Access control and fingerprint terminals: what you are buying besides the device

A company buys a fingerprint terminal, it goes on the wall beside the door, and it records attendance exactly as promised. Then the questions nobody asked before the purchase arrive: the door does not close by itself, so it stands open half the day; the power goes off and nobody knows whether that door is supposed to open or stay locked; and at the end of the month a report is printed and typed back into a spreadsheet line by line. The terminal works. The system does not.

What follows is for a company about to buy and not yet sure what the pieces are. The order of the decisions is deliberate: it starts at the door, not at the terminal.

The terminal is one of five parts, and four of them decide whether it works

The thing that carries a model name in the catalogue is the reader: what you put your finger or card on, or look at. It is the only part the user ever sees, and the least decisive of the five in whether the system succeeds or fails. The other four:

• The lock — the part that actually holds the door. Two common types fail in opposite directions, and this is the single most important decision in the whole subject.
• The power — a supply for the lock and the reader, and a written decision about what happens to the door when it fails.
• Getting out from inside — a button, a motion sensor or a second reader. This is what makes the door a door rather than a trap, and what decides whether you have an exit time or only an entry time.
• The cabling — what runs between all of the above, and the only part that cannot be changed once the ceilings are closed and the finishing is done.

A quotation that says "lock, exit button and installation" on one line has hidden four decisions in five words.

The lock: a magnetic lock and an electric strike fail in opposite directions

A magnetic lock is a steel plate that holds the door magnetically and needs continuous power to keep holding, so when the power goes it releases the door and the door opens. An electric strike, in its common form, works the other way round: the door is held mechanically, and power reaches it as a short pulse that only releases it, so when the power goes it stays locked. Strikes are also made in a version that releases on power loss, which is why the right question to a supplier is not "magnetic or strike?" but "does this door open or stay locked if the power is cut?" — with the answer written down for each door by name.

This is a fire-safety decision before it is a security one. A door on an escape route must not hold anybody inside because the power failed or the controller died. A stockroom or a server room, on the other hand, you may want locked under all circumstances, and then a mechanical override with a key whose location is known is essential. Which door is which is for the person responsible for safety at the premises to review, and it goes into the plan before the lock is bought.

Tying the locks to the fire alarm is what settles it in practice: a signal from the panel cuts power to the locks on escape doors so they all release, regardless of the state of the mains or of the controller. Alongside it, a manual emergency release on the inside face that breaks the lock circuit directly. Both items fall out of most quotations because nobody asked about them.

Power: the battery reverses the behaviour you were relying on

The lock has a supply rated for its current, with a battery to carry it through a power cut. The battery is an obvious item, and it has a consequence people miss: if you were relying on a power cut to release the magnetic lock, adding a battery has removed that entirely. The door now stays locked while the mains is down, which is the opposite of what you were counting on. So emergency release must not depend on the power failing. It depends on the alarm signal and on the manual button.

Two further points spoil a lot of doors: a magnetic lock cannot hold a door that is not shut, so a door with no closer stands open while the lock works away meaninglessly. And thin cable over a long run drops part of the voltage, so the lock holds weakly or can be pushed open. The distance from the power panel to the door is part of the specification, not a detail left to the technician on fitting day.

Add a door position sensor: a small part that knows whether the door is open or shut. Without it the system knows nothing about a door wedged open with a block of wood, which is the commonest way people get round the whole system.

Getting out from inside: a button, a sensor or a second reader

Most doors read in one direction: a reader outside, and free exit from inside.

• The button is simple and fails visibly: a stuck button leaves the door unlocked, and you can see that fault.
• The motion sensor is cleaner to use and harder to set up: too wide a field opens the door for anyone walking down the corridor, and it can respond to movement from outside through a gap or a glazed panel. Its field is adjusted after fitting, not assumed before it.
• A second reader on the inside face costs more and adds work, and it is the right choice if you want an exit time for every person, or if you want to stop a card being passed back out to somebody waiting outside.

This is where most attendance projects break: a door with one reader records the entry and not the exit, so you reach the end of the month with half a record. If what you need is an in and an out that a wage is calculated from, reading both directions is decided before the purchase, not after the first month-end close.

The cabling: the decision you do not get to redo

Every door needs more than one run: a data or network run to the reader, a separate run to the lock with a conductor sized for its current over the real distance, a run to the exit button, and a run to the door sensor. A spare run at each door costs little now and saves tearing walls open a year later.

And one detail decides whether the system is secure or decorative: the lock wiring must not be within reach of somebody standing on the unsecured face of the door. An exposed wire that can be cut or bridged from outside turns the door back into an ordinary door and the reader into a box on the wall. These runs go in with the network cabling before the ceilings close; postponing them until after the painting means paying for them twice.

