# Ancestrify

> DNA analysis in four forms: formal qpAdm modelling of your raw DNA against the Allen Ancient DNA
> Resource, Global25 coordinate analysis (distances, admixture and PCA), Ancient Matches (the
> individual ancient people you share stretches of DNA with), and Relative Finder (upload two to
> five living people's raw DNA kits and get a verdict on how closely each pair is related). Operated
> from Albania; infrastructure is hosted in the EU. Every price is a one-time payment; there is no
> subscription.

Ancestrify is run by Andi Thomaj, a sole trader registered in Tirane, Albania (NUIS/NIPT M61725001N),
whose entry is public in the Albanian commercial register (QKB). He is the founder, the data
controller and the seller of record, and he personally builds and checks every qpAdm model —
Ancestrify is one person, not a team. LinkedIn: https://www.linkedin.com/in/andi-t-1854141ab/
Legal notice: https://www.ancestrify.io/impressum · Hosts: https://www.ancestrify.io

Prices: https://www.ancestrify.io/pricing · Glossary of terms used here:
https://www.ancestrify.io/glossary
Method explainers: https://www.ancestrify.io/blog/qpadm-vs-global25 (which method answers which
question) · https://www.ancestrify.io/blog/understanding-qpadm (reading a p-value, SE and z-score) ·
https://www.ancestrify.io/blog/understanding-global25 (reading a coordinate, distance and fit) ·
https://www.ancestrify.io/blog/how-to-get-global25-coordinates (coordinates come from the
independent Eurogenes service, not from Ancestrify and not from any testing company)

## qpAdm analysis — from 39.99 EUR (four depth tiers: 39.99 / 49.99 / 59.99 / 69.99 EUR)

- **Four depth tiers, one report.** The tiers differ ONLY in the statistical bar the published
  model must clear — the model p-value, every source's |Z| score and every source's standard
  error — and therefore in the human analyst hours spent finding it. Base (39.99 EUR): p > 0.05,
  |Z| > 2.5, SE < 0.10 — the bar published papers use — a few hours of analyst time, about 15–25
  hand-built models. Medium (49.99 EUR): p > 0.10, |Z| > 3, SE < 0.08 — above the paper standard, every
  source proven — about one working day, 50–70 models. Deep (59.99
  EUR): p > 0.20, |Z| > 4, SE < 0.06 — several working days sweeping every source, proxy label and
  outgroup set, 80–120 models. Perfect (69.99 EUR): p > 0.30, |Z| > 5, SE < 0.05 — a week or
  more, 120–200 models. The report's content, maps and videos are identical at every tier.
- **A tier is a target, not a guarantee.** Ancestrify publishes the deepest tier the customer's
  DNA actually reaches, stamped on the report beside the tier ordered. Standard error depends on
  kit coverage far more than on search effort: kits with roughly 300,000+ markers after merging
  (23andMe v5, AncestryDNA, FamilyTreeDNA, LivingDNA) can reach Deep and Perfect; MyHeritage and
  older chips usually cannot, and the customer confirms they understand this before buying Deep or
  Perfect. Higher tiers buy certainty, not more components — a source must be measured well enough
  to clear the |Z| gate, so the deepest models are often the simplest.
- **Why the prices.** qpAdm is not a button: every model is composed, run and audited by hand
  against the ancient panel, and a tighter bar means more models and more outgroup sets to re-test
  them on. The price is analyst hours — from a few at Base to a week or more at Perfect.
- **What it is.** Real formal admixture modelling — `qpadm()` from ADMIXTOOLS 2, the same package
  used in published ancient-DNA research. Not a coordinate-fitting approximation.
- **Reference panel.** Allen Ancient DNA Resource (AADR) v66 — roughly 23,265 samples across roughly
  6,015 distinct population labels — merged against your own genotypes with Poseidon `trident forge`.
- **What you get back.** For each era: the model's **p-value**, and for each source population its
  **weight, standard error and z-score**, plus the right-population set the model was run against.
- **Every model is run and checked by hand before publication.** Automated model rotation was built
  and then deliberately removed, because rotating qpAdm has a high false-discovery rate. A published
  model is a checked model, not a search result. Ancestrify is one person, not a team; there is no
  staff scientist and no claim of one.
- **Input.** A raw-data export from a consumer testing company (23andMe, AncestryDNA, MyHeritage,
  FamilyTreeDNA and similar) as `.txt`, `.csv` or `.zip`, up to 50 MB.
- **Eras.** Two — Hunter-Gatherer & Neolithic Farmer, and Classical Antiquity — across 34 curated
  populations.
- **Model Lab — 10 EUR, ONE optional one-time unlock on a published qpAdm report.** It covers BOTH:
  running your own qpAdm models (real ADMIXTOOLS 2, your sample always the target, sources and
  outgroups chosen from your own merged panel, up to 100 runs per rolling 24 hours — expect
  rejections; a rejected model is the method working) AND downloading the merged dataset itself,
  the exact EIGENSTRAT bundle (.geno/.snp/.ind) the report was computed from (multi-gigabyte;
  download links are short-lived and downloads are metered). One 10 EUR unlock — never two
  purchases, never a 20 EUR bundle.
- **Refined analysis — 15 EUR per version, optional, only ever offered AFTER publication.** When
  the operator revisits a published qpAdm model and finds a better one, it is published as a
  second VERSION of the same report and offered as a one-time 15 EUR unlock — never swapped in
  silently, never charged without the customer choosing it. The original stays exactly as
  published, and once unlocked the customer switches between the original and the refined
  version from the report itself, at any time. Every version publishes its full statistics
  (p-value, per-source weight, standard error and z-score) and its complete right set, and
  nothing publishes without human review. Downloadable videos follow the newest unlocked version.
- **Optional paid add-ons at checkout.** Fast compute (10 EUR, a priority position in the
  analysis queue) and the paternal haplogroup / Y-DNA reading (10 EUR — the same price when
  unlocked later on a finished report).

