Catalogues list antenna gain as if it were a quality score. In practice gain is a trade: every decibel of concentration is paid for in coverage angle, and the angle is what determines whether the antenna sees one dock door or half the warehouse.

Choosing an antenna is therefore not a purchasing decision but a layout decision. Get the zone right on paper and the part number follows; get it wrong and no amount of gain will recover the reads.

Gain is concentration, not power

The single most expensive misconception in UHF deployments is that a higher-gain antenna reaches further because it is stronger.

Gain in dBi describes how tightly the antenna focuses the energy it is given; total radiated power does not increase.
Adding three decibels doubles the effective power in the main lobe and multiplies ideal free-space range by about 1.41.
The same three decibels remove roughly half the solid angle the antenna can cover.
Regulatory limits cap effective radiated power, so gain does not let you transmit more; it lets you transmit more narrowly.
A twelve dBi antenna on a two-watt reader is a narrow, long corridor of energy, not a stronger version of a six dBi antenna.

The arithmetic of a read zone

Range is a chain, and the antenna is only one link in it.

Reader output power, minus cable loss, plus antenna gain, must stay under the legal EIRP for the region.
A three-metre cable run at 900 MHz typically costs one to two decibels, which a careless installer gives away by choosing thin cable.
The tag's own sensitivity and the polarisation mismatch matter as much as the antenna; a poor inlay caps range regardless of gain.
Range grows with the square root of the power delivered to the tag, so doubling gain does not double distance.
The honest specification is not maximum range but the zone in which read rate exceeds the business threshold.

Beamwidth is the parameter that decides layout

Beamwidth is the number that converts an antenna into a floor plan.

A six dBi patch covers roughly seventy degrees in both planes, which at three metres is a zone about eight metres across.
A twelve dBi panel narrows that to forty or forty-five degrees, giving a zone about two and a half metres wide at the same height.
A wide beam sees more tags and more neighbours; a narrow beam sees fewer of both, and both effects are usually wanted.
Vertical beamwidth controls how much floor the antenna sees at a given tilt, which is what stops ceiling-mounted antennas reading adjacent aisles.
Draw the cone before ordering, including the blind spots, because a cone on paper is cheaper than a re-drill.

Circular versus linear polarisation

Polarisation is where most "the antenna does not work" complaints actually originate.

Circular antennas read tags at any orientation but give up about three decibels against a perfectly aligned linear tag.
Linear antennas gain that three decibels back but lose far more when the tag rotates, which apparel and parcel tags routinely do.
Mixed estates are common: a linear tag on a pallet and a circular antenna is a three-decibel tax on the whole project.
Circular is the safe default for handhelds and person-facing portals; linear is worth it only when tag orientation is mechanically fixed.
Whatever you choose, measure with the actual inlay, because one tag family can be far more polarisation-sensitive than another.

Six, eight, twelve dBi: what each is actually for

The honest answer is that each has a zone, and the mistakes come from using one where the other belongs.

Six dBi: shelf edges, sorting chutes, close-range batch reading, and handheld accessories where angle matters more than metres.
Eight to nine dBi: the general-purpose workhorse for doorways, conveyor portals and mid-height rack faces at two to four metres.
Twelve dBi: long, narrow reads such as dock doors, vehicle gates and tall single-face racks viewed from a distance.
Very high gain is rarely a good choice indoors, because the beam becomes so narrow that a pallet shift of thirty centimetres misses it.
Two medium antennas often beat one high-gain antenna, because they cover the zone from two angles and can be aimed separately.

Mounting: height, tilt and the Fresnel zone

Installation errors cost more decibels than any specification choice.

Mounting height sets the incidence angle; a ceiling antenna pointed straight down reads the floor, not the rack face.
Tilt of fifteen to thirty degrees is typical for reading a rack face from above and behind.
Never mount behind metal: a steel beam or a mesh shelf is a reflector, and the reflection is rarely where you wanted it.
Keep the first Fresnel zone clear; a handrail or a cable tray inside the beam attenuates more than its size suggests.
Fix the bracket properly. A slowly rotating antenna produces a slow decline in read rate that looks like a tagging problem.

Antenna gain cannot fix a tag problem

Half of the antennas sold as upgrades are bought to solve something gain does not address.

A tag detuned by metal or liquid loses more than any antenna can add; the fix is a spacer or an on-metal inlay.
Dense tag populations cause reader collisions, and a narrower beam does not help once the tags are inside it.
Weak inlays with low sensitivity cap the link regardless of gain, and the antenna simply wastes its budget on them.
Multipath nulls create holes inside the zone; moving the antenna twenty centimetres can fix it, and more gain cannot.
If the read rate is low at short distance, the problem is almost never the antenna.

A selection checklist

A short checklist prevents most of the expensive mistakes.

State the business zone in metres, then choose beamwidth, then choose gain; never the reverse.
Calculate the EIRP chain with cable loss included, and confirm it against local regulation before ordering.
Verify polarisation against the actual inlay used on the product, not against a datasheet assumption.
Prototype with one antenna on a temporary mount and move it, rather than installing six at the ceiling on day one.
Record the final position, tilt and model on a drawing, because the next installation will copy it whether or not it was right.