evryproduct.com/how-we-verify

Methodology

How we decide that two listings are the same object

Every price comparison on this site rests on one claim: that the product on one marketplace and the product on another are the same thing. That is a strong claim, so this page sets out exactly what we check, what each check can tell you, and — the part most comparison sites leave out — what we cannot establish at all.

What this page is

The reasoning behind every match confidence figure on the site.

What it is not

A guarantee. It is a statement of how much evidence sits behind a match.

The checks

Four things have to agree

A name and a photograph are the two easiest things for a merchant to change, so neither settles a match on its own. Each check below answers a different question, and each has a limit we are explicit about.

01

Image match

Every photograph of the product is compared against every photograph of every candidate, by two independent visual models. Products made from the same mould photograph alike in ways that survive different lighting, backgrounds and crops.

What it gives usA measure of how closely two objects resemble each other, across every image either listing publishes.
What it does not proveThat they are the same object. Merchants reuse manufacturer photography freely, so the same picture can sit above a different size, material or product entirely.
02

Product type

The candidate has to be the same kind of thing. This is the check that catches what image matching alone lets through — a storage bag photographed like a bath toy scoop, beads of one material listed under beads of another.

What it gives usA hard stop. If the type does not match, there is no match, whatever the pictures look like.
What it does not proveThat the quality is the same. Two items can be the same type and the same shape and still differ in what they are made of.
03

Merchant trace

We look for every merchant carrying the unit rather than stopping at the first. A product listed by several independent merchants at similar prices is a different proposition from one listed once.

What it gives usHow many merchants carry it, how long each has been trading, and what buyers report about them.
What it does not proveHow your particular parcel will arrive. A merchant's history describes the merchant, not your order.
04

The same variant, and what is in it

A listing is not one price. It is a set of options — size, colour, count, bundle — each priced differently, and the figure you see first is usually the lowest-priced of them. We read the option matrix on both sides, pin the option that corresponds, and read how many units are actually in the box before any percentage is calculated. A six-pack against a single unit is not a saving.

What it gives usA comparison between two specific options, on a stated basis — per item, or one unit against a whole pack — shown on every product.
What it does not proveWhat it costs to get it to you. We compare item prices; we do not calculate shipping, duties or taxes to your country.

Options and contents

Why a listing price is not a price

The single number on a listing is the lowest-priced thing that listing sells. It is rarely the thing you came for.

79%
Of listings we check have options that span at least half again in price
2.3×
Median gap between a listing's lowest-priced option and its highest
48%
Of matches ship a different number of units on the two sides

A pack of ten and a single unit can sit on the same page under the same photograph, and the price shown in search results belongs to whichever costs least. The widest single listing we have measured runs 79× from its lowest-priced option to its highest. Comparing two listings by their headline figures therefore compares two objects that are not the same object, which is how a comparison can be arithmetically correct and still wrong.

So the option is pinned before anything is compared. We read what each side actually ships — how many, what size, what is included — and state the basis on every product: per item when both sides sell the same unit, or one unit against a whole pack when they do not. Where the contents cannot be read on both sides, no percentage is published at all.

This is the part that changes an answer. A pack of six watch bands at $8.49 works out at $1.42 each, which beats a $2.57 single elsewhere — so the mainstream listing wins, and we say so. Read on the sticker price alone, the same two listings look like a saving of more than two thirds in the other direction.

The score

What the confidence figure means

The four checks combine into a single number between 0 and 100. It describes how much the evidence agrees — not how much we would like it to. The median across everything we have matched so far is 79%.

Above a floor, a match is described one of two ways, and the difference matters more than the number does:

Same item

The same object, under a different name

Same mould, same construction, usually the same manufacturer listing with a different logo on the box. This is the only case where we use the word identical, and the only case a saving figure is described as being for the same item.

Equivalent

The same kind of thing, made by someone else

It does the same job to the same specification, but it is not the same object. Worth buying, often worth much less money, and we label it separately so you are never told you are getting the same thing when you are not.

Roughly 47% of the matches we publish are Same item; the rest are Equivalents. Below the floor we publish nothing at all — an uncertain match is worse than no match, because it is the one that sends you to the wrong product.

Limits

What we cannot verify

We never hold the product. Everything on this site is drawn from listings, photographs and what buyers report, which leaves real gaps. These are the ones worth knowing about before you spend money.

Materials

A photograph shows shape and finish. It does not show what a plastic is, how thick a steel is, or how a fabric is woven. Two objects can match closely on every image and still differ in what they are made of.

Safety testing and certification

We do not test anything, and we do not verify certification claims. In categories where testing and liability are a genuine cost — child safety, electrical, anything load-bearing — that is a large part of what a brand premium buys, and we say so rather than treating the objects as interchangeable.

Batch variation

A merchant who shipped well last year may not this year. Trading history and buyer reports describe a pattern, not your particular parcel.

Whether a listing stays the same

Listings change under the same address — a merchant can swap the unit behind one. We re-check on a fixed cycle, but between checks a listing can become something else.

Review authenticity

We read what reviews say rather than what they average, which helps. It does not let us certify that any individual review is real.

Delivered cost

We compare the price of the item. Shipping, duties and local taxes are not calculated, so the figure you see is the price of the object and not the price of getting it to your door.

Whether a branded unit is authorised

For a branded product we can tell you an identical-looking unit is listed elsewhere. We cannot tell you whether that particular merchant is authorised to sell it.

Anything about the future

Prices move, stock runs out and merchants close. Every figure on this site carries the date it was checked, and that date is the whole of what we are claiming.

When the evidence is thin

What we do when we are not sure

Uncertainty is common. What matters is what happens next, so these are fixed rules rather than judgement calls.

WhenWhat happens
The product type does not matchNo match at all, whatever the images look like
The visual match is not strong enough to claim identityPublished as an Equivalent, never as the same item
The option on one side has no counterpart on the otherNo comparison is published for that product
Pack size cannot be read on both sidesNo percentage is shown
The same item costs more at the alternative merchantWe tell you to buy from the mainstream listing
A price is older than one full re-check cycleIt is withdrawn rather than shown with a stale date
Only one merchant carries the unitThe match is shown, and the fact that it rests on one merchant is shown with it
The fourth rule is not a formality. On the categories we have worked through so far, close to half of everything we check ends with us telling you to stay where you are. A comparison that can only ever reach one conclusion is not a comparison.

Freshness

How current a price is

The catalogue is re-checked completely every 10 days. One tenth of it is checked each day, so something is always being verified and nothing waits longer than one cycle.

12345678910

Each block is one day of the cycle. Today's slice is highlighted.

Every listing shows its own date. We would rather tell you a price was checked six days ago than imply it is live. Prices move, and a comparison you cannot date is a comparison you cannot check.

Worked example

The whole process, on one product

Everything above, applied to a product you can open and check for yourself.

TOHIBEE swimming turtle bath toys, 3 pack — Amazon lists a pack of 3 at $9.99, which is $3.33 each. We found the same item at $2.09 each, carried by 8 merchants.
Image match
86%
across all images
Product type
Same item
type confirmed
Merchants
8
carrying the unit
Per item
−37%
$2.09 vs $3.33

Combined confidence 90%. The saving is stated per item, because Amazon sells this one in a pack and the alternative does not — comparing the sticker prices would have produced a much larger number and a much less useful one. See the full comparison →

Where you will see this

Every confidence figure on the site links here

On a product page, a cluster page or the home page, the match confidence figure is a link back to this page. If a number is being shown to you, the reasoning behind it is one click away.

Questions this page does not answer are covered in the common questions on the home page, or you can write to us.