A bit is one halving of the field
One bit of decision information means a shopper can rule out half the shelf. It is not a score we invented and scaled to look impressive; it has a fixed meaning you can check.
The method
Every ranking on this site comes from one measurement, and it is reproducible from our repository with a single command. This is what it does, what the unit means, and where it stops being reliable.
Products on the shelf we measured
216
Where you are
Bits of decision information, from the shelf we measured. Processor is absent from three quarters of the shelf and still ranks third, which is the part of the method worth explaining.
They should. A vendor telling you which fields matter is making a claim about your shoppers, and claims like that are usually built from a proprietary score nobody outside the company can inspect.
Ours is not. The measurement is information theory applied to a shelf of products and a set of stated buying requirements, and both halves are checkable. If the method is wrong you can tell us why, which is a property we would rather have than mystique.
What follows is the whole of it. There is no second, better model behind this page.
If we cannot show you how a figure was produced, we do not print it. That rule is why there is no conversion statistic anywhere on this site.
The same number, after the repair
One attribute populated on the shelf we measured, and the same method run twice.
Products still sharing every recorded answer
Identical method on both runs, nothing re-tuned between them. That is the only condition under which a before and an after mean anything.
What it looks like
A bit is one halving of the shelf. That fixed meaning is why two attributes can be compared at all.
One bit halves the shelf. That is the unit, and it is why two attributes can be compared.
What happens
For each attribute, compute how much knowing it reduces uncertainty about which product a shopper means. A field recorded on nothing removes none. A field everyone shares the same value for also removes none, which is why completeness is a poor proxy.
Simulate the same attribute recorded everywhere and score again. The difference between the two is what that field is worth to you today, expressed in bits.
An attribute only matters where a real decision depends on it. Each one is scored against the stated requirements for that category, so a field no decision rests on scores near zero however empty it is.
What you get
Because the question is genuinely about information, and bits are what information is measured in.
One bit of decision information means a shopper can rule out half the shelf. It is not a score we invented and scaled to look impressive; it has a fixed meaning you can check.
Screen size and processor are not comparable as percentages complete. They are comparable as how much each one narrows a choice, which is the thing you are actually trading off.
Two attributes can be equally missing and differ enormously in what recording them would buy. Bits capture that; a completeness report cannot.
The same computation on a repaired catalogue gives a directly comparable number, which is what makes verification possible without traffic.
Before you commit
Questions
The method is only interesting if it says something about your catalogue. One category export is enough to find out.
Request a Decision Audit