Keepa vs Helium 10 data: which number to trust
Keepa records what a listing actually did; Helium 10 models what it probably sells. Trust Keepa for history and offer counts, Helium 10 for current demand.
Last reviewed 5 min readSupplyForge LLC
Keepa and Helium 10 answer different questions: Keepa reports what a listing actually did, and Helium 10 estimates what it probably sells now, so neither number replaces the other. Keepa is an observation log. Helium 10's unit and revenue figures are model output, because Amazon does not publish per-ASIN unit sales to third parties.
Trust the observation for historical questions — did this listing ever move, and who has held the buy box. Trust the estimate for current demand on a live listing you do not sell yet.
What each source is actually measuring
Keepa measures observed listing state over time, and Helium 10 produces a modeled estimate of present-day sales. Verified vendor research compiled 2026-08-03 describes Keepa as an Amazon price, sales-rank and offer-history database with a token-metered API, and records that Keepa does not require an Amazon seller account. That matters for wholesale: Keepa can describe a brand's whole catalog before you have any relationship with the supplier.
The same research describes Helium 10 as an all-in-one Amazon seller suite covering product research, keyword research, listing optimization, PPC and profit analytics, aimed primarily at private-label brand owners, with Amazon-side discovery tools including Black Box and Xray. An observation can be stale, but it is never wrong about the past. An estimate can be current and still carry error that nobody publishes per ASIN.
Where a rank-based estimate loses accuracy
A modeled sales estimate is least reliable exactly where wholesale sellers spend most of their time: slow, deep-catalog ASINs from a distributor list. A rank-based model converts sales rank into units through a category curve, and near the head of a category that curve is steep and densely observed. Down in the tail one sale moves a rank thousands of places, so the same rank maps to very different unit counts week to week — a precise-looking figure that is really a range.
A rank-based estimate also cannot see how units split across sellers, so an ASIN doing 300 units a month across six sellers looks identical to one doing 300 alone. Forge Command's sourcing engine applies a hard floor of $1,000 per month of ASIN revenue and 10% net ROI before anything reaches a shortlist, which keeps the coarsest end of the estimate range out of buying decisions.
Why seller counts differ between Keepa, Helium 10 and SP-API
Three tools report three different seller counts on the same ASIN because each counts a different population over a different window, and all three can be correct at once. The common mistake is treating a listing-level number as a competition-level number: how many offers exist and how many sellers take revenue are separate facts, and only the second predicts your share.
| Source | Population counted | Window | Reads low when | Reads high when |
|---|---|---|---|---|
| Amazon SP-API offer count | Offers live on the listing at the moment of the call | Right now | Competitors are out of stock | Sellers sit above the box and never win it |
| Helium 10 seller count | Active sellers in your last CSV export | The export date | The export predates a new entrant | Sellers exited since the export |
| Keepa buy-box winners | Distinct sellers that won the buy box | Trailing 30 days | Listed sellers never win the box | Sellers rotate the box during stockouts |
Worked example, illustrative rather than measured: an ASIN shows nine live offers through SP-API, six sellers in a June Helium 10 export, and two distinct buy-box winners over Keepa's trailing 30 days. Nothing is broken. Nine includes sellers priced well above the box, six is June's population, and two is the count that took revenue. For thresholds, see how many sellers on an ASIN is too many.
Dead ASINs: the case where history beats the estimate
On a dead ASIN, history is the only useful data, because a current-demand estimate correctly reports close to zero and tells you nothing you can act on. That is how real opportunities fall out of shortlists: a listing that sold hard for two years and stopped six months ago reads as dead weight to a live-demand filter, while its review count, listing age and rank history say demand was real and supply stopped.
Forge Command defines the revival case numerically — at least 200 units in the last year and under 5 a month now — and surfaces those ASINs as one of its four opportunity types rather than filtering them out. Most research tools discard these listings by default, and the listing equity is already built. See finding and reviving dead Amazon ASINs.
Where Helium 10's number is the one to use
Helium 10's estimate is the better number for any live ASIN you do not sell yet, which is most of a supplier catalog on the first pass. A modeled monthly revenue figure gives one comparable value across ten thousand rows from ten brands, and history alone will not do that. Verified research checked 2026-08-03 also credits Helium 10 with the deepest keyword dataset in its category through Cerebro, Magnet and Keyword Tracker, plus PPC automation, profit and inventory modules, and an MCP server exposing 30-plus tools.
As of 2026-08-03, helium10.com/pricing lists Platinum at $129 per month or $99 billed yearly, Diamond at $359 per month or $279 billed yearly, and Enterprise from $1,499 per month billed annually. Watch the lifetime caps rather than the headline price — Platinum is capped at 20 tracked ASINs for the life of the account. Forge Command is not a product-research tool, does not replace Helium 10, and consumes Helium 10 data by CSV upload.
How Forge Command keeps three seller counts side by side
Forge Command stores the SP-API live offer count, the Helium 10 snapshot and Keepa's 30-day distinct buy-box winners in three separate fields and never averages them into one number. Merging them would erase the observation window, and the window is the entire meaning of the reading. Two detection patterns depend on that separation: the FBM-opportunity pattern needs a live count of 3 or fewer with an FBM buy-box winner, and the brand-as-only-FBA-seller no-go needs a live count of one alongside FBA fulfillment.
SP-API enrichment runs in batches of 20 identifiers per request against each tenant's own credentials, and Keepa lookups are opt-in and spend the tenant's own paid tokens, so a large catalog does not quietly drain a Keepa plan. See the FBM buy box entry pattern and why bring-your-own API keys matter in seller software.
Frequently asked questions
Does Keepa require an Amazon seller account
No. Verified vendor research compiled 2026-08-03 records that Keepa does not require an Amazon seller account, which is why Keepa can describe an ASIN or a whole brand catalog you have no relationship with. Helium 10's profit and inventory modules run off a connected Amazon seller account.
Why does Forge Command not merge the three seller counts
Forge Command keeps the SP-API live offer count, the Helium 10 snapshot and Keepa's 30-day distinct buy-box winners in three fields and never blends them. A blended figure hides which window produced it, and a seller count without its window cannot support a competition judgment.
What data replaces both tools on SKUs I already sell
Your own Amazon reports. For a SKU you already stock, unit-level order history and buy box percentage from Seller Central beat any third-party estimate, and both are free — see the Amazon reports every wholesale seller should run.