In 2013, as part of an internal reproducibility study, one company sent the same person two test kits with slightly different DNA samples. The ethnicity composition results differed somewhat. Not dramatically, but enough to raise the question: what exactly does this test measure — and how literally should the results be read?
DNA ethnicity analysis is one of the most popular consumer genetics products. Beautiful interactive maps, percentages to discuss over dinner. But behind the interface sits a statistical model with real limitations that advertising materials rarely mention.
Your DNA contains millions of single nucleotide polymorphisms (SNPs) — positions in the genome where people differ from each other by a single letter. Ethnicity analysis uses the fact that the frequency of certain SNPs has historically varied between geographic populations. If your set of SNPs resembles the set found in people whose ancestors lived in Norway, the algorithm assigns you a 'Scandinavian' component.
Technically this works through comparison of your genome against reference panels — sets of DNA from people in different regions whose ancestry is well documented. The larger and more representative the reference panel, the more accurate the estimate. The smaller and less representative, the greater the margin of error.
It does not mean that 47% of your ancestors were Scandinavian. It means: 47% of your genome shows SNP patterns that, in current reference panels, are associated with people from Scandinavia. The difference is fundamental.
First, 'Scandinavian' is not an ethnic or cultural category — it is a geographic clustering of contemporary genomic data. In the Viking Age, the Bronze Age or the Neolithic, these patterns would have looked different. The population of the region changed — migrated, mixed, was displaced. The DNA of a modern Norwegian carries traces of all these waves, including steppe pastoralists, Anatolian farmers and hunter-gatherers.
Second, estimates are unstable. The same person will receive somewhat different results from different companies — because they have different reference panels and different algorithms. And the same company can change your results without a new test — simply by updating its database. This is not an error. It is a normal model update as data grows.
The quality of ethnicity analysis depends directly on the quality of the reference panel. And reference panels are not a neutral cross-section of humanity. They substantially over-represent Western Europe and East Asia relative to Central Asia, the Middle East, sub-Saharan Africa, indigenous peoples of the Americas or Pacific islands.
This has direct consequences. A person of Ashkenazi Jewish descent may receive a significant 'Eastern European' or 'Middle Eastern' component — depending on how a specific company has classified the Ashkenazi population in its panel. A person of mixed African and European ancestry may see the same DNA patterns classified into different categories by different companies.
For populations well-represented in the panels (Western European, East Asian), estimates are reasonably stable. For everyone else — with a substantial margin of uncertainty.
Specific names of ancestors — DNA does not know who your great-great-grandfather was. The archive does. Exact countries and towns — '47% Scandinavian' does not mean 'from Bergen rather than Stockholm'. Cultural identity — DNA describes population ancestry, not culture, language, religion or nationality. Membership of a specific historical lineage — being a 'descendant of Vikings' or 'descendant of knights' cannot be established through ethnicity analysis. That is marketing, not science.
A special case: indigenous peoples. For many indigenous communities, DNA tests produce deeply problematic results: their genomes are either poorly represented in reference panels (and their ancestry gets 'spread' across adjacent categories), or colonial admixture introduces unpredictable complexity. Questions of tribal or national belonging are legal and cultural matters, not biological ones.
Broad biogeographical patterns — from which parts of the world your ancestors broadly came — the test shows with reasonable reliability, especially for families with an undivided regional history. Discovery of unexpected components — if someone only knew their ancestry from one side, the test may point to origins from the other. This is genuinely valuable.
DNA matches with relatives — perhaps the most practically useful function: a list of people whose DNA is similar enough to yours to suggest kinship. This, rather than the ethnicity percentages, often proves most useful for genealogical research. Deep evolutionary ancestry — haplogroups (Y-chromosome and mitochondrial) tell of migratory routes thousands of years old. This is not genealogy in the conventional sense, but an interesting history of your DNA lineage.
DNA ancestry tests sometimes reveal what a family did not know — or did not want to know. NPE (non-paternity event) — a discrepancy between biological and social paternity — occurs in an estimated 1–3% of cases per generation. Mixed racial or ethnic ancestry that was not discussed in the family. Connections to historical events — displacements, diasporas, genocides — that were not part of the family narrative.
This is not a reason to avoid the test. It is a reason to take it consciously — understanding that the information is sometimes more complex than expected.
DNA ancestry analysis measures biological heritage — patterns of genetic material transmitted across generations. It does not measure or define cultural, national, ethnic or personal identity.
Identity is shaped by language, culture, history, community and personal experience. A person can be '100% Irish' by DNA test result and never have set foot in Ireland, speak no Irish and not consider themselves Irish at all. And the reverse. Genetics describes one dimension of human history — important and interesting, but far from the only one.
DNA ethnicity analysis is a fascinating and genuinely informative tool when its capabilities and limitations are properly understood. It shows broad biogeographical patterns well. It works poorly as a substitute for genealogical archival research. The percentages are statistical estimates, not biographical facts. And the best use of test results is as a starting point for further investigation — not as a final answer.