p. The result shall be shown against the applicable regulatory ceiling with the margin stated in dB.”

The conversion worth getting right is that e.r.p. and EIRP are two scales for the same radiated power, with EIRP running 2.15 dB above e.r.p. RAIN states the ceilings explicitly: 2 W e.r.p. is 33 dBm e.r.p., which is 35.15 dBm EIRP; the FCC 4 W EIRP ceiling is 36 dBm EIRP; the European upper band’s 4 W e.r.p. is 36 dBm e.r.p., or 38.15 dBm EIRP.

Worked, for an Indian read point with a 2 W e.r.p. ceiling. A reader set to 30 dBm conducted, 1.5 dB of cable and connector loss, a 6 dBi antenna: 30 − 1.5 + 6 = 34.5 dBm EIRP, which is 32.35 dBm e.r.p., or 1.72 W — inside the ceiling with 0.65 dB of headroom. Swap in a 9 dBi antenna and the same reader gives 37.5 dBm EIRP, or 35.35 dBm e.r.p., which is 3.4 W and 2.35 dB above the limit; set the reader to 27.65 dBm conducted and it sits back inside. That is the useful lesson buried in the arithmetic: in a power-capped market a higher-gain antenna buys a narrower beam at the same radiated power — which is also what suppresses cross-reads at the next dock door. Sometimes a narrower beam is exactly what you want. Requiring the calculation in the bid puts that trade in the open, and it lets you compare a 6 dBi and a 9 dBi proposal on the same terms.

Clauses 11–13: tags

11. Tag read sensitivity, derived rather than asserted. Start from the read distance you need and work backwards. At 915 MHz the free-space path loss over 6 m is about 47 dB; with 35.15 dBm EIRP arriving at a tag with an assumed 1.5 dBi of inlay gain, the power at the tag chip is about −10.6 dBm. Treat that as a planning number: the margin it appears to leave is the budget for everything the item itself adds — detuning from moisture, absorption by the contents, orientation, and multipath in a metal-framed doorway.

So write the clause in two parts: “The Supplier shall state the inlay read sensitivity in dBm and the test method used, and shall separately state the measured read sensitivity of the tag applied to the Purchaser’s sample item supplied under clause [x].” The second figure is the one that predicts your read rate. Requiring the test method to be named alongside it gives every bidder’s two numbers a common measurement basis, so a reader-sensitivity figure and a tag-sensitivity figure can be read against each other across the whole bid set.

12. Memory. “Tags shall provide an EPC bank of not less than [96/128/196] bits. A unique, factory-programmed and non-modifiable serial number in TID memory [is / is not] required. User memory of not less than [x] bits [is / is not] required, with the intended payload described below.” Serialised TID is what lets you detect a cloned or re-encoded tag; because it is factory-programmed, it is cheap to require at purchase and belongs in the first order. User memory costs money and read time — ask for it where you have named the payload it carries.

13. The grading vocabulary the bid must answer in. Require tag performance in one of two published grading systems, so every response arrives in the same units.

The GS1 Tagged-Item Performance Protocol (Release 1.2, May 2024) names a grade with four designators: a letter for test configuration, a number for performance level, a letter for family, and an optional frequency designator. S means the item is tested singly, M means tested as a stack of 2 and a stack of 11, J means a dense stack (7 items by default). The number rises with performance in steps of five, and within one family a tagged item that passes a higher number also passes the lower ones — M15B implies M10B. The fourth designator is the one that matters for procurement: an absent designator means the grade was tested across both 865–868 MHz and 902–928 MHz; -ETSI means it was tested over 865–868 MHz, and -FCC over 902–928 MHz. So for an Indian or European deployment, ask for a bare S15B or an S15B-ETSI grade, since the -FCC designator means the grade was tested only over 902–928 MHz; a bare S15B carries the widest evidence of the three. Rank a grade only against others carrying the same frequency designator.

The alternative vocabulary is an Auburn RFID Lab ARC use-case spec, which large retailers name directly in supplier playbooks. ARC tests are worth understanding before you cite one: per the ARC Benchmarking Methodology, inlays are submitted in rolls of 4,000, sampled at random, and measured on a platform rotated from 0 to 270 degrees in 30-degree steps against four antennas at 0, 30, 60 and 90 degrees, sweeping 800–1000 MHz in 1 MHz steps at powers giving −25 dBm to +10 dBm at the inlay. That is a far stronger statement than a datasheet number.

