1. Pick the one a buyer compares you to
Not the biggest name in the category, and not the one whose ads annoy you. The one that turns up in the same shortlist when somebody is deciding between you.
One is enough to start. Research on five rivals that you abandon after a month is worth less than research on one that you keep.
2. Read what they publish, not what they say
Every company has a story about itself. It is on the About page, and it is the least informative thing they own.
What they publish is different: it is a decision they had to commit to. The price on the product page. The variant that went out of stock. The job ad that names the tool they just bought. The sitemap. Zapier publishes 122,800 URLs, measured 1 September; that list is a map of what they decided to build, drawn by them.
Their live ads are the same kind of evidence. An ad that has run for months is a test they already paid for: one public consumer brand we measure had 7,504 live ads with an average lifespan of 91 days on 6 September, and 6 per cent of them carried a discount.
3. Take the same reading twice
One reading tells you where they are. It does not tell you where they are going, and that is the part you can act on.
Take Gymshark's catalogue. Two readings seven days apart, 5,000 products in each: the same 4,827 products appear in both, and 154 of them changed availability. That is not a story you can see in either reading alone, and it took no extra work, only the discipline of keeping the first one.
Same method both times, or the difference you find is a difference in how you looked. If the first reading was the first twenty results and the second was the first two hundred, the growth you just found is your own.
4. Write down what you could not see
This is the step that gets skipped, and it is the one that decides whether the other three survive contact with a decision.
A source that failed and a source that came back empty look identical in a spreadsheet a month later. "We could not read their reviews that week" becomes "they have no reviews", and then somebody builds a plan on it.
Keep the difference. Write the hole down next to the row, with the date. It costs a line, and it is the only thing standing between a research habit and a confident guess.
What it looks like when it is done properly
Two brands, one method, both sides collected in the same run. Our own reference pair carries 197 measures across audience, products and price, paid, reviews, search, reach and AI answers, and each one names the source it came from and the window it covers. Estimates are written as bands, never as single confident numbers.
It runs every 14 days, which is the other half of step three: the second reading only exists if something takes it without being asked.
Pick one competitor. See what we find.
Your first benchmark is free. One domain, measured against you, no card and no call.