The Glass Is More Than Half Full—2017 Version

“Ten thousand matches and only 10 of them have trees!”

“Why do people spend money on a DNA test if they’re not interested in genealogy?”

“I only test at Company X because the people at Company Y just want their ethnicity estimates. What good are they to me without trees?”

 

We hear these complaints all the time. All. The. Time.  But are they reasonable?  Personally, I’ve made some important discoveries in both my own research and for my clients using matches without trees, but that’s hard to quantify.  What we can quantify, though, is the percentage of DNA matches who have trees associated with their results.

First, how do you know if a DNA match has an associated tree? It’s different at each company.  The matches at Family Tree DNA (FTDNA) have a small tree icon near their names; the icon is blue if they have a tree in their account.

 

At AncestryDNA, you can not only see at a glance whether a tree is linked, but also how many people it contains.

 

23andMe makes the tree harder to find. For each match, you need to click on their name in your list of DNA Relatives, then scroll to the bottom of the next page. If they have a tree, you’ll see something like this:

 

GEDmatch users may also have associated trees. If so, they will have an entry in the GED/Wikitree column in the One-to-Many tool’s results.

(MyHeritage users can also link trees to their DNA results, but I didn’t include them in this analysis because I don’t have access to enough accounts there.)

 

What Percentage of Matches at Each Company Has Trees?

To answer this question, I looked at the DNA results of 12 unrelated people who are in the following four databases: AncestryDNA, Family Tree DNA, 23andMe, and GEDmatch. Because all 12 people were in all four databases, the possibility that one database is better for a particular group of people is minimized. Most were American, of European, African, Mexican, and Ashkenazi heritage, and one was Scottish.

For each person at each site, I determined the ratio of matches with trees to the total number of matches. (See below for details.) For FTDNA, 23andMe, and GEDmatch, the total percentage of matches with trees was straightforward to determine. However, at AncestryDNA, an account can have a tree that isn’t necessarily “linked” to the DNA (meaning the user has associated the DNA test with a specific person in the tree). At AncestryDNA, I first determined the percentage of all matches with linked trees, then I looked only at the ones saying “No family tree” (see screenshot above) to see how many did in fact have a tree that was unlinked. Finally, I used the two figures to calculate the total percentage of matches at AncestryDNA with trees (linked + unlinked).

Here are the results.

The differences were stark.  And consistent.  For every one of the 12 people, the rank order of matches with trees was AncestryDNA > FTDNA > GEDmatch > 23andMe. That rank order applies even if I only consider the linked trees at AncestryDNA and not the unlinked ones. What’s more, the lowest percentage at AncestryDNA (42.8%) was higher than the highest percentage at FTDNA (42.1%) and the same applies for FTDNA over GEDmatch and GEDmatch over 23andMe.

I was surprised at how few people at GEDmatch—only about 13%—have trees, given that it’s a specialized site that presumably attracts more dedicated genealogists. The low 5% rate at 23andMe is surely because that the company doesn’t host trees at all and makes the option to link an online tree difficult to find.  (For instructions, see the end of this post.)

The great news here is how many matches do have trees:  more than 75% at AncestryDNA (if you include unlinked trees) and nearly 40% at FTDNA. Although it sometimes feels like we all sitting around in the rain with half-empty glasses at Eeyore’s birthday party griping about trees, the reality is far sunnier.  There is so much family history out there, just waiting for us to use it!

 

How I Got the Data

Here are the gory details of what I did at each site, in case you’d like to repeat the experiment yourself.  If you do, please post your results in the comments!

AncestryDNA: I scanned the match list for each person using the DNAGedcom Client. The resulting csv file for a given kit lists all of its DNA matches and includes a ‘treeurl’ column that is populated only if the matching kit has a linked tree. The ratio of matches at AncestryDNA with linked trees is then:

 

Family Tree DNA: The DNAGedcom Client was used to scan the matches (Gather Match/ICW scan) and their trees (Gather Trees scan) for each person. I used the number of entries in the Family_Finder_Matches file and the number of unique entries in the ‘resultid’ column of the Family_Finder_Trees file as follows:

 

 

23andMe: I know of no scan that will survey all of the matches at 23andMe for those with trees. To get the percentage, I opened the profile pages for the top 100 matches for each person and checked to see whether they had a tree linked.

 

GEDmatch: I ran a One-to-Many analysis on each person. The percentage of matches with trees was:

 

How to Link a Tree to Your DNA Results at 23andMe

Log into your account and click on your name at the top right. Select Settings.

 

On the next page, scroll down to Privacy/Sharing and click Manage your profile.

On the Update DNA Relatives Preferences page, scroll down to Enhance your profile and click Family Tree URL. You copy and paste the web address (a.k.a., URL) from your online family tree at any of the following sites into the empty field: MyHeritage.com, Ancestry.com, Findmypast.com, FamilySearch.org, Genealogieonline.nl, Geni.com, WikiTree.com, RootsWeb.

Scroll to the bottom of the page and click the blue Update button.  Now you have a family tree linked to your test results and 23andMe. Your DNA Relatives will appreciate your efforts!

25 thoughts on “The Glass Is More Than Half Full—2017 Version”

  1. A lot of the trees I have for matches at AncestryDNA are very minimal and lack any useful information. Many of those with unlinked trees starting with a pedigree with “Private” and not the ID of the testee – so you can never be certain if the pedigree is the pedigree of the match! Personally, of my 100 closest matches – 42 have linked public trees (not all are very useful) and another 9 have linked but private trees. I don’t mind messaging anybody, but my rate of reply is low! But then, you don’t always need a tree, especially with a very close match, to get a good idea who that match may be!

    1. I intentionally did not consider tree quality in this analysis, although it’s an important consideration and I’m thinking of looking at that next. Useless trees are not unique to AncestryDNA; I’ve got matches at FTDNA with just themselves in the tree! The question then becomes: are the trees at AncestryDNA better, worse, or equivalent to those at FTDNA. (I wouldn’t include the other sites in this comparison, because I can’t pull data with the DNAGedcom Client.)

  2. In keeping with the half-full theme, 23andMe does have a mechanism for adding family surnames and locations to a profile. This may draw in some people who are intimidated by the very idea of creating a family tree. I downloaded my aggregate data just now, and 598/1095 (54.6%) of my named matches included surname information, and about half of those also include place information. These are text fields that can be searched in the DNA Relatives match list.

    1. I was just coming in to say the very thing Ann Turner did..The “new Experience” provides me with enough tools that if the match is decent size, the “shared matches” and surnames ( when they have them) and genetic data give me a great deal of information

  3. I’ve noticed the number of trees on Gedmatch has increased since the link to the tree was included in the one-to-many table.

    I tested my mother recently at Ancestry. Fifty eight of her 101 4th cousin or closer matches don’t have trees. It was very disappointing since most of them appear to occur on her major brick wall lines. I’ve worked out how we’re related to three of them so far. I plan to spend my summer holidays finding second cousins to test to hopefully resolve some more of them.

  4. I have heaps of matches on GEDmatch to people who have tested