The short answer: a hand-checked qpAdm analysis from Ancestrify costs 29.99, 39.99, 49.99 or 59.99 EUR, paid once, with two optional 10 EUR add-ons and a 15 EUR re-analysis if a better model is ever found for your report. A do-it-yourself subscription tool costs a few dollars a month instead, and buys a different thing. The rest of this post explains what each figure buys, what does not move the price, and how to decide which of the two products is the right spend.
The four tiers#
qpAdm is sold at four tiers. Every tier gets the same report, the same maps and videos, the same downloadable model record, and the same publish bar: for every source in every era, p > 0.05, |Z| > 3 and SE < 0.10. Nothing about the standard changes with the price.
| Tier | Price | In the customer's words |
|---|---|---|
| Base, "The focused search" | 29.99 EUR | Your strongest straightforward model, found, checked, published. |
| Medium, "The search past good enough" | 39.99 EUR | We keep searching past the first model that works, until a clearly better one stops turning up. |
| Deep, "The full sweep" | 49.99 EUR | Every plausible version of your ancestry tried, the model that survives being pushed. |
| Perfect, "The exhaustive search" | 59.99 EUR | The search ends only when nothing beats it, the strongest model your DNA can give. |
Those lines are the tier cards' own copy. They were written when the tiers bought different search budgets: roughly 15 to 25 hand-built models at Base, 50 to 70 at Medium, 80 to 120 at Deep and 120 to 200 at Perfect, with analyst time from a few hours to a week or more.
What has changed, and what I want to be plain about: since 2026-08-29 every order, whatever tier was bought, receives the same exhaustive search. Every proxy label of every candidate source is tried, several hundred hand-built runs in parallel, and the model published is the one nothing else beat. That is now how I work on every file, because a smaller search on a Base order produces a defensible model less often than I was comfortable with, and the cheaper way to fix that was to stop doing smaller searches.
So what do the tiers buy now? Presentation depth. A Base report publishes the winning model with its full record and a written explanation. The higher tiers carry more of the search into the report: more of the alternatives that were tried and why they lost, more of the reasoning about proxy choice, more written explanation of what each source does and does not mean for your genome. The numbers are identical because the search is identical. If you want the model and the record, Base is the honest buy. If you want to read how it was reached, the higher tiers are where that goes. Medium is the tier the order form pre-selects; it is not the one you must pick.
The add-ons#
Model Lab, 10 EUR, one unlock. After the report is published, this puts your own merged
sample in the workbench: you choose sources and outgroups, qpadm() from ADMIXTOOLS 2 runs with
your genome as the target, up to 100 runs per rolling 24 hours. The same unlock includes the
download of the exact EIGENSTRAT bundle the report was computed from, for reproducing everything
on your own machine. One price, never two. Page: /qpadm/model-lab; walkthrough:
Run your own qpAdm models.
Whole-genome VCF upload, 10 EUR. If you sequenced your genome (Nebula, Dante Labs, tellmeGen
and similar) you have a .vcf or .vcf.gz rather than a chip export. The add-on converts it to
the reference panel's markers at upload, lifts GRCh38 to GRCh37 where needed, and does not store
the VCF. Details: Upload a whole-genome VCF.
Refined re-analysis, 15 EUR. If I revisit a published report and find a model that beats the one on the table, it is offered as a second version. The original stays exactly as published and you switch between them in the report. You are never charged for this without choosing it, and most reports never receive one.
Two further 10 EUR options at checkout, the paternal (Y-DNA) and maternal (mtDNA) haplogroups, are separate products bundled onto the same order rather than parts of the qpAdm analysis.
What does not change the price#
Your file's coverage. A 23andMe v5 export and an older AncestryDNA v1 export cost the same to analyse, though they will not support the same standard errors. Coverage bounds what a report can say, not what it costs. Check it for free before paying at the Raw DNA File Check; if the file cannot support the bar, the check says so and you have spent nothing.
Your vendor. 23andMe, AncestryDNA, MyHeritage, FamilyTreeDNA, LivingDNA and any export in the usual microarray text layout are accepted at the same price. Files are validated by content, not by brand.
Your region. The analysis covers every sovereign state, 199 countries and 1,115 regions, at the same price. A genome from a region with sparse ancient sampling is not cheaper to model; it is harder, and that is absorbed.
How many eras. The report covers both eras (Hunter-Gatherer and Neolithic Farmer; Classical Antiquity) at every tier.
What a DIY subscription costs instead#
The alternative to buying an analysis is renting the environment to do it yourself. Insights on Ancestry, as stated on their site, charges 5.99 USD or 10.99 USD per month, each plan with a compute balance. Illustrative DNA's AdmixLab is a subscription whose price is not stated on their site. Genoplot's tools are free. The full directory is Where to buy a qpAdm analysis.
The comparison is not "cheap versus expensive". A subscription is cheaper if you run models every week and already know how to choose a right set, read a nested-model table, and recognise a passing model that is wrong. It is more expensive, in both money and hours, if what you want is one defensible answer: a month or two of subscription plus the evenings spent learning the method adds up to roughly the price of a Base analysis, and at the end nobody has checked your model. The free AdmixTools 2 Lab runs real ADMIXTOOLS 2 over the public panel and is the right place to find out which kind of user you are before spending on either.
Why a person is in the loop#
The price of an analysis is analyst time, so it is fair to ask why a person is needed at all. Automated rotation, trying source and outgroup combinations until one passes, is the obvious way to sell qpAdm at a few euros. I built it, measured how often it published a passing model that was wrong, and removed it. Harney and colleagues (2021) documented the same problem in simulation: run enough models and one clears any threshold by chance, and the best-scoring model is frequently not the right one. The person in the loop is what stops a confident wrong answer reaching you with a p-value attached. Every model I publish is composed, run and audited by me, with my name on it, and the full record is published so anyone can check it. That is what the one-time price pays for, and it is why there is no cheaper automated tier.
Refunds#
An analysis is refundable before it runs and not after the report has been delivered, except where the law requires otherwise. The full policy is at /refund, and the free file check exists so that the main reason for wanting a refund, a file that cannot support the bar, is caught before any money changes hands.
Deciding#
If you want one answer with a stated bar and a name on it, start at Base, 29.99 EUR, at /buy-qpadm-analysis; the numbers are the same at every tier. If you want to read how the answer was reached, choose a deeper tier. If you want to run models yourself, add the Model Lab afterwards for 10 EUR. If you would rather learn first, the free Lab and the demo cost nothing. The full product page is /qpadm, and every price on this page is also on /pricing.
