Keypad, fingerprint, app or key card: how each one fails

Stand at your own front door with a bag of groceries in one hand and you find out quickly which unlock method you actually trust. Most keyless locks offer three or four ways in. Owners settle on one of them within about a week and meet the rest only on the morning the favorite refuses.

So the useful question is not which credential is fastest or most modern. It is how each one fails, how often, and what you are left holding when it does. Keypads, fingerprint readers, phone apps and key cards fail in completely different ways, and the combination worth paying for depends more on who lives in the house than on anything printed on the box.

Why nearly every lock ships with more than one method

A credential is the thing you present. The lock is the thing that decides. Those are two separate pieces of engineering and they fail separately, which is why a correct code can be refused by a reader that is not registering presses, and a working reader can refuse a code that was quietly deleted three months ago.

Manufacturers stack methods because every credential type has a group of people it suits badly. A keypad is hopeless for a four-year-old who cannot keep a secret. A fingerprint reader is hopeless for a contractor you will never meet in person. An app is hopeless for a visiting parent who does not want another account on their phone. So the lock carries two or three readers and usually a keyway, and you assign credentials per person rather than per household.

If you have not settled on a lock yet, it is worth sorting out the five broad categories sold as keyless entry first, because the reader options differ sharply between an interior retrofit unit and a full replacement deadbolt.

Keypads: the pad is rarely the part that fails

The keypad is the default on almost every keyless lock sold in the US and the method most owners end up using daily. There are three physical forms. Mechanical pushbuttons drive the bolt through a mechanical linkage with no battery involved at all. Rubber or membrane buttons sit over a small circuit board. Capacitive touchscreens respond to the electrical properties of a fingertip rather than to pressure, which is why they ignore a gloved hand.

Worn buttons and the four digits on display

The oldest attack on a keypad needs no equipment. Four buttons out of ten get pressed several times a day for years and the other six do not, so wear, or a film of finger grease, narrows ten thousand combinations down to a couple of dozen arrangements. Manufacturers answer this two ways: touchscreens that require a full-hand wake gesture, which smears the surface evenly, and firmware that accepts padding digits, so you can type a long run of noise with the real code buried inside it. Both help. Neither helps if the code is 1234, and wear on a mechanical pushbutton lock is permanent.

Shoulder surfing is the version that actually happens: a delivery driver, or a teenager’s friend waiting on the step behind you. Thermal residue attacks, where a camera reads the warmth left on the pressed keys, turn up regularly in security research and almost never on a residential porch. The code that gets used against a house is nearly always one that was handed out deliberately and never withdrawn.

Lockouts, expired codes and codes that never saved

The other keypad failure is the lock refusing a code you are certain is right. Most models impose a timed lockout after a run of wrong entries, commonly around thirty to sixty seconds after three to five failures, with the exact figures set out in the manual. During that window the correct code fails too, which is baffling if you did not know the feature existed.

Codes also go missing. Programming sequences are unforgiving, and missing the confirming press at the end leaves the lock beeping as though it accepted a code it never stored. Scheduled codes expire on a date somebody set and forgot. A factory reset wipes every user code and puts the lock back on its default programming code. When a code that used to work stops working, the first job is to establish whether the keypad is dead or the code is being rejected, because those are two unrelated faults that arrive as one complaint.

Fingerprint readers: the false rejects are the problem

A fingerprint door lock has obvious appeal. Nothing to remember and nothing to carry. The sensors fitted to residential locks are almost always capacitive: a dense grid of tiny plates measuring the difference in charge between the ridges touching the surface and the valleys sitting a fraction of a millimeter above it. That mechanism explains every one of its failure modes.

Skin that will not cooperate

A capacitive sensor wants a fingertip carrying a normal amount of moisture. Too much water and the valleys fill in, flattening the image the sensor gets. Too little and the contact is patchy. So a reader that behaved all summer starts refusing in January, when hands are dry and cold and the ridges are pinched. Washing up, gardening, plaster dust, a fresh cut across the enrolled finger: all of it shifts a fingertip outside the range the sensor likes, and a cut takes weeks to grow out.

Gloves are a flat no on a capacitive reader, and so is a thumb still wet from washing the car. Neither is a defect, but both matter if a fingerprint reader is going to be the only way in.

The number on the box is the wrong number

Fingerprint locks advertise a false acceptance rate, usually something on the order of one in fifty thousand, and it is the least interesting figure in the listing. False acceptance means a stranger’s finger being read as yours, which requires a stranger at your door pressing fingers against the sensor. The number that governs daily life is the false rejection rate: how often your own finger gets refused. Manufacturers rarely publish it, and the threshold is adjustable in firmware, so a lock tuned to be hard to fool is also a lock that says no to you more often.

Enrollment does more for daily reliability than any published figure. Manufacturer instructions generally ask for the same finger to be recorded several times at different angles and pressures, and the owners who skip that step are the ones reporting a reader that works one time in three.

Fingers that read badly

Some fingerprints do not read reliably on any sensor. Young children have small, shallow ridge detail that keeps changing as they grow. Biometrics research has long recorded the same difficulty at the other end of life, where ridges flatten and skin dries out. Work that abrades the fingertips, from bricklaying to kitchen shifts to constant hand washing, produces prints that read inconsistently from one week to the next.

None of that makes a fingerprint reader a bad choice. It makes it a bad sole choice.

The phone app: two batteries and a handshake

App unlocking covers two quite different mechanisms. One is a Bluetooth handshake at the door, where the phone talks to the lock directly and nothing else is involved. The other is a command sent over the internet to a Wi-Fi lock or a bridge, which is what lets you open the door for somebody while you are at work.

