How to analyze competitors with Claude: one table that shows where you win

You open a competitor's site, read the pricing, scroll through reviews, and half an hour later your head is mush: one is cheaper, another ships faster, a third shouts louder in ads. Turning that into a decision about what to change in your own business never quite happens. Claude helps turn scattered notes into one table where you can see at a glance where you win and where you fall behind. Here it is step by step: what to collect, what to ask, and which file to keep.

Why analyze competitors at all if it already feels obvious?

The feeling that it's all obvious usually rests on a couple of vivid details: someone's banner stuck in your mind, someone's angry review did. Decisions built on scraps like that come out crooked. The point of the analysis is not to copy others but to see your own position soberly: where your offer is genuinely better and worth saying louder, and where you sag and the customer leaves without a word.

Claude doesn't replace you here, it does the boring part: it folds your observations into a structure, spots contradictions, and highlights the empty cells where you know nothing about a competitor. You still do the thinking, but now over laid-out facts instead of memory.

What to collect before opening Claude?

Claude doesn't go guessing for you, it works with what you give it. Gather raw material on 3 to 5 competitors, 10 to 15 minutes each:

• What they sell and at what price. Copy the pricing or plans straight from the site, don't paraphrase.
• How they describe themselves. The headline on the homepage, the promise, who the product is for.
• What customers say. Five to ten reviews, both glowing and angry. The angry ones are more valuable: they show where the competitor has a hole.
• How they sell. Free trial, installments, delivery, guarantee, response speed.
• Where they show up. Social media, ads, blog, newsletter.

Put it all in one text file, marking whose is whose. The more concrete the raw material, the less Claude invents on its own.

What prompt should you give Claude?

The prompt (the text request you send to the AI) should set a role, the data, and the shape of the answer. Copy it and plug in your own:

"You are a market analyst. I'm an entrepreneur, my business is [short description of your business and product]. Below are my notes on competitors, each block tagged with a name. Fold this into a comparison table with rows: price, main promise, strength, weakness from reviews, sales channels. Make my business the first column. Where there is no data, write 'no data', don't make it up. After the table give three conclusions: where I'm stronger and it's worth putting into ads, where I'm weaker and customers leave because of it, what I can test within a week without big spending. Notes: [paste the file]".

The key here is the ban on making things up. Without it the model fills empty cells with plausible nonsense, and you end up deciding based on a fact that doesn't exist.

How to turn the answer into a working table?

Claude's first answer is almost always rough, and that's fine. Don't rewrite it by hand, refine it through dialogue:

• "The weaknesses are too vague. For each competitor give a concrete quote from a review that your conclusion is based on". That cuts off the fantasy: no quote, no conclusion.
• "Add a row: what does a customer overpay for at the competitor, and do I deliver it?". Your advantage often hides right there.
• "Assemble the final table in a format that opens in Google Sheets or Excel". Claude gives you a ready file or text that pastes into a spreadsheet without manual fuss.

This is easier to run inside one Claude project (a folder that holds the chat and files on a single topic): you drop the notes there, keep the table there, and come back to it a month later when competitors change their prices. Go to claude.ai, create a "Competitors" project, and the whole history stays at hand.

What to do with the finished table?

The table isn't the goal, the decision is. Take the three conclusions Claude gave and turn each into one concrete action: rewrite the site headline around a strength, close the weak spot that annoys customers, test one hypothesis within a week. Don't grab ten items at once, pick one and see it through.

A month later you open the same project, update prices and reviews, ask Claude to rebuild the table and show what changed. A one-off analysis goes stale within a quarter, while a living table becomes your radar on the market.

Common mistakes

• Raw material from memory, not from sites. "I think they're pricier" is not data. Copy real pricing, or the analysis falls apart.
• Trusting empty cells. If there's little data on a competitor, it should say "no data". Don't let the model paper over a gap with a guess.
• Copying others' decisions. The goal of the analysis is to find your own position, not to become a clone of the leader. Your advantage is often exactly where you don't look like everyone else.

Want to work through this on your own business with prompts at every step, start with our free materials: guides and the first free lesson about your digital twin.

AGINE Academy is an independent product, not affiliated with Anthropic. Claude belongs to Anthropic.