Close the tag clauses with a substrate and environment table: material the tag sits on (metal, liquid-filled, textile, carton, glass), attachment method, wash or autoclave cycles to survive, temperature range, and required service life in years or cycles. One row per item type. This table drives more of your asset tagging cost than the reader specification does.

Clauses 14–16: software, protocols and integration

14. Host interface. Name the interface, because a specification is the interface named. “The reader shall expose [one or more of] LLRP over TCP, a documented HTTP/REST interface, an MQTT client with configurable broker, topic and QoS, or a documented Modbus TCP register map. Interface documentation shall be supplied with the bid, not on award.”

LLRP version 1.1 was ratified by the GS1 EPCglobal board on 13 October 2010 and remains the established vendor-neutral low-level reader interface. Its default TCP port is 5084, with 5085 for TLS, both registered with IANA. It is worth naming explicitly because it standardises the things you will later want to change in-house: the ROSpec that defines when and on which antennas the reader inventories, the AccessSpec that defines what it reads or writes on a matched tag, the transmit power level table, and GPI events. Where a PLC is the consumer rather than a server, ask for the register map instead — a documented map is reusable at the next site, which is what makes it worth specifying.

15. SDK platforms and licence terms. “The Supplier shall provide SDKs for [Android, Windows, Linux, Flutter, .NET, Java], each with sample applications, and shall state the licence: whether it is perpetual, whether source is included, whether redistribution within the Purchaser’s applications is permitted, and whether any per-seat, per-device or annual fee applies.” The licence sentence is the one that saves money. An SDK with a per-device runtime fee changes the ten-year cost of a 400-handheld deployment materially, so make it a stated line in the bid.

16. Where the logic lives, and what happens when the link drops. “The Supplier shall state which of the following execute on the reader and which require a server: duplicate suppression and its time window, RSSI thresholding, EPC prefix or mask filtering, tag-population smoothing, and business-rule evaluation. The Supplier shall state the on-reader buffer depth in tag reads and the behaviour when the network link to the host is lost and subsequently restored, including whether buffered reads are forwarded with their original timestamps.”

That last requirement is short and decisive. Store-and-forward with original timestamps is the behaviour to require, because it is what keeps a shift of data intact through a switch reboot — the kind of event a site meets long after a factory acceptance test on a clean network has been signed off. It is also why edge management on the reader earns a clause of its own on any site with more than a handful of read points. Add the fleet questions here too: how firmware is pushed to 40 readers, how configuration is backed up and restored to a replacement unit, and how a reader reports its own health.

Clauses 17–19: documentation, evidence and support window

17. Certification and test evidence, listed as documents. The phrasing matters: ask for documents and you get documents. “The Supplier shall attach to the bid, per model number offered and per destination market, the following documents, each showing the offered model number on its face:” then a list — the radio type-approval certificate or grant, the safety and EMC test reports, the declaration of conformity naming the standards applied, the RoHS declaration, the IP-rating test report, the country-of-origin declaration, and the datasheet revision the bid is based on. Add: “Where a document names an entity other than the Supplier, the Supplier shall state its relationship to that entity.” That sentence, on its own, tells you who makes the hardware.

18. Firmware support window and end-of-life notice. “The Supplier shall commit to a firmware support window of not less than [60] months from the date of final acceptance, during which security patches shall be issued for vulnerabilities affecting the reader’s network stack or management interfaces. The Supplier shall give not less than [12] months’ written notice before a model is declared end-of-life, and shall state the last-time-buy window in months.” Sixty months of firmware support and twelve months of notice is the level we suggest writing in for fixed infrastructure; set the number against the horizon you plan to run the readers for. Ask also for the release cadence and the changelog format, because shipped releases are what turn a support commitment into a support record.

19. Spares, repair turnaround and second source. “The Supplier shall state: the spares holding to be maintained for this contract and where; the repair turnaround in working days for an in-warranty and an out-of-warranty unit; whether repair is board-level or whole-unit replacement; and the advance-replacement terms. Where any item is single-sourced, the Supplier shall identify it and state the mitigation.” Board-level repair and whole-unit swap are both good answers — they are simply very different cost structures over ten years, and the bid is where you choose which one you are buying.

Clauses 20–22: acceptance testing written as arithmetic

Here is where a number earns the test that stands behind it. State the population, the passes, the pass mark as a count, and the confidence.

