How to do a competitive analysis, step by step

Step 1. Write down the decision it has to serve

Not "understand the landscape". A decision: whether to launch a cheaper tier, which market to enter next, whether to move the price, what to publish this quarter.

An analysis without a decision attached expands to fill whatever time it is given and ends as a deck that gets admired once. With a decision attached, most of the work turns out to be unnecessary, which is the point.

Step 2. Pick the set, and keep it small

One direct rival that buyers weigh you against. One or two companies growing faster than the category, wherever they are. Optionally one adjacent player whose economics differ from yours.

Three or four names is enough. A wider set produces averages, and an average competitor is not a company anybody is losing deals to. If the list is all rivals, expect confirmation rather than ideas, for reasons worth reading before you start.

Step 3. Take the free sources first

A surprising share of the answer is public and costs nothing: their tech stack, their legal and policy pages, the shape of their site from its own sitemap, the catalogue and the prices on it, the claims on the homepage, recent press, open roles, changelog or release notes.

Do these before paying for anything. They are complete rather than sampled, which means you can say which pages, which prices and which products, not just how many. That distinction decides how much weight a finding can carry later.

Step 4. Add paid sources only where the decision needs them

Ad libraries, review text at volume, social and AI answer engines cost money per run. They are worth it when the decision depends on them and wasteful when it does not.

Two rules keep this honest. Never let a sampled source pretend to be a census: if you read fifty of their ads, that is a sample, and reporting fifty as their ad count reports your settings rather than their spending. And never present a modelled number, such as a traffic or spend estimate, without the label, because an estimate standing bare next to a measured figure is the fastest way to make a good analysis unusable.

Step 5. Turn observations into gaps

An observation is "they publish weekly". A gap is "they publish weekly, we publish monthly, and their comparison pages rank for terms we have never written about".

Go category by category: price and packaging, catalogue, content and search, ads and creative, reviews and reputation, hiring, technology, presence in AI answers. For each, write the difference and the direction, and do it before you start judging which ones matter.

Step 6. Sort by width against cost, and cut the list

Wide and cheap is this quarter. Wide and expensive is a strategy discussion, not a task. Narrow is a distraction no matter how irritating it is.

Then cut what you should not copy at all. Some gaps exist because your competitor made a decision that suits their cost base and would be wrong for you, and closing those turns you into a worse version of them. The output of this step should be three or four moves, each with a measure and a date, not a list of forty parity items.

Step 7. Repeat it, or accept the shelf life

This is the step that separates an analysis from a habit, and it is the one almost nobody does.

A competitive analysis is a photograph. It is accurate on the day, and it decays quietly: prices move, pages appear, the messaging drifts, a tier shows up without an announcement. Six weeks later people are still quoting it, and it still sounds right, which is the dangerous part. Weekly is enough. The value is not in any single reading but in the line, and the archive is worth more at month six than the first run is on day one, which is the only real argument for starting now.

The mechanical version of steps three, four and seven is what this product is. The free sources run for nothing on a first audit, the comparison lands on Benchmark, the week over week movement on Changes, and the sorted list of moves on Actions. Step five and six are still yours, and should be.