built on the amazon dataset · sells the decision, not the rows

what to sell.
will it make money.
who is the soft incumbent.

the studio runs scheduled sweeps of the amazon catalog so you don't. you get a ranked list of niches with a one-line thesis each, an honest sales estimate (with the confidence band the model deserves), and an alert the morning your rival drops price.

  • · BSR-to-units bands with a real confidence interval — never a fake promise
  • · the time-series we accumulate is the moat — your one-off renter cannot replicate it
  • · no raw rows, no spreadsheets to sift — you buy a decision
  • · REST + MCP for outside agents · sub-200ms calls when cached
w1w2w3w4w5w6BSR ↓ / reviews ↑B0X1B0L2B0Y3B0Z9#1 · soft incumbent · 4.1 stars · 600 reviews · BSR ↑ 40% in 8w · beatable.
★ category sweep · live

six products · one feed · shared time-series

built on the dataset. priced for the decision.

every product reads from the same scheduled-sweep history. a single snapshot is a commodity — the accumulated time-series is the product.

niche.finder

$39/mo

the question every would-be seller pays jungle scout for: what should I sell. ranked sub-niches with a one-line thesis each.

  • ·category sweep on demand
  • ·demand vs competition scored
  • ·history seeded for free
start →

demand.track

$29/mo · 100 ASINs

BSR over time is the closest public proxy for sales volume. we maintain the weekly series and convert it into a units/month band with an honest confidence.

  • ·weekly time-series per ASIN
  • ·category-specific BSR→sales curves
  • ·low/med/high confidence on every estimate
start →

review.mine

$19/report

the category leader's reviews are the redesign brief. recurring complaints, 'I wish it had' patterns, exemplar quotes, severity-ranked.

  • ·top-N ASINs in any niche
  • ·complaints clustered by pattern
  • ·the build-a-better-version spec
start →

launch.guard

$49/mo · 25 ASINs

watch your ASINs and your rivals'. price drops, stockouts you can capitalize on, BSR surges, buy-box hijacks. every six hours by default.

  • ·before/after on every alert
  • ·email + webhook delivery
  • ·configurable thresholds per rule
start →

arbitrage.map

$99 once · or $29/mo

products that win on .com but barely exist on .de / .co.uk / .fr. the studio scrapes six marketplaces on one schema — this product just reads the gap.

  • ·any marketplace pair
  • ·absent / thin / present, ranked
  • ·rolling refresh available
start →

listing.doctor

$0.25/ASIN · min $9

your listing vs the category top performers, dimension by dimension. title, bullets, images, price position, rating. prioritized fix list.

  • ·category median benchmark
  • ·severity-ranked findings
  • ·concrete fix per finding
start →

how it works

the spine is the same for every product.

  1. 01

    give us a category URL or an ASIN list.

    you don't write SQL, you don't sift rows. you state the input and pay for the answer.

  2. 02

    the agent sweeps. the wrapper enforces.

    every Amazon Scout call is capped at the boundary — max products, max URLs, every call logged. a hallucinated 200,000-product sweep is impossible, not just discouraged.

  3. 03

    the model runs on accumulated history.

    BSR→sales, demand vs competition, review-gap synthesis — all computed against the shared time-series we've been collecting, never from a single snapshot.

  4. 04

    you get a decision.

    ranked opportunities with a one-line thesis. velocity bands with a confidence interval. a redesign spec with exemplar quotes. an alert with the before and the after.

REST · MCP

for the agents you're building.

every product surface is also a REST endpoint and an MCP tool. bearer-token auth, the same prices as the browser surface, deterministic guardrails at the wrapper boundary so your agent can't accidentally invoice you for a 200,000-product sweep.

// niche.finder — REST

POST /api/agents/niche-finder
authorization: Bearer si_xxx...
content-type: application/json

{
  "category_url": "https://www.amazon.com/Best-Sellers-Pet-Supplies/zgbs/pet-supplies",
  "marketplace": "com",
  "max_products": 120
}

→ 200 OK
{
  "runId": "0b2c...",
  "opportunities": [
    {
      "sub_niche": "Cat Trees",
      "demand_score": 71,
      "competition_score": 38,
      "thesis": "the category leader sits at 4.1 with 612 reviews; BSR climbing — beatable.",
      "top_asins": ["B0X1...","B0Y3..."]
    }
  ]
}

questions

honest answers.

isn't BSR-to-sales a guess?
it's a model, and we say so. every estimate carries a low/med/high confidence band derived from how many history points we have and how volatile the BSR has been in our window. we never launder a model into a promise. "estimated 800–1,400 units/month, medium confidence — BSR is volatile in this category" is correct. "guaranteed 1,200 sales/month" is wrong.
where does the data come from?
amazon-scout.0p.studio — a sibling product the studio runs. every seller-intel product consumes it through its public MCP server, the same way an outside developer would. amazon scout sells the rows; seller intel sells the judgment.
what's the moat?
the accumulated time-series. a single snapshot is a commodity anyone can buy. a maintained history of BSR, price, and review-velocity deltas across six marketplaces is something only a scheduled-scrape operation can assemble. we scrape the category once per cadence and amortize across every customer interested in that category.
do I have to learn an API?
no — the browser surface is the primary one. but every product is also a REST endpoint and an MCP tool, and the price is the same. agents are first-class customers.
can I cancel?
any time, from the billing page. failed payments don't delete data; per the studio's policy, the lights stay on.
is this affiliated with amazon?
no. amazon is the data source. there is no endorsement, no API key, no partnership, and we don't use any of amazon's marks.