Reader wherever it matters who entered, keypad where simplicity and no per-user cost outweigh that - and many installations sensibly have both.

A keypad needs no credential to issue, lose or replace, and costs nothing per user. For a small building with stable occupancy, or as a backup method, it works.

Its weakness is inherent. A shared code spreads: former residents, contractors, delivery drivers and anyone who observed it entering. Because changing it requires telling everybody, it is changed rarely, and the effective access list only grows.

A reader issues a credential per person. Access is granted and revoked individually, a lost card is disabled without affecting anyone else, and the system records who entered - which matters for investigating incidents.

Combined panels offer both, with the keypad as a fallback or for time-limited visitor codes.

For multi-occupancy residential, readers with individually issued credentials are the sound arrangement, and shared codes should be regarded as a temporary measure.

Modern encrypted credentials or mobile credentials - not the legacy low-frequency proximity cards that remain widely installed.

Older proximity technology transmits a fixed identifier with no encryption. Devices to read and clone such a card are cheap and readily available, and cloning takes seconds from a card in a pocket. Any system using it should be regarded as providing convenience rather than security.

Contemporary smart credentials use mutual authentication and encryption, so the card and the reader verify each other and the exchange cannot simply be copied. This is the current standard for new installations.

Mobile credentials on a smartphone are increasingly common: nothing extra to carry, issued and revoked remotely, and protected by the phone's own authentication.

Migration is the practical difficulty, since readers and cards must change together. Multi-technology readers that accept both during a transition are the usual approach - but the transition should be completed rather than left indefinitely, since the weakest accepted technology defines the security.

Individual codes where the system supports them, changed on a schedule and on occupancy change - and never a single shared code that nobody ever alters.

A shared code is the default and the problem. Every former resident, every contractor and everybody who saw it entered still knows it, and there is no way to revoke one person's knowledge.

Systems supporting individual codes remove most of that: each user has their own, it can be revoked without affecting others, and the log shows which code was used.

Time-limited codes for contractors and deliveries are a useful feature, expiring automatically rather than depending on somebody remembering to remove them.

Where a shared code is unavoidable, change it on a defined schedule and whenever an occupant leaves, with a process for informing residents.

Physical measures help: keypads with a restricted viewing angle, scrambled key layouts on higher-security installations, and positioning that is not overlooked from the street all reduce observation.

Free egress must always be maintained, and the entry behaviour depends on the lock type and the fire strategy - which is a regulated matter rather than a preference.

The absolute requirement in every jurisdiction is that people can leave. Exit must never depend on power, on a credential, or on a working system - which is why mechanical override on the inside, or a fail-safe arrangement, is part of the design.

For entry, fail-safe locks release when power is lost and fail-secure locks stay locked. Which is correct depends on the door's role in the fire strategy and on the local regulation, and it is determined by the fire risk assessment rather than by the security preference.

Doors on escape routes and those required to release on fire alarm must be interfaced to the alarm system so they do so.

Battery backup keeps the system operating through short interruptions, which avoids residents being locked out unnecessarily.

Confirm all of this with the fire strategy for the building before specifying, and record the arrangement.

Following an authorised person through an open door - and a reader panel alone cannot prevent it, though the wider system can discourage it.

The reader verifies a credential and releases the door. It has no knowledge of how many people pass through.

In residential buildings this is the normal way unauthorised people enter, and it is a social problem as much as a technical one - residents hold doors open out of politeness.

Technical measures exist. Anti-passback rules prevent one credential being used twice for entry without an intervening exit, which limits credential sharing. Door-open-too-long alarms flag doors propped open. Speed gates and turnstiles physically permit one person at a time, which is practical in commercial lobbies and not in residential entrances. Sensors that count people through a doorway can raise an alert.

For residential buildings the realistic measures are a door closer that works, an alarm on doors held open, and resident awareness - the entrance is only as secure as the willingness to let it close.

Wherever an access control system exists, yes - two separate credential databases for one building is an administrative failure waiting to happen.

A standalone reader keeps its own user list. Adding and removing people means doing it twice, and the two lists diverge - so a departed resident's credential is revoked in one system and still works on the entrance.

Integration means one credential per person, one place to grant and revoke, one audit trail covering the entrance and internal doors, and consistent permissions.

It also allows richer rules: time-based access for contractors, temporary credentials that expire, and access levels by area.

The requirement is compatible technology and a supported integration - reader protocols and credential formats must match, and this should be confirmed before purchase rather than assumed.

Where no access control system exists, a standalone panel with good management software is adequate, but the credential database should still be exportable so a future migration is possible.

More than for a video station, because this panel is touched every time somebody enters - many times a day for years.

Keys and readers wear. Backlit keys fade, membranes perforate, and printed legends disappear. Metal keypads with engraved or moulded characters last far longer than printed plastic in any position with real traffic.

Weather and hand contact combine: rain, dirt and grease on a keypad in constant use degrade it faster than the ingress rating alone would suggest, so the seals matter and so does the cleanability.

Backlighting is essential for night use and is a component that fails - check it is serviceable rather than sealed.

Impact and vandal resistance matter at street-facing entrances, and tamper detection reporting to the system provides notice of interference.

Keep a spare for any entrance serving multiple dwellings. A failed reader at a shared entrance locks out every resident, which becomes urgent immediately.