The arithmetic. A read test produces read opportunities: tags presented multiplied by passes. If a test yields n opportunities with zero misses, the upper 95% confidence bound on the miss rate is approximately 3/n — the rule of three, set out by Hanley and Lippman-Hand in JAMA in April 1983, which states that the upper 95% confidence limit of a 0/n rate is approximately 3/n and is accurate to the nearest percentage point for n above 30. Invert it and you get the test size you actually need:

To demonstrate 99% at 95% confidence with zero misses: about 300 opportunities. 100 tags × 3 passes.
To demonstrate 99.5%: about 600 opportunities. 200 tags × 3 passes.
To demonstrate 99.9%: about 3,000 opportunities. 200 tags × 15 passes.

That third line is why a 99.9% clause is worth writing deliberately rather than copying. It is a two-hour test with a large tag population, so put a 99.9% clause in when you are prepared to run the test that backs it, and choose 99% or 99.5% where the shorter test is the right one for the site.

20. Read-rate acceptance. “A pass consists of moving the loaded test unit through the read point at [1.5] m/s in the normal direction of travel. The test population shall be [200] tags applied to the Purchaser’s own items per clause [13], loaded in a mix of orientations representative of normal operation, and shall be re-randomised between test sets. [3] consecutive passes shall be run. Acceptance requires zero missed tag reads across all [600] read opportunities. Additionally, no single tag may be missed on more than one pass in any test set.”

The second sentence of the pass mark is the one that finds dead spots. It catches the case where the same six tags at the bottom of the pallet are missed on every pass — a read-zone geometry matter, best resolved while the installer is still on site, and one a per-tag pass mark surfaces on the day.

21. Stray-read acceptance. Cross-reads deserve their own acceptance test, because a stray read arrives as data rather than as a visible gap. “With the read point operating in its accepted configuration, a static population of [100] tags shall be placed in the adjacent [lane / bay / staging area] at a distance of [3] m from the nearest antenna, at heights from [0.2] m to [2.0] m. [10] cycles of the read point shall be run with no tag present in the read zone. Acceptance requires zero reads of any tag from the static population. During the clause 20 passes, no tag from the static population may appear in any transaction.” Add the neighbouring-reader case where two portals sit within roughly 10 m: run both simultaneously for the clause 20 test, so dense-reader conditions are inside the test rather than outside it.

22. Test conditions and what may change between attempts. “Tests shall be run at the installed site, in the accepted final configuration, under normal site lighting, material handling and personnel activity, with all adjacent readers operating. Before the first attempt the Supplier may adjust reader transmit power, antenna aim, session and target settings, and tag placement. After the first attempt, tag placement and tag model are frozen; the Supplier may adjust reader settings only, and every changed setting shall be recorded on the log sheet before the next attempt. The Supplier is permitted [3] attempts, after which [remedy].” Freezing tag placement after attempt one keeps the accepted configuration the same one operators will reproduce every day.

Attach this log sheet as an annex and require it signed per pass:

Pass	Tags presented	Settings changed since last pass
1	200	—
2	200
3	200
Set total: ____ of 600 opportunities read. Acceptance requires 600 of 600 and no EPC missed more than once.

The same structure scales to a warehouse rollout unchanged — you run one set per read point and the annex becomes the commissioning record.

Scoring the responses

Publish the weighting in the tender. Bidders write better responses when they know what is being measured, and a published matrix supports you when you award to the second-cheapest bid.

Criterion	Weight	What earns the marks
Mandatory compliance	Gate	Clauses 2, 3 and 17 answered with documents. A gate rather than a score: a bid clears it on documents and proceeds to scoring.
Price	30	Normalised total per the scoring rule below.
Evidence quality	20	Certificates and calculations attached. Clause 10 link budget submitted per read point.
Integration effort	20	Interfaces present out of the box, SDK licence terms stated, buffering behaviour described.
Continuity	20	Firmware window, EOL notice, spares holding, repair turnaround.
Acceptance plan	10	Bidder’s own proposed test method and its willingness to be measured by clauses 20–22.

Normalising a bid that has moved a requirement into an option. Comparability comes from one rule: before comparing, add back into every bid the bidder’s own quoted price for every item another bidder included as standard. Where an optioned item is unpriced, price it at the highest figure quoted by any bidder for that item and add a note. Publish this rule in the tender, and full scope stays in the base price from the first draft rather than arriving in the last negotiation.

The two questions that identify a manufacturer. Put both in the response form, each with a one-line answer box.

“On which legal entity’s name is the radio type-approval certificate for the exact model number quoted issued, and what is that entity’s relationship to the Supplier?” A manufacturer names itself. Anyone else names a third party, which tells you who you will actually be waiting on when a variant is needed.
“Name the party that can build and sign a firmware image for this model, and state the lead time to deliver a build configured for a region profile not currently shipped.” Whoever can build and sign the firmware is the party who can fix your problem in the field. For multi-market rollouts this question predicts your year-two experience better than any other line in the document.