Fingerprint, card and face: three methods that fail three ways

No method of identification is better than the others in the abstract. One of them suits this door and these people. Each fails in its own shape:

• Fingerprint is the cheapest and commonest, and it fails on the same hands every morning: fingers that are too dry or wet, prints worn down by manual work, hands that handle cement, oil or cleaning chemicals. A terminal that rejects one attempt in every few does not just produce complaints; it produces a queue and the habit of one person holding the door for everyone else — and that habit alone cancels both the record and the permissions.
• A card works with any hand in any conditions and fails at something else: a card can be lent. It records that the card went in, not that its owner did, and nothing fixes that except a camera or a reader on both faces. On top of that, older low-frequency proximity cards are copied with cheap duplicators that are easy to obtain, so if a card is what protects a room that matters, ask what type it is and what encryption it uses rather than what it looks like.
• Face solves the hand problem completely: it works with gloves, with dirty hands, without contact. It fails on light: a terminal mounted facing a window or a glazed door meets strong backlight and refuses. Enrolment distance and mounting height matter too, and so do helmets and masks in some places.

A factory floor is not an office: hands there are harder, and the terminal is sometimes mounted outdoors in sun and dust, so the device's dust and water rating is a line in the specification. The honest conclusion is that one site may need two methods: card or face at the production gate, fingerprint in the office. That is two devices, not one, and we say so before the quotation rather than after a month of complaints.

Standalone or networked: where the data lives

A standalone terminal keeps everything in its own memory: the enrolled staff, the fingerprint templates and the entry log. It is administered from its own keypad, and data comes off it on a flash drive. A networked terminal sends its log to software and a database on a server or a machine of yours, and all the doors are administered from one screen.

The difference shows itself on one day: the day the terminal is stolen or burns. The standalone one takes the entire attendance history and all the templates with it, so you enrol every employee again from scratch and you have nothing to rely on in a dispute about a month that has passed. The networked one loses a reader, you buy another, and the record is still in the database. And fingerprint templates generally do not move from one brand to another, so keeping one brand across the doors saves re-enrolling your staff — which saves more than picking any particular brand does.

Even so, standalone is the right answer in real cases, and we say so although it makes the job smaller: one door, a handful of people and no link to payroll does not justify a server, a database and software to administer. Networked earns its trouble when the doors multiply, when permissions change often, or when attendance has to reach payroll. Either way, the terminal stores its log while the network is down and sends it when the network returns, and that means nothing if nobody ever pulls the data off. Fingerprint and face data belongs to the people it describes, not to the operation, so write down who may read it, how long you keep it and where its backup lives.

Attendance is worth nothing unless something reads it afterwards

The terminal does not produce attendance; it produces time stamps: a name, a time and a direction. Turning those into worked hours, lateness, absence and overtime needs written rules: shifts, the grace period, rest days, work on a public holiday, and how a missing punch is handled when somebody forgets to clock out. Those rules are the real system, and the terminal is a source of numbers for it.

The software bundled with the terminal is usually enough to print a report and not enough to close a payroll. And a report that is printed and then typed back into a spreadsheet reintroduces exactly the error you bought the terminal to remove, with the difference that the error now looks documented. Ask one question before buying: who corrects a wrong record, and does the system log who changed it and when? A record that can be edited without a trail stops being evidence in a dispute, which is the purpose it was bought for.

And here is the truest thing we can say about our own service: ZEMAVO Technology fits access control and fingerprint attendance systems and also writes ZHR, our HR and payroll system, which reads them. The company that puts the terminal on the door is the one that writes the system reading its log, so attendance reaches payroll without anyone retyping it into a file. The session where the rules are written comes before the installation, because it is what decides whether you need both directions read on that door.

The administrative failures outnumber the technical ones

Systems that fail after a year do not fail because the hardware broke. They fail because nobody revoked the access of a person who left, so the enrolled list becomes a list of everyone who ever worked there. Because the administrator password is one password everybody knows, so nobody knows who added whom. And because nobody reads the log, so rejected attempts at the stockroom door at dawn sit in the database and are never seen.

So the most important part of the handover is not the hardware: a door plan with each lock's behaviour written down, a file of the permissions as configured, and training for whoever will add a joiner and revoke a leaver — a job that comes round every month and should not wait on an engineer. What follows that is covered by a written support and maintenance contract naming who reviews the log and who checks the batteries, because a dead battery in a power supply does not report itself until the first power cut.

The brand, and the price question

Most people who ask, ask by brand name, and ZKTeco is the name typed into search more than any other. It is a real brand and a common one in the Egyptian market, and we are not here to recommend it over another or to recommend a model: the brand is the smallest decision on the list. An excellent terminal on a door with the wrong lock, no exit button and thin cable is a failed system with a good device in it.

As for price, there is no figure to give before the following is known, and every item here moves the cost more than the model name does:

• How many doors, and the type of each: timber, glass or metal, hinged or sliding.
• The lock type per door and its behaviour on power loss, and tying escape doors to the alarm panel.
• Reading one direction or both, and the method of exit from inside.
• Whether the cabling already exists or has to be run, and whether the place is finished or still under construction.
• Standalone or networked, and where the database and its backup live.
• The identification method at each point, how many staff are enrolled, and whether attendance has to reach payroll.

Two quotations that each carry a single line reading "access control and fingerprint system" are not comparable. Ask for the first thing in the quotation to be the door schedule, not the model name.

We are in New Maadi, beside the Technology Village, and the survey happens at your site because this schedule cannot be filled in over the phone: we walk the doors, see the type of each one, the cable route and the power source, and write down how each lock behaves before it is bought. Start with one door if you like, but start at the door.