Wednesday 24 October 2012

What next for

The big news in the family history world this week is the announcement that the genealogy website is to be acquired for $1.6 billion by the European private equity firm Permira. It remains to be seen what effect this will have on the company. Ancestry's third quarter accounts were released today, which show that the company continues to make a healthy profit and, as of 30 September 2012, had around 2,020,000 subscribers. Within the genetic genealogy community we are wondering what impact this acquisition will have on Ancestry's DNA testing programme. In the official press release Ancestry and Permira stated the following:
There are no anticipated changes in's operating structure.'s focus will continue to be on investing in content, technology and its user experience, expanding its product offerings in areas like DNA, and building the brand and the family history category, all on a global basis. will remain headquartered in Provo, Utah, with a continued large presence in San Francisco, Dublin, London and other international markets.
Ancestry started offering Y-chromosome DNA (Y-DNA) and mitochondrial DNA (mtDNA) tests in 2007 after acquiring the assets of the now defunct company Relative Genetics. These tests are no longer being actively marketed, and Ancestry have instead focused all their attention on their new autosomal DNA test which was launched in May 2012. It remains to be seen whether this test will be marketed outside the US. As I commented when I reviewed the test on this blog earlier this year, it is very much geared towards the American market with its emphasis on ethnicity percentages. The database is also 99.9% American and anyone who does not have Colonial American ancestry will gain little if any benefit from the test at present.

A significant drawback of the Ancestry autosomal test, in comparison with the equivalent offerings from Family Tree DNA and 23andMe, is that you are not able to access your raw genetic data and you are also not given any segment matching data. This makes it impossible to interpret your results in any meaningful way, and prevents you from uploading your results to third-party tools such as GedMatch for further analysis. I wrote to Ancestry about this problem in August. Other bloggers, including CeCe Moore and Judy Russell, have also been campaigning for Ancestry to make the raw genetic data available to customers. It is therefore gratifying that Ancestry have listened to the criticisms and have now announced that they will make the data available to customers some time in 2013.

However, the problems with Ancestry's admixture tools have yet to be resolved. My own documented ancestry, as far back as I can trace it, is all from the British Isles, yet according to the Ancestry test I am 58% Central European, 25% British Isles, 13% Eastern European and 4% unknown. I wrote about this aspect of the testing in more detail in a previous blog post. In contrast many of the Americans who are my genetic cousins have much higher percentages of  "British" ancestry, despite having a large proportion of their documented ancestry from other European countries. For example, the American blogger Roberta Estes, who has also commented on the deficiencies of Ancestry's admixture results, reports that according to Ancestry she is 80% British Isles, 12% Scandinavian and 8% unknown. However, Roberta's genealogy research shows that only 22% of her ancestry is from the British Isles, with her remaining known ancestry being from a variety of different European countries. There is, of course, a perfectly simple explanation for these discrepancies. The ethnicity predictions are based on a very limited number of publicly available datasets from reference populations, and the sequences that are in the public domain often have little in the way of accompanying ancestral information. For those people of British ancestry more detailed analyses should be possible once the People of the British Isles Project start to publish their results later this year. Other companies are more open about the limitations of the tests and publish details of the reference populations they use. For example, Family Tree DNA use a population from the Orkney Islands as a proxy for British DNA. Ancestry unfortunately have not released details of the reference populations they are using for their calculations. They have access to the same public datasets as the other companies. In addition, Ancestry purchased the assets of the Sorenson Molecular Genealogy Foundation (SMGF) in March 2012, and may have access to data from the SMGF database. However, they almost certainly do not have any data from any large British reference populations and it seems that the most plausible explanation for the bizarre results that we are seeing from Ancestry is that their "British" populations are probably Americans from Salt Lake City with mixed European ancestry rather than people living in the British Isles like me with many generations of documented British ancestry.

We are all aware of the limitations of these tests, and there will no doubt be many improvements in the years to come as more reference populations become available allowing the companies to provide more accurate results. Ancestry's autosomal test is in any case still in the beta-testing phase as is FTDNA's Population Finder tool. I am, however, concerned that Ancestry are compounding the problem by presenting the data in a very misleading and dishonest way. A link was posted on the ISOGG Project Administrators' mailing list this week to a new video that is now available on Ancestry's YouTube channel.

In this presentation Ancestry's blogger Crista Cowan provides a somewhat dumbed down overview of the Ancestry autosomal DNA test. However, what is of most concern is that nowhere in the video does Crista admit the limitations of Ancestry's admixture result. Instead she is trying to suggest that the reason why people are getting these unexpected percentages is because they haven't yet researched all their family lines. In other words it is not Ancestry's fault but the customers' for not doing their research properly! We are supposed to believe, for example, that there was some significant migration from Germany to Scotland which might account for some of these unexpected results. I was told by an Ancestry rep that there might have been a large migration from Eastern Europe to account for the 13% of my DNA that is supposedly from Eastern Europe. This is all of course total nonsense. It might well suit Ancestry's purpose to encourage people to search for their phantom ancestors throughout Europe to explain their mysterious DNA results. They are undoubtedly selling their test at $99 as a loss leader and are hoping to make up the deficit by increasing their subscription revenue. It is in their interests to encourage people to renew their subscriptions and to reduce the churn level. I just hope that most people have the good sense not to be taken in by this misleading hype.

For anyone who has had unexpected results from the Ancestry test I would advise that you provide the company with feedback and, more importantly, ask them to publish the information about the reference populations they are using. If enough pressure is brought to bear on Ancestry then perhaps they will make this information available so that the admixture results can be put into context.

© 2012 Debbie Kennett


Anonymous said...

You are right-on about the the DNA ancestry ETHNCITY results of the test. They are way off base and it is foolish for ancestry to deny it. It's almost crazy for them to deny it. HOWEVER, the "cousin matching" DNA part of the results to find "shared ancestors" in common with others who have taken the DNA test, I WOULD SAY THAT PART IS A GREAT SUCCESS. I am finding "shared ancestors" with "matching" cousins all over the place. Especially in the low and very low confidence part where you are looking for distant cousins. Yes, there are quite a few where I can't find a "shared" ancestor. But it is probably because I have not been able to find the "shared" ancestor, because of the family trees "missing" the names of those ancestors. So.. the ETHNICITY test results.. they should quit trying to "defend" them. And, the "cousin" matching part.. it's great. Maybe they should give up on the ethnicity part and just go with the "cousin" matching part until they get better information.

Debbie Kennett said...

The cousin-matching part of the test seems to work well for people in the US, though it would be better if Ancestry could provide the segment data and access to the raw genetic data so that matches can be confirmed. However, the test is only on sale in the US so it is essentially useless for me in the UK as I have no hope of finding a match with my distant US cousins, all of whom seem to have Colonial American ancestry going back to the 1600s.