The Bluetooth version fails slowly rather than dramatically. Waking the phone, opening the app and waiting for the connection takes a few seconds, which sounds trivial until you have done it while holding a sleeping child.

Auto-unlock, where the door opens as you walk up to it, depends on a geofence, on location permission set to always, and on background activity that both iOS and Android are willing to suspend to save power. Owners commonly report it working for months and then quietly stopping after a phone update reset a permission. Treat it as a convenience rather than as a way in.

The failure everybody eventually meets is a dead handset. A phone at zero percent is not a credential. Phones carrying a door key in the wallet are a partial exception, since Apple documents a power reserve that keeps express credentials working for a while after an iPhone shuts down on low battery, though the window is short and the feature has to be set up in advance.

Guests are the other soft spot. Sharing an app key means the guest installs an app, creates an account and accepts an invitation, which is a lot to ask of a plumber who will be at the house for forty minutes.

Key cards and fobs: cheap, convenient, easy to copy

Cards and fobs are the quietest credential. Nothing to remember, nothing to charge, and a card in a coat pocket works with cold hands and gloves on, which is exactly where a fingerprint reader gives up. Residential locks that accept them nearly all use passive RFID at one of two frequencies, and that difference matters more than anything else printed on the card.

Low frequency 125 kHz cards, the old proximity format still bundled with plenty of hardware, carry a fixed identity number and no meaningful cryptography. Duplicators that copy them cost less than a takeout meal and are sold openly. High frequency 13.56 MHz cards can perform real authentication, but only where the lock uses a current implementation; the older MIFARE Classic scheme has been publicly broken for years and still turns up in inexpensive hardware.

Owners often say a fob has demagnetized. Passive RFID has no magnetic storage to lose, so that is not what happened. Cards fail because they crack in a back pocket, because the antenna wire inside snaps, or, far more often, because the card was lost and nobody removed it from the lock. A lost card that is still enrolled is a working key in a stranger’s hand.

Cloning belongs to the same family of close-range attacks as the radio relay tricks in our look at what can realistically be hacked on a smart lock, and it has the same shape: possible, proximity-bound, and much less likely than a credential you issued and forgot.

Choosing a pairing for the people who use the door

Locks are sold on the number of methods they offer. What matters is which two you will actually rely on, and whether those two fail for unrelated reasons. Some common households:

• Young children: a code wins. Small fingers give shallow ridge detail that changes as they grow, and a child with a phone key is a child with a lost phone.
• Cleaners, dog walkers and trades: a scheduled or single-use code, deleted when the job ends, with no hardware handed over and nothing to chase up.
• Guests staying a few nights: a code again, and their own rather than yours, for the same reason an app invitation is a poor fit.
• Older residents, or anyone with dry or worn hands: large mechanical buttons with a positive click, or a card. A capacitive touchscreen with no tactile feedback is the worst option here.

Underneath all of that sits one pattern. Fingerprints suit the two or three people who live there and use the door every day. Codes suit everybody else. Cards suit people who need entry without a phone and without memorizing anything. Apps suit remote administration far better than they suit standing at the door.

What is left when the electronics stop answering

Every credential here depends on the lock having power, and a lock running on weak cells misbehaves long before it dies: slow reads and refused codes turn up well ahead of a motor that will not finish throwing the bolt. What drains a lock, and how long a set of cells realistically lasts, is a subject of its own. Before the day it matters, find out which of these your lock has.

• A mechanical keyway, with a key kept somewhere other than inside the house.
• External contacts, usually on the underside of the exterior escutcheon, that take a 9V battery held against them to supply enough power for one entry. Several major brands provide this and barely mention it.
• A code that works entirely offline, independent of the app, the bridge and the internet. Almost every keypad lock has this. An app-only lock, by definition, does not.

Common questions

Is a fingerprint door lock safer than a keypad?

For the threats a normal house faces, the gap is smaller than it looks. A fingerprint cannot be shoulder-surfed, passed around a group chat or worn into the buttons, which is a real advantage. Against that, most fingerprint locks also carry a keypad and a keyway, so the lock is only as strong as the weakest credential enrolled on it. Physical security still comes from the bolt and the frame it goes into.

Why does my fingerprint work sometimes and not others?

Almost always the state of the fingertip rather than a fault in the lock. Cold, dry, wet or freshly washed skin all change what a capacitive sensor sees. Delete the enrollment and record the same finger two or three times at slightly different angles and pressures, then add a backup finger on the other hand.

Can someone clone my key fob?

If it is a 125 kHz proximity card, yes, using hardware anyone can buy, though they need the card in hand or a reader held very close to it. Where the lock uses a current 13.56 MHz implementation, copying is considerably harder. Either way, the risk that actually turns up is a card that went missing and stayed enrolled.

What happens if my phone dies and the lock has no keypad?

You do not get in on the phone. Interior retrofit units keep the original cylinder, so you still have the original key, which is one of the better arguments for that design. A lock with no keypad, no keyway and a flat phone is a locked door, so keep a physical backup somewhere off the property.

Do I need a lock with all four methods?

No. Two is the useful number: one you use daily, one that works when the first does not, chosen so they fail for different reasons. A keypad with a keyway behind it covers most homes. Fingerprint plus keypad suits a household that hands out access often.

The short version

Keypads fail through worn buttons and codes that were handed out and never withdrawn, which makes them an administration problem rather than a hardware one. Fingerprint readers fail on the state of your fingertip, so enroll several fingers several times and never let one become the only way in. Phone apps fail with the phone, and auto-unlock is a convenience feature rather than an entry method. Cards fail by going missing while still enrolled, so delete a credential the day it stops belonging to someone. Pick two methods that fail for unrelated reasons, and keep a mechanical key off the property.