Our methodology

Our homepage title carries the word best. A qualifier like that is worth nothing unless the process behind it is written down, so this page writes it down: which products get considered, which document each spec comes from, exactly how the arithmetic works, what gets a product thrown out, how the scores are built, and how often each page is checked again.

What we do not do, stated first

We do not test hardware. We do not own it, install it, mount it, drill for it, or leave it on a door for six months to see what happens. We do not accept free units from manufacturers, and no brand has ever seen one of these pages before it published. There is no lab, no jig, no battering ram, no soak room.

That is not an apology. It is the reason the rest of this page can exist. A site that has genuinely tested fifteen locks is telling you something you cannot verify — you have to take the claim on trust, and the claim is the single easiest thing on any review site to fake. What we publish instead is entirely checkable: a spec with a link to the manual it came from, a piece of arithmetic with every input on display, and a price pulled live from Amazon with a date stamp on it. You can disagree with our weighting. You cannot be misled about our inputs.

Every product page on this site repeats the same line, because the fact belongs next to the recommendation and not buried on a policy page. Where a page would genuinely be better for someone having used the thing, we say that too — and we cite owner reports and manufacturer support documentation as third-party evidence, clearly attributed to whoever actually said it. We never launder someone else's experience into the first person.

Step 1 — which products enter consideration

A product is eligible for a Sensor & Latch ranking only if all of the following are true on the day the page is built:

1. It is currently sold on Amazon.com and resolves to a live ASIN through the Amazon Creators API. Not a marketplace listing that existed last year — a live, buyable record returned by the API when our build runs.
2. The manufacturer publishes a specification document or user manual that we can link to. If the only thing available is a retailer bullet list, the product is not eligible, because a retailer bullet is marketing copy and we cannot cite it as a spec.
3. It belongs to the category the page is about at the level of mechanism, not marketing. A camera that stores footage on a required cloud plan is not a local-storage camera, whatever the box says.
4. It is not a duplicate SKU of something already on the list. Where a manufacturer sells the same hardware in four finishes or bundle sizes, one record represents the family and the page says which variant the spec figures describe.

The candidate pool for each page starts from the brands that actually sell into the category in volume, plus anything that keeps appearing in the smart-home communities where these questions get asked in earnest. It is not a search of other review sites. Restating another publisher's claim as a fact is one of the things our editorial policy forbids outright.

Step 2 — which specs we collect, and from which document

Each spec has exactly one acceptable source type. We do not average across sources or take the more flattering figure.

When a manufacturer simply does not publish a figure, the cell reads Not published. It never reads zero, and it never reads a dash that could be mistaken for zero. A missing spec is information — usually it tells you the manufacturer would rather you didn't compare that number.

Step 3 — the cost-to-run math, with every input shown

This is the number almost nobody in this category publishes, and it is the reason our rankings sometimes disagree with everyone else's. The purchase price is a one-off. What a device costs to keep alive is not.

The formula has three terms:

five-year cost = live purchase price + consumables over five years + subscription over five years

Term 1 — the purchase price

Comes from the live Amazon layer, refreshed on a schedule, and shown on the page with the date it was checked. We never type a price into prose or into a table. If the data is more than 48 hours old the number disappears from the page entirely and the link falls back to “check price on Amazon” — a vanished price is honest, a stale one is not.

Term 2 — consumables

Batteries, mostly. Two inputs come from the manual (cell type and cell count) and two are published constants we hold identical across every page so that products are compared on the same basis:

• Usage constant: 10 lock or unlock cycles a day for a lock; state change only for a contact sensor; continuous standby for a leak sensor.
• Consumable cost constant: 50 cents per AA alkaline cell, 100 cents per lithium coin cell. These are assumptions, not prices we looked up — substitute your own and every figure below scales linearly.

A fully worked example

Take a keypad deadbolt whose manual specifies four AA cells and states a 12-month life at 10 cycles a day. Every input, in order:

• Cells per set: 4 (installation manual, battery section)
• Sets per year: 1 (manufacturer's rated life at our usage constant)
• Years modeled: 5
• Cells over five years: 4 x 1 x 5 = 20 cells
• Consumable cost constant: 50 cents per AA cell
• Consumables over five years: 20 x 50 cents = 10 dollars
• Subscription required for codes, app unlock and notifications: none
• Five-year cost to run: 10 dollars, plus the purchase price shown live on the page.

Now run the same three terms on a camera whose recordings are only reachable through a per-camera monthly plan. Term 1 is often lower — subscription cameras are frequently the cheaper box. Term 3 compounds every month for five years and does not stop when you stop paying attention to it, and in most plan shapes it overtakes the hardware cost long before year five. That single piece of arithmetic is why a subscription-free camera that looks expensive on the shelf usually is not, and we lay the plan shapes out in our subscription cost comparison.

Step 4 — what disqualifies a product

These are hard exclusions. A product that trips one is removed from the ranking, not demoted within it.

• No live ASIN. If the Creators API cannot resolve it, we cannot show you an honest price or an honest availability, so it does not rank.
• Discontinued, or a permanent third-party-only listing. We sweep for this quarterly.
• No locatable specification document. If we cannot cite it, we cannot rank it.
• Core function silently gated behind a plan. A device where local recording, notifications or codes stop working when a subscription lapses is disqualified from any page in the no-subscription section, regardless of how good the hardware is.
• A specification that contradicts itselfacross the manufacturer's own documents, and that we cannot resolve with a certification listing or an FCC ID lookup.
• A published safety or security grade below the floor for its class where the product is marketed as if it met it.

Step 5 — the scoring rubric

Every ranked product carries a small set of scores from 0 to 10, and the overall figure is their arithmetic mean, rounded to one decimal place. Nothing is weighted behind your back. The metrics vary by category because the things that decide a lock purchase are not the things that decide a camera purchase.

These scores are judgments compiled from published documents. They are not bench measurements and we never describe them as such. Two people reading the same manuals could reasonably score Ecosystem 8.5 and 9.0. Nobody reading those manuals should get a different cell count, a different backset range, or a different answer on whether the thing works offline — and those are the parts of the page that actually decide the purchase.

Every ranked product also carries a Skip this if line, and it has to be genuinely disqualifying — a real reason a real buyer should walk away, not a soft non-objection. A page where nothing has a downside is a page that has stopped being useful.

Step 6 — how often each page is re-checked

Every page shows both a published date and a last-updated date, and a correction is made in place with a dated changelog line rather than edited away quietly.

Where a monitored system is the right answer

A contact sensor reports that a door opened. That is all it does, and no amount of app polish changes it. Where life safety is genuinely involved — fire, carbon monoxide, a medical alert for someone living alone — a professionally monitored system may be the correct choice even though every link on this site pays us nothing when you buy one. We would rather say that plainly than let a page imply a twenty-dollar sensor covers it.

How to check our work

Open any roundup. Take the comparison table, click through to the manufacturer documents in the sources block at the foot of the page, and confirm the cells. Then take the cost-to-run figure, plug the manual's battery numbers into the formula above, and see whether you get what we got. If the ranking still comes out differently for you, that is a legitimate disagreement about weighting rather than a dispute about facts — and it means the page did its job.

If a figure is wrong, tell us. Write to info@sensorandlatch.com or use the contact form, and point at the document. Corrections made from a primary source are the fastest thing we do.

Related reading: who writes this site and why, our editorial policy, and how we make money. If you want to see the methodology in use, the door and window sensor roundup and the smart lock roundup are the fullest applications of it.