Both answers are easy to give and hard to fake, which is what you want from a scoring question. If you are drafting a tender now and want a supplier’s-eye view of which clauses are routine and which will move your price, that is a conversation worth having before the document goes out rather than after.

Frequently asked questions

What should an RFID tender specification include?

At minimum: scope and read-point counts per site; the destination market for each unit and the radio approval regime it is supplied against; the band and channel plan; reader antenna ports, PoE class and IP rating per location; antenna gain, polarisation and beamwidth per read point; cable grade and loss budget in dB; the calculated radiated power at each read point against the market ceiling; tag memory and a published performance grade; the host interface named explicitly; certification documents attached to the bid; a firmware support window and end-of-life notice period; and acceptance tests written as a tag count, a pass count and a pass mark. The 22-clause skeleton above covers all of these in an order you can paste and edit.

How do you write measurable RFID read-rate acceptance criteria?

Convert the percentage into read opportunities and state a count as well as a rate. Opportunities equal tags presented multiplied by passes. With zero misses in n opportunities, the upper 95% confidence bound on the miss rate is about 3/n — the rule of three, published by Hanley and Lippman-Hand in JAMA in 1983 — so roughly 300 opportunities demonstrate 99%, 600 demonstrate 99.5% and 3,000 demonstrate 99.9%. A workable clause reads: 200 tags on the buyer’s own items, three consecutive passes at normal travel speed, zero misses across all 600 opportunities, and no single tag missed on more than one pass. That last condition is what catches a read-zone dead spot, and it surfaces one while the installer is still on site.

Should an RFID tender specify a frequency band or a destination country?

State the destination country and require the supplier to name the band, power and channel plan it will supply against. Each plan is country-specific, and a correctly specified band is what lets the equipment be switched on the day it lands. India’s allocation is 865–868 MHz at 2 W e.r.p. under G.S.R. 853(E) of 10 December 2021, on four channels centred at 865.7, 866.3, 866.9 and 867.5 MHz; the FCC band is 902–928 MHz at 4 W EIRP; the European upper band is 915–921 MHz at 4 W e.r.p. The RAIN Alliance region codes (IN8A, EU8A, EU9B, US9A) give you a compact vocabulary for the clause. For multi-country rollouts, also ask whether region profiles are field-selectable by firmware or fixed at manufacture.

What documentation should I ask an RFID reader manufacturer to provide?

Ask for documents, each showing the offered model number on its face: the radio type-approval certificate or grant for each destination market, safety and EMC test reports, the declaration of conformity naming the standards applied, a RoHS declaration, the IP-rating test report, a country-of-origin declaration, and the datasheet revision the bid is based on. Add one sentence requiring the supplier to state its relationship to any entity named on a document other than itself — that single line tells you who manufactures the hardware and who is reselling it.

How many antennas should I specify per dock door?

Specify the read zone and the tag population, then let bidders propose the geometry and justify it with the clause 10 link budget and a read-zone footprint drawing. Two antennas per side of a standard dock door is a common starting point in our own portal designs, which is why four-port readers are common there, and the right answer for your site follows from pallet height, tag placement, travel speed and how close the neighbouring door is. Requiring a plan and elevation drawing with aiming angles gets you a comparable answer from every bidder while leaving the design to the people quoting it.

What firmware support window is reasonable to require?

Sixty months from final acceptance is the level we suggest writing in for fixed reader infrastructure, paired with a commitment to issue security patches for vulnerabilities in the reader’s network stack and management interfaces. Add an end-of-life notice period of at least twelve months and a stated last-time-buy window, and set both against the horizon you plan to run the readers for. Also ask for the release cadence and the changelog format, because shipped releases are what turn a support commitment into a support record.

Should a tender name a specific tag chip?

Specify the outcome and you keep the bidder pool wide while tightening the performance you get, because the same chip performs very differently on different items. Name the EPC length, whether a serialised TID is required, any user-memory payload, and a published performance grade. Use a GS1 TIPP grade — noting that the optional frequency designator matters, since a grade ending in -FCC was tested over 902–928 MHz, so for an 865–868 MHz deployment ask for a bare grade or an -ETSI grade — or name an Auburn ARC use-case spec. Then require the measured read sensitivity of the tag applied to your own sample item, which is the figure that predicts your read rate.