A competitive intelligence framework for product teams

A framework starts with a decision someone has to make, because the decision sets which rivals and which pages matter. Write it as one sentence with an owner and a due date, then write the one question about rivals it depends on. Suppose a product team in work management is deciding whether its pricing page should say what its AI does with customer data, and on which plan. The question for the rivals is specific: where do two rivals that sell AI features show their AI controls to a buyer, and on which plan? Work without a decision behind it tends to collect readings nobody uses, and a question with no decision has no point at which the reading is finished. A decision with a due date also tells you when the reading is needed, which is the deadline that matters.

Each question has a source that can answer it, and choosing it first saves reading pages that cannot. What a rival says and offers to a visitor is on its public pages: the homepage, the pricing page and the comparison table under it, the documentation. Why buyers choose one product over another is not; it comes from customer interviews, sales calls and win-loss work. How much of a market a rival holds comes from market reports and data you may not have. For the AI-controls question, the sources are Notion’s features page and monday.com’s homepage, because those are the pages a buyer reads before talking to anyone. A question that needs two kinds of source is better split in two, so each half can be checked against its own.

Public pages change without notice, so a reading is only as good as the copy of each page it was made from and the date beside it. coExploro captured the Notion sample and the monday.com sample on 29 October 2026, both as a visitor without an account sees them, so the two can be read like for like. Keep the whole page, not only the part you expect to quote: the cards that answer this question sit in a section of their own, not in the hero. When two rivals are captured on different days, write both dates beside every sentence that sets them side by side. Without the page and its date, nobody can check the reading later, including you.

An observation states what a page shows; a reading states what you conclude from it, and the two belong in separate sentences. From the captures of 29 October 2026: Notion’s features page headed twelve cards with ‘Built for teams.’ and the line ‘Enterprise-grade security, admin controls, and agent governance.’, and one card read ‘No training on your data by default; ZDR on Enterprise.’ monday.com’s homepage listed six AI controls under ‘Full control’, among them ‘No data training’, which read ‘Your data is never used to train AI models’. Those are observations. ‘Notion keeps its strongest data promise for Enterprise customers’ is a reading. It may be right, but a card of one line is not a contract, and nothing behind a login was captured. Mark it as a reading and write down what would confirm it. coExploro’s methodology states what a capture can and cannot show.

Deliver the reading in the format the decision already uses, on one page: the question, the dates, the observations, your reading and what stays open. For this decision the result is short. On 29 October 2026 one rival listed its AI controls among security and admin cards on its features page, with a default on data training and a term for Enterprise, the other listed six of them on its homepage, among them a promise about data training, and neither capture showed the controls at work inside the product, which sits behind a login. Then set the date the reading goes stale, tied to the decision: if the pricing change ships next month, read both pages again before it does. On 29 October 2026, after the captures, the two observations below held. Competitive intelligence for product managers lists the questions a product team asks most often, and How to do a competitive teardown covers a single reading in detail.

• Observed by us

  Under the heading ‘Built for teams.’, Notion’s features page listed ‘No training on your data by default; ZDR on Enterprise.’ among its cards.

  verified Oct 29, 2026 · Source: notion.so/product/features (opens in a new tab)

• Observed by us

  monday.com’s homepage listed ‘No data training’ among the six items under ‘Full control’.

  verified Oct 29, 2026 · Source: monday.com (opens in a new tab)

coExploro does step 3 for you and drafts the observations of step 4: paste a rival’s address, and in 2–4 minutes it captures its main public pages with their date, and four AI analysts, PM, Marketing, UX and Designer, write findings from that capture. The decision, the choice of sources and the delivery stay with you, and so do the questions only customers can answer; the product teardown is what you bring to step 5. To read a rival again before the next decision, run a teardown of the same site again; each one carries its capture date. You can try one run a day per network without an account, a free account includes 3 teardowns a month, and paid plans start at $16 a month billed yearly.

• Twelve cards under 'Built for teams' cover security and admin · Notion

  Evidence: Enterprise-grade security, admin controls, and agent governance. · Seen on www.notion.so/product/features, features page, line under the 'Built for teams.' heading · Captured Oct 29, 2026 · Direct evidence

• The homepage lists six AI controls, among them no data training · monday.com

  Evidence: No data training · Seen on monday.com, 'Full control' block, second row, third item · Captured Oct 29, 2026 · Indirect evidence