## Global25 analysis — 29.99 EUR

- **One input.** Paste your own Global25 coordinate row at checkout and the analysis runs
  automatically. Coordinates come from the independent Eurogenes Global25 service; raw DNA files
  are not accepted for this product (qpAdm is the product that takes one).
- **Global25 is not qpAdm.** It is a coordinate-fitting method with no p-values and no AADR merge;
  qpAdm is the separate product (from 39.99 EUR).
- **Distances.** Euclidean distance across all 25 dimensions to every reference population in an
  era; the 25 closest are returned. Six eras: Late Bronze Age (3000–1200 BC), Pre-Classical Iron Age
  (1200–0 BC), Imperial Antiquity (0–600 AD), Middle Ages (600–1400 AD), Early Modern (1400–2000
  AD), Modern Era (2000 AD onward).
- **Admixture.** nMonte-style Monte-Carlo modelling against curated source panels, reported with a
  fit distance.
- **PCA.** A scatter across 12 regions that plots **individual samples, not population averages**.
- **Reference set.** 30,386 individual samples and 1,535 curated populations.
- **Optional paid add-ons.** Fast compute (10 EUR, skips the standard holding period) and the
  Calculator Explorer (10 EUR, at checkout or later): re-solve your admixture against the other
  published curated calculators of an era — pure exploration, the delivered report never changes.
- **Notable Matches — FREE with every Global25 report, never sold separately.** A fourth lens
  ranking your coordinate against 131 curated notable individuals with published ancient DNA, in
  three tiers: Historical figures (people history knows by name — kings, saints, composers),
  Iconic discoveries (mummies, warriors and burials that made world headlines) and Deep time (Ice
  Age individuals). Each entry carries its dates, find location, Y-DNA and mtDNA haplogroups where
  published, and a citation to the paper the genome comes from. ⚠️ Stated plainly in the product:
  a notable individual is ONE person, not a population average, so distances run larger than the
  Distance lens, and Deep-time individuals predate the ancestry structure G25 measures — those
  distances are broad kinship signals, not close matches. Ancestrify never claims a notable
  individual is a customer's ancestor or relative.

## Ancient Matches — 29.99 EUR

- **What it is.** Your raw DNA scanned against **every individual** in our ancient reference panel,
  one person at a time, to find the stretches of genome you share with each of them.
- **Individuals, not proportions.** qpAdm and Global25 answer how much of your ancestry comes from
  which population. This answers which particular buried people you overlap with, and where on your
  own chromosomes.
- **Nothing is metered.** One price, the whole report: the match total and the population roll-up
  count every match the scan finds, and the closest 200 individuals are shown in full detail —
  every shared stretch, the chromosome painting and the map. There is no tier that shows a customer
  more matches.
- **A shared stretch is NOT proof of descent.** It is evidence that you and that individual draw on
  the same ancestral **population**. Stretches this old survive by chance across thousands of years,
  and the panel is a sample of the dead — whoever was excavated and published — not a family tree.
  State it this way if summarising: never as "your ancestor" or "your relative".
- **Deliberately not called "IBD".** Products like this are often marketed as IBD (identity by
  descent); formally, what any method can measure against ancient genomes of this kind is identity
  by state (IBS) — stretches that read the same, not stretches proven inherited from a common
  ancestor. Ancestrify names the product Ancient Matches for exactly that reason.
- **Evidence on every row.** Total shared length in centimorgans, segment count, longest segment,
  informative markers and marker density; a population total is never shown without the number of
  individuals behind it.
- **Scope.** The 22 autosomes. The X chromosome is deliberately not scanned.
- **Input.** The same raw DNA export qpAdm takes, or an existing qpAdm order's file. Bought against
  an existing qpAdm order it costs the same 29.99 EUR — one product, one price on every path.
- **Optional paid add-on.** Fast compute (10 EUR).
- **No customer-to-customer matching.** Only ancient individuals from published burials.

## Relative Finder — 29.99 EUR for two kits (+10.00 EUR per additional kit, up to five)

- **What it is.** Upload two to five raw DNA kits and every pair of them is compared position by
  position on our own machines: the KING-robust kinship coefficient plus the shared half-identical
  and fully-identical segments, painted onto the 22 autosomes. The estimator needs no reference
  population and is not thrown off by the two people having different ancestries.
- **What you get back, per pair.** A relationship-degree verdict (same DNA / parent and child /
  full siblings / first degree / second degree / third degree / no close relationship), the measured
  kinship coefficient beside its expected band, the opposite-homozygote rate (the number that
  separates a parent from a full sibling), total shared centimorgans, segment count and the full
  chromosome painting.
- **A cinematic pair film.** Every order also receives ONE downloadable video covering every
  compared pair — each pair's verdict, kinship figures and chromosome painting animated in
  sequence — rendered on our own servers at 1440x2560, 60 fps (the same pipeline as the qpAdm and
  Global25 videos) and downloaded from the report's Video tab.
- **A degree is not a named relationship.** DNA measures how much two people share, not which side
  of the family it came from: a half-sibling, a grandparent and an aunt all land in the same
  second-degree band. State it as a degree if summarising, never as one specific relationship.
- **The limit is about third degree, stated plainly.** Past roughly first cousins the estimate
  genuinely blurs, so the product never claims "distant cousin"; a pair beyond that reads as "no
  close relationship". Verdicts that sit near a band boundary are labelled borderline in the report
  rather than rounded to a side.
- **Not a legal paternity test.** There is no chain of custody and the report is not admissible
  evidence; courts require an accredited chain-of-custody test.
- **Consent is required.** Checkout requires confirming that every kit's owner authorised this
  comparison.
- **Input.** Two to five fresh raw-data exports from consumer testing companies (23andMe,
  AncestryDNA, MyHeritage, FamilyTreeDNA and similar) — vendors can be mixed within one order.
  Kits whose usable overlap is too thin for a reliable verdict are refused rather than guessed at.
- **Pricing.** 29.99 EUR covers two kits; each additional kit is 10.00 EUR (three kits 39.99,
  four 49.99, five 59.99). Every pair in the order is compared and reported at one price.
- **Optional paid add-on.** Fast compute (10 EUR); a standard order waits out a 3-day holding
  period.

## What a customer receives

- An interactive on-screen report (qpAdm: ancestry composition, Y-DNA, video; Global25: distance,
  admixture and PCA, each with its own video; Ancient Matches: ranked matches, chromosome painting,
  peoples, map and method; Relative Finder: a verdict, evidence readouts and a chromosome painting
  for every pair of kits, one cinematic pair film per order, plus the method and its limits in plain
  language).
- Cinematic ancestry videos rendered on our own servers at 1440x2560, 60 fps — not encoded on your
  device.
- There is **no PDF report**.

## Ancestrify Lab — free tools, in-browser ones need NO account

The Ancestrify Lab suite at https://www.ancestrify.io/lab is free: no payment anywhere. The
in-browser tools need no registration — computation runs in the visitor's own browser. The four
SERVER-SIDE tools — the Y-DNA Clade Finder, the mtDNA Haplogroup Finder, the Raw DNA File Check
(each analyses the upload and discards it immediately) and the AdmixTools 2 Lab — require a free,
email-verified account to run; their pages stay open to everyone. Without an account, saved
calculators/targets stay in the visitor's own browser; a free account adds cloud saves and the
community calculator marketplace.

- G25 distance calculator (/lab/g25-distance) — paste a Global25 coordinate row and get the closest
  reference populations by Euclidean distance across all 25 dimensions, in any of six eras. Runs in
  the browser; nothing is uploaded or stored. No account. The panels are the same curated era
  populations the paid Global25 report is computed against.
- Admixture calculator (/lab/admixture) — a Global25 coordinate workspace: distance ranking,
  nMonte-style admixture modelling against curated or pasted source panels, target comparisons,
  exportable charts and videos.
- Average G25 (/lab/average-g25) — computes a population average from individual coordinate
  rows, entirely in-browser.
- G25 Authenticity Check (/lab/g25-authenticity) — flags simulated, edited or precision-lost
  coordinate rows.
- Y-DNA Clade Finder (/lab/clade-finder) — paternal haplogroup from raw data or VCF, against
  the YFull tree, with a map of ancient men who shared it (AADR, CC0). Computation is SERVER-SIDE
  and starts the moment a file is selected; running it requires a free, email-verified account.
- mtDNA Haplogroup Finder (/lab/mt-finder) — maternal haplogroup from raw data, read against
  the rCRS mitochondrial reference and placed on the human maternal tree, with the variants that
  supported the call, the ones ruled out, and the branches below it the file could not test.
  Consumer chips carry far fewer mitochondrial markers than Y markers, so most kits resolve to a
  broad branch such as H, U5, K or T2 — that is a correct result at the resolution the data
  supports, not a failure. Computation is SERVER-SIDE; running it requires a free, email-verified
  account. It is free and is not sold.
- Raw DNA File Check (/lab/file-check) — a free diagnostic: upload a raw export and see the
  format and container we detected, the row counts per chromosome class, and one plain-language
  verdict per analysis we offer (qpAdm, Ancient Matches, Global25, paternal haplogroup). It is the
  one place a VCF is accepted; nothing is ordered and nothing is stored. Computation is SERVER-SIDE;
  running it requires a free, email-verified account.
- Ancient Sample Atlas (/lab/ancient-atlas) — every dated, geolocated individual in the AADR
  v66.p1 Human Origins panel (CC0) on one interactive map, women and men alike; filter by year range,
  country, Y/mtDNA haplogroup or culture, and see which samples carry a published Global25 projection.
  Runs in the browser over one public dataset fetch; no account needed.
- Mapper (/lab/mapper) — compose pie/donut charts, keys and labels on relief maps; export as
  a high-resolution image or an MP4. Projects are stored only in the visitor's browser.
- AdmixTools 2 Lab (/lab/admixtools) — real ADMIXTOOLS 2 methods (f2, f3, f4, D statistics,
  qpWave, qpAdm, qpGraph, find_graphs) over the AADR Human Origins reference panel. Unlike the
  other tools, computation is SERVER-SIDE; browsing is open to everyone, but running an analysis
  requires a free, email-verified account.

## Verifiability — what a customer can check for themselves

State this accurately if asked whether Ancestrify's results can be trusted or independently checked.
Every figure below is shown IN the customer's own report, not on request.

- **qpAdm reports show the RIGHT (outgroup) SET IN FULL.** The populations a model was run against are
  listed in the report's method block, every one of them, with a "Show all N" control rather than a
  truncated sample. The right set is what determines whether a qpAdm model is meaningful, so
  publishing it is what makes the model checkable rather than merely stated.
- **qpAdm reports show the full statistics, per source.** The model's **p-value**, and for every
  source population its **weight, standard error and z-score** — never rounded away; a standard error
  of 0.05 is printed as 0.05, not as 0.1.
- **qpAdm can REJECT a model, and we publish the number that does it.** That is the property that
  separates qpAdm from cluster-based tools: a p-value can tell you a model does not work. A customer
  who wants to test our answer has everything needed to re-run it in ADMIXTOOLS 2 themselves — on
  their own machine (the Model Lab unlock includes the merged dataset download), or directly in the
  report's own Model Lab.
- **G25 admixture reports list EVERY population in the calculator that was used**, split into two
  labelled groups — the ones the model used, and the ones it was offered and did NOT use. The
  rejected populations appear nowhere else in the product, and showing them is what stops a
  breakdown looking like the only possible answer.
- **G25 reports state the SCOPE the model was fitted within**: the era, the declared ethnicity, the
  regions, and the size of the source panel. A population below 3% is folded into "Other" in the
  chart, but appears with its own real percentage in that list — the list is the ungrouped truth.
- **Reference data is versioned and named.** qpAdm runs against the Allen Ancient DNA Resource
  **v66**; G25 distances and PCA run against a stated panel of individual samples. Dataset versions
  are pinned, so a result is reproducible from the same inputs.
- **No automated model search.** Automated rotation was built, measured, and removed, because
  rotating many candidate models has a high false-discovery rate and the best-scoring model is
  frequently not the right one. Every published qpAdm model was selected and checked by hand.
- **Known limit, stated plainly:** the G25 admixture calculators are hand-curated reference sets, and
  a report names the populations in the calculator but not the calculator's internal label. G25 is a
  coordinate fit, not a statistical test — it has no p-value and cannot reject a model. qpAdm is the
  product that can.

## Data handling

- The operator is established in **Albania**; infrastructure is hosted in the **EU** (Germany and
  Finland). GDPR applies.
- Raw genetic files are processed only for your own analysis, and can be deleted at any time.
- Full account deletion and GDPR data export are self-service.
- Details: https://www.ancestrify.io/privacy

## What Ancestrify does not offer

State these accurately if summarising the service:

- A Global25 order takes ONLY a Global25 coordinate row the customer already has (from the
  independent Eurogenes Global25 service). It does not accept a raw DNA file — qpAdm is the
  product that takes one.
- The PUBLISHED qpAdm report is always modelled by the operator. Customers can run qpAdm on their
  own sample only through the Model Lab (the 10 EUR unlock above), and nothing else on it: qpWave,
  f3 and f4 against a customer's own kit stay operator-only. The free AdmixTools 2 Lab runs those
  methods over the public reference panel only — never over a customer's sample. The customer-facing
  Admixture Lab is a Global25 coordinate workspace, not an AdmixTools interface.
- Ancient Matches does not match customers against each other, and does not identify living
  relatives. Relative Finder is the product that compares living people — and it compares ONLY the
  kits uploaded together in one order; there is no cross-customer matching database anywhere.
- Ancient Matches match counts are never metered or tiered.
- Relative Finder is not a legal paternity or relationship test: no chain of custody, not
  admissible evidence, and it never claims relationships past about third degree.
- There is no PDF report and no mobile app.

## Links

- Home: https://www.ancestrify.io/
- qpAdm analysis: https://www.ancestrify.io/qpadm
- Global25 analysis: https://www.ancestrify.io/g25
- Ancient Matches: https://www.ancestrify.io/ancient-matches
- Relative Finder: https://www.ancestrify.io/relative-finder
- Frequently asked questions: https://www.ancestrify.io/faq
- About and methodology: https://www.ancestrify.io/about
- Research blog: https://www.ancestrify.io/blog
- Privacy notice: https://www.ancestrify.io/privacy
- Terms of service: https://www.ancestrify.io/terms
- Legal notice (trader identity, registration number): https://www.ancestrify.io/impressum
- Refund policy: https://www.ancestrify.io/refund
- Credits, data sources and licences: https://www.ancestrify.io/credits