All studies
DNA·November 8, 2026·5 min read

Do I Need a DNA Test With a Big Database If I Only Want My Ethnicity and Not Relatives

Relative matching needs a huge database. An ethnicity estimate does not. Here is how to tell which job you want and which kind of test fits it.

No. A DNA test does two separate jobs, and only one of them, relative matching, gets better as a company's database grows. Your ethnicity estimate depends on the reference populations you are compared against and the markers the chip reads, not on how many customers the company has. If you also want relatives, database size does matter, and that changes the answer.

Ethnicity depends on the panel, not the crowd

An ethnicity estimate compares you to a reference panel. A chip reads several hundred thousand positions in your genome where people commonly differ by a single letter. The company holds a panel of people with known origins, sorted into regions and populations. Software asks which panel groups your pattern of letters resembles, and in what proportions along your chromosomes. The output is percentages.

The core method compares you to that panel, not to every customer. What matters most is how many reference samples exist per population, how well they represent the place they stand for, and how many markers were read.

Two companies often give the same person different percentages because they use different panels and statistical choices, not because they disagree about your DNA.

Relative matching is a numbers game

Two people are flagged as relatives when they share long unbroken stretches of DNA, measured in centimorgans (cM). The closer the relationship, the more they share, and the share falls fast.

Average DNA shared with a cousin, by relationship
1st cousin866cM
2nd cousin229cM
3rd cousin (range 0 to 234)73cM
4th cousin (range 0 to 139)35cM

The 1st-cousin bar dwarfs the others: distant cousins share very little DNA. Averages from the Shared cM Project; individual results vary widely. Reported ranges run from roughly 0 to 234 cM for 3rd cousins and 0 to 139 cM for 4th cousins.

A first cousin matches you on a large amount of DNA and is easy to spot. By third and fourth cousins the shared amount is small and varies widely between people, so the company needs enormous numbers of customers before any of your distant cousins turns up. And the databases generally do not talk to each other. A cousin who tested somewhere else is usually invisible to you.

So for this job the question is simply: how many of my relatives are likely to be in this company's database? AncestryDNA and 23andMe are better than HealthOSX for finding relatives, because their databases are generally the largest. Database sizes change, so check current figures before you buy.

Database size decides who you can find. It does not decide what percentage of you comes from where.

Where size still touches ethnicity, a little

There is one indirect link. Company-built reference groups, drawn from customers whose families stayed in one place for generations, can help sharpen regional results, but the size of the benefit is uncertain. A bigger database is not a guarantee of a better answer.

The real limits of an ethnicity estimate are different:

  • Neighbouring regions blur. Separating, say, two adjacent countries is much harder than separating continents.
  • Recent migration muddies things, because panels describe where people lived, not where your great-grandmother was born.
  • Percentages are estimates with uncertainty. A 3% trace may be noise.
  • No autosomal test can name a tribe or settle a lineage. Maternal and paternal line results follow a single line each, deep but narrow.
Two jobs, two things to check
Relative matchingAncestry estimate
What decides the resultNumber of people in the same databaseReference populations and markers read
Check before buyingWhether your relatives are likely to have tested with that companyNumber of regions and populations, ancient groups, raw file download
Bigger database helpsYes, directlyOnly indirectly
Across companiesGenerally no, matches stay inside each company's databaseEstimates differ because panels differ

Scroll the table sideways for the rest →

What to look at before buying, depending on what you want from the test.

When the simple answer does not hold

Relatives become the point in these cases:

  • You are adopted, or a parent is unknown. Buy where the biggest pool is.
  • You want to confirm a family story, such as a half-sibling or a rumoured cousin. That needs the other person in the same database.

If you might want relatives later, buy where the biggest pool is now.

What to check on an ethnicity-only test

If the goal is ancestry and nothing else, compare tests on these:

  • How many markers the chip reads. HealthOSX reads about 650,000 on an Illumina GSA chip. Ask the same of any other test.
  • How many regions and reference populations the estimate uses. HealthOSX uses 15 regions and 157 reference populations.
  • Whether it compares you to ancient groups, not only present-day ones.
  • Whether maternal and paternal lines are reported.
  • Whether you can download your raw file.
  • Where the sample is analysed, what happens to it afterwards, and whether there is a subscription.

Expect estimates to differ between companies, even for the same person. Use these numbers to choose a test, not to judge which percentages are right.

Next step

Pick the job first, then the test. If you want relatives, go where the crowd is. If you want ethnicity, ignore the customer count and look at the panel and the markers. If you already have a raw file, try the free reader at healthosx.com/dna first, before paying for anything.

If you have no file, the HealthOSX DNA Kit is the second step. It reads about 650,000 markers from a saliva sample and reports ancestry across 15 regions and 157 reference populations, plus the closest of 298 present-day and 133 ancient groups, maternal and paternal lines, a Neanderthal variant count and traits. The raw file is yours to download. The kit is not built around finding relatives, so if relatives are your goal, use a company with a big matching database. There is no health report, so it is an ancestry and traits test, not a medical one. On the checklist's data question: the sample is analysed in Germany, results are stored encrypted in Germany until you delete them, and the lab destroys the sample after analysis.

The kit ships to Germany, the rest of the EU except France, and the US except New York and Maryland. US readers can find it at healthosx.com/dna/test/us and German readers at healthosx.com/dna/test/de.

New to raw DNA files? Start with the full guide: 23andMe Raw Data Interpretation Free in Your Browser

Written by

HealthOSX Research

Keep reading