about

we sell judgment, not rows.

the amazon dataset is a commodity. you can scrape it yourself with playwright on a tuesday; you can buy clean rows from half a dozen vendors. that's not the product. the product is the answer to the question every would-be seller pays jungle scout for: what should I sell, and is it going to make money.

so seller intel is six tools, all reading from the same shared time-series we accumulate one sweep at a time. niche.finder asks "what's the demand-vs-competition gap." demand.track turns BSR into a units-sold band with the confidence interval the model honestly deserves. review.mine reads the category leader's complaints back to you as the redesign brief. launch.guard watches your rivals on a six-hour cadence. arbitrage.map finds the cross-border gap. listing.doctor benchmarks your own listing against the top performers.

the studio runs the scraper that feeds all of this — amazon-scout.0p.studio — and seller intel is its first reference customer, consuming through the public MCP server the same way an outside developer would. a single snapshot of amazon is a commodity; the accumulated time-series is the product. that's the part nobody else has the patience to build.

what we won't do

• launder a model into a promise. every estimate ships with its confidence band. "estimated 800–1,400 units/month, medium confidence" is correct; "guaranteed 1,200" is wrong.
• ship a dashboard you have to study. you buy a decision, ranked, with a one-line thesis each. if you wanted rows, you'd buy amazon scout.
• add a free plan. scrape cost is real. so is what the model does on top of it.
• imply we're affiliated with amazon. amazon is the data source. we don't carry their logo and we're not their partner.

the rank was always public; the sales it implied were the part nobody had the history to read.