The question a single competitor cannot answer
Your traffic fell nine percent. Watching one rival tells you whether they fell too, which is useful and easily misread: they might have had a bad month for their own reasons.
Watching the category tells you something stronger. If eight brands out of ten fell in the same window, the month did it. If you fell while the pack held, you did it. Same starting number, two opposite conclusions, and only one of them is worth a meeting.
That is the entire case for market intelligence over competitor intelligence, and it is why the set matters more than any single name in it.
Share of traffic, defined so it cannot drift
Share of traffic is your estimated visits as a proportion of a set of brands you chose. Three conditions make it honest, and dropping any one of them makes the number worse than nothing.
The set is fixed and written down. Add a brand and everybody's share falls without anybody having lost anything. If the set changes, the series restarts, and the chart says so.
Every brand is measured the same way. Your own analytics against their modelled estimate is not a share, it is two instruments divided by each other. Use the modelled figure on both sides, or state clearly which side is which.
It carries the estimate label. Panel-modelled traffic is never a fact about a competitor, and a share built from estimates is an estimate too, however precise the percentage looks.
Read the direction and ignore the decimal. A share that moves half a point is inside the model's own error, and the limits of traffic estimates apply to every brand in the set at once.
Trends are only visible against a fixed frame
An industry trend is not a headline, it is a line: the same measure, on the same set, in the same window, repeated. Publishing cadence across the category. The share of brands running a discount. How many have added a free tier. Where prices sit this quarter against last.
Two traps sit here. The first is a scale that is normalised per brand: some popularity indices rescale each brand to its own peak, so the numbers look comparable and are not. Adding those across a category produces a chart that means nothing at all, confidently. The second is survivor framing: a category set built from today's visible brands quietly omits whoever left, which makes every trend look healthier than it was.
What it is good for
Attribution. Your move or the market. This is the main one and it pays for the rest.
Timing. A category that starts adding a cheaper tier all at once is telling you something about demand months before anybody writes an article about it.
Sizing. Share tells you which fights are winnable. Two percent of a category is a different strategy from twenty, even when the revenue looks similar this quarter.
Sanity. A pack norm is a check on your own targets. It is not a target, and treating it as one is how companies end up chasing an average nobody actually is.
What you can do today, and what is honestly still ahead
The parts that exist now: every tracked brand is measured on the same schedule with the same collectors, the Benchmark puts them side by side with estimates labelled, and the category column groups them so a comparison stays within a like set. That is enough to answer the attribution question for a set you define.
The part we will not pretend about: a rolled-up industry aggregate, published as a category index, is not a product surface yet. Building it before the collection is symmetric across every brand in a set would produce a number that describes our coverage rather than the market, which is the failure this whole product is organised against.
Pick one competitor. See what we find.
Your first benchmark is free. One domain, measured against you, no card and no call.