Help & docs — product research, explained straight

The field manual for doing product research with Kestrel — how market scans, scores and signals decide whether a product is worth selling. Everything on this page describes what the radar actually does — no marketing gloss. If something here doesn't match what you see in the app, tell us and we'll fix one of the two.

Your first verdict — is this product worth selling?

  1. Sign in with a magic link. Free accounts start with 200 units — enough for 20 free market scans, no card.
  2. Open Discover and type a product you'd actually sell — "dog paw cleaner" works as well as anything.
  3. Hit Run scan (10 units). About forty seconds later you have a 0–100 score and a verdict — Hot, Promising, or Weak — with the evidence behind every point: ads running in Meta's ad library, Google search demand, and a live Amazon retail check.
  4. Hot or Promising? Track the keyword and the radar watches it every morning for free. Weak? You just saved yourself weeks — scan the next idea.

Units are only spent when you press scan — nothing ever runs against your balance on its own. The rest of this page explains every number you'll meet along the way.

Scan units — what costs what

One meter for everything. Watching niches, the daily sweeps, the signal feed, dossiers, and the morning edition are included in every plan and never consume units. Units only pay for research you trigger on demand:

ActionCostWhat you get
Market scan (Discover)10 unitsAd-landscape sweep + search economics + live Amazon retail check + 0–100 verdict
Niche blueprint (Generate)15 unitsKeyword basket with 12-month pattern analysis, entry point, buyer questions, category leaders
Blueprint, ad probe unavailable10 unitsSame research minus the ad market — the verdict is withheld rather than guessed, and you're charged less
Cached draft (research down)freeA starting draft built from the radar's own recent research, clearly labelled
Prospector dossierfreeThe radar researched it overnight on its own initiative

Every charge appears in Settings → Usage history with the exact reason. If a data source fails mid-research, the result says so in plain words — you are never billed for data that wasn't delivered.

The signal types

demand rising / demand fading
The keyword's last 3 months of search volume vs its prior 9-month average moved ±25% or more. Rising demand with modest competition is the radar's core "go look at this" signal.
keyword spike / keyword drop
A sudden jump far outside the keyword's normal range between collection runs — catches short-term surges the monthly curve smooths over.
competitor scaling / competitor ramp
A watched advertiser's active-ad count jumped ≥30% in a day, or rose across 3+ consecutive captures. Advertisers scale spend on what works.
creative push
A watched advertiser launched 3+ new creatives within a week — fresh creative volume usually precedes a scaling push.
community surge / social heat
A watched community is running at 2×+ its 7-day engagement baseline, or a single video/post is gaining unusual velocity.
prospect
The Prospector published a new dossier — a niche candidate that survived its overnight scoring. Free to read, one tap to track or pass.
retail hot / retail rising
Amazon retail proof. Retail hot: a tracked keyword is selling strongly right now (proven sellers, high "bought in past month", top Best Seller Ranks). Retail rising: a product's Best Seller Rank jumped ≥30% between sweeps — sales accelerating.

The opportunity score (0–100) — how the radar decides if a product will sell

Every market scan returns a transparent score with its full breakdown — you always see exactly where every point came from:

The five parts above add up to 92, not 100 — we publish the raw points rather than stretch them to a rounder number, so 92 is a perfect sheet. Bands: 70+ Hot · 40–69 Promising · below 40 Weak. Confidence is "high" only when search-volume data was available. If the ad-market probe fails during a scan, you get an error and no charge — never a fake verdict built on zeros.

Demand shapes & seasonal timing

Wherever the radar has 12 months of history it classifies the curve, not just the direction:

You'll see these as badges on blueprint keywords and as a seasonal chip + timing note on market scans.

Retail Intelligence — does it actually sell?

Search demand and competitor ads tell you people want a thing and someone advertises it. Amazon tells you people actually buy it. Every market scan now runs a live Amazon check alongside the ad sweep and returns a Retail check:

The reconciliation is the payoff: ad longevity proves someone profits from ads; Amazon sales prove people buy. When both agree, conviction is high. When ads are hot but Amazon is thin, you're looking at an emerging angle — or a fad on hype. When Amazon sells but ads are quiet, that's a possible opening. The Prospector attaches the same retail proof to every dossier, free, and a background sweep tracks BSR movement on your tracked keywords to raise retail signals.

Niche blueprints (the generator) — niche research on demand

Type an interest into Niches → Let the radar build it — or type nothing and let it pick. In about ten seconds it assembles:

When you generate without a seed, the idea comes from the radar's own overnight research or from a large idea catalog — always excluding niches you already have, seeds you recently generated, and prospects you passed on. The result tells you where the idea came from.

The Prospector — research you didn't ask for

Every morning around 04:40 UTC the Prospector expands demand maps around rotating categories, filters for novelty, probes the ad library for proof of spend, and scores each candidate. Only candidates scoring 45+ are published — most mornings that's a handful or none, and that's the point: silence means nothing passed, not that nothing ran.

Dossiers appear at the top of Discover. Track adds the keyword to your active niche and tells the Prospector you want more like it; Pass hides it forever and it will never suggest that idea again.

The morning edition & the dispatch

The app's Today view assembles an edition each morning after the sweeps land (all times UTC: community/video every 6 hours; search demand 06:00; competitor ads 08:00; analysis 07:00). It leads with what changed since your last visit, keeps your reading streak — with grace days, life happens — and counts down to the next sweep.

Prefer it in your inbox? Toggle Morning dispatch in Settings and a plain-text brief arrives at 08:30 UTC — only on mornings when there's actual news. No signals, no email.

Niches & watchlists

A niche bundles keywords, communities, video searches, and advertisers. Create as many as your plan allows and activate the one the radar should focus on — collectors re-aim automatically. Add items from a scan with one click ("Track keyword", "Watch"), from a blueprint, or by hand in the Niches tab. Removing an item stops future collection immediately; history is kept.

Reading the winners gallery

Scans rank the longest-running active ads first. Check the media badge (video vs image tells you the niche's creative meta), the variant count (an advertiser running 10+ variants of one concept is spending real money), and the store domain (a myshopify.com domain usually means a dropshipping competitor — beatable). Every ad links back to the public ad library so you can verify it yourself.

Account & sign-in

Sign in with a magic link — no passwords. Your workspace, niches, and signals are private to your account. Workspace tokens (Settings) are an alternative unlock for shared/ops use; rotate them any time.

Link not arriving? Check spam first — first emails from a new sender often land there. The link is single-use and expires; if it's been a while, just request a fresh one. Still stuck? Email us and we'll sort it.

Data & privacy

Everything on the dashboard comes from public market surfaces: ad transparency libraries, search-demand databases, and public community activity. We don't scrape private data, we don't estimate revenue we can't observe, and we never ask you to connect your own accounts or keys — the radar runs its own intelligence network. Numbers you can't verify are numbers we don't show.

Contact

Questions, upgrade requests, or something in the docs that doesn't match the app: write to us — read daily, answered by a person.

Upgrading? Pick a plan in the app and it opens a prefilled email — we reply with a secure payment link and your plan is live within a day. Card checkout is still being wired up, so it's a human on the other end for now.