How to Compare Prices Online Without Losing the Evidence

What has to be true before two offers can be compared at all, and why that is rarer than it sounds.

A comparison needs more than two prices

Two numbers can always be subtracted. That does not make them comparable. For a price comparison to mean anything, the two offers have to be the same product in the same configuration and condition, from sellers whose obligations to you are similar, with cost components measured the same way, checked recently enough to still be true.

Most of what is presented online as price comparison is missing at least one of those, and the missing one is rarely disclosed.

Identifiers do not join retailers as often as you would think

The obvious way to prove two listings are the same product is a shared identifier. In practice, the major shopping feeds issue a different product identifier per merchant. We measured a case where four offers for a single product carried four distinct identifiers, only one of which matched a catalogue entry.

That is why identifier matching alone cannot group offers across retailers, and why grouping has to fall back to the manufacturer model number carried in the title. It is also why a comparison engine that claims to match on identifiers is either doing something more complicated than that, or matching far less than it implies.

Condition and seller are part of the product

A new unit from a retailer and a used unit from a marketplace seller are not the same product at two prices. They differ in warranty, in returns, in who answers when something is wrong, and in what happens in month nine.

Any comparison that places them side by side ordered by price is asking you to treat those differences as worth exactly the price gap, without ever stating them.

Why two retailers is the threshold

One retailer with several listings is not a comparison, it is a catalogue. A like-for-like comparison needs at least two independent sellers offering the same item, both recognised, both with enough confirmed price components to be measured against each other.

That bar is high enough that many searches do not clear it. We say so when it happens rather than presenting a single-retailer result as though it settled something.

Keep the evidence attached to the number

The failure mode in manual comparison is the same as in software: the number survives and the context is lost. You remember 219 dollars and forget that it was the 256 gigabyte model, from a marketplace seller, with delivery unresolved, checked on Tuesday.

Write the context next to the number. Configuration, seller, condition, what is included in the figure, and when you looked. Four short notes, and the comparison stays honest a week later.

A sequence that holds up

Write down the exact configuration, including capacity, colour, generation and condition. Identify which listings are the same item rather than the same product family. Separate confirmed costs from costs that will only appear at checkout. Note when each price was last checked. Compare only what is left.

Then finish at the retailer, because the retailer basket is the only place the real number exists.

Related guides