AncestryDNA dominates the genetic genealogy market in America, but until 2016 they only sold tests in the USA, the UK, Ireland, Australia, New Zealand, and Canada. Even today, their tests are only available in 35 countries. That’s the smallest geographic footprint of any of the major players. Thus, although the AncestryDNA test is generally the best place to start for someone with deep American roots, the same advice does not apply universally.
But which test is best outside the US? Turns out that’s a difficult question to answer. The companies don’t release country-by-country sales data. The best we can do at the moment is crowd-source information from genealogists around the world and try to make sense of it.
To that end, I created a survey to gauge how many matches one can expect based on grandparental country of origin. For each tester, the survey asked for the country of birth of all four grandparents, how many matches were at each testing company, and which company had been most valuable for that person’s genealogical research.
Problems with this Approach
Right off the bat, I admit that there is a major flaw in this approach: a large number of matches doesn’t necessarily translate to genealogically useful matches. That is, someone from Sweden might have more matches at AncestryDNA but closer ones at Family Tree DNA. A hundred closer matches will likely do more to advance one’s genealogical research than 1,000 or even 10,000 distant ones.
Another problem is that someone of mixed background (say, grandparents from Trinidad, Guadeloupe, and the US) will have skewed results. Most of their matches will be from the US, and it’s hard to tease out which company is best for their Caribbean relatives.
Third, those who took the survey are not an accurate representation of global test sales. For example, more than 10% of reported grandparents were Scandinavian, but surely Scandinavians don’t make up more than a tenth of all DNA testers around the world.
A final concern is that the “most valuable” question is entirely subjective. For example, one respondent said they refused to test with two of the companies, so their preference for a third reflects preconceived bias rather than which company might actually be best for their research.
When writing surveys, there are trade-offs between the information you really want and the information you think respondents are willing to give. In this case, I asked these particular questions to make the survey easy to answer. A better question than “How many matches do you have?” might have been “How many matches greater than 50 cM do you have after you’ve excluded segments shorter than 7 cM?”. However, given that more than 10% of respondents didn’t answer the simple “how many matches” question for any of the companies, it’s unlikely that any but the most hardcore genetic genealogists would have bothered to dig up the more precise numbers.
And the respondents are already biased: of the more than 10 million people who have taken a genealogical DNA test, only 281 had answered the survey as of 13 March 2018.
With those weaknesses disclosed, let’s play with the data.
First, I omitted from consideration any response that did not have numbers for at least two companies. After all, how can you have a valid “preference” if you’ve only used one company? That left 187 unique respondents.
There were also some individual data points that were not credible. One person reported 25,400 pages of matches at AncestryDNA. (There are 50 matches per page). I assumed they meant actual matches. Another reported a single page of matches at Ancestry, 288 matches at Family Tree DNA, and 2,392 matches at MyHeritage, yet said that AncestryDNA was the best company for them; I excluded the AncestryDNA number for this person as an error but kept the other information.
Next, I added filters for “one or more grandparent born in the US” and “one or more grandparent born in the Anglosphere”, which I defined as the USA, the UK, Ireland, Australia, New Zealand, and Canada (the early markets for AncestryDNA). I wanted to see whether either category made a difference in the number of matches or in the preference for one company over another.
Let the Numbers Speak
The grandparents for the 187 respondents were from the following countries:
First, let’s look at the numbers of matches at each site based on grandparent origins. No matter where the grandparents were from, every respondent who tested at AncestryDNA had more total matches there than elsewhere. That “advantage” was skewed toward people with at least one grandparent from the US and was even more skewed toward people with at least one grandparent from an Anglosphere country.
Conversely, if none of the grandparents were from an Anglosphere country, the advantage of testing at AncestryDNA was not as extreme. Testers with one or more grandparent from the Anglosphere had an average of 12.7 more matches at AncestryDNA than at Family Tree DNA, but for testers with no grandparents from the Anglosphere, the difference was only 4.8 fold.
Now, let’s look at company preference, or as was asked on the survey, “Which autosomal DNA test has been most valuable for your own family research?”
Regardless whether or not respondents had at least one grandparent from the US, they generally thought that AncestryDNA was more valuable. However, Family Tree DNA and, to a lesser extent, MyHeritage increased in importance to people with no American grandparents, and 23andMe decreased in relevance.
A really interesting pattern emerges when we consider whether any grandparents were from the Anglosphere.
AncestryDNA was preferred by 71% of respondents with one or more grandparent from the USA, the UK, Ireland, Australia, New Zealand, and Canada, but Family Tree DNA was the favorite of 67% of those with no grandparents from those countries. MyHeritage also saw a substantial gain outside of the Anglosphere, although the total numbers of respondents who preferred them was small (7 in each category).
As someone with roots in North America that predate the American Revolution by more than a century, I have a vantage point on DNA matching that is surely quite different than someone from, say, Malaysia. I tend to recommend AncestryDNA as the first step in genealogical testing, but I should do a better job of asking where that person’s ancestors are from before I weigh in. If they’re not from an Anglosphere country, they should probably test first with Family Tree DNA first then transfer to MyHeritage.
This exercise also gave me some ideas about survey questions that might better gauge match quality rather than match quantity. I hope to write an improved survey in the future. If you’d like to contribute to the current version, you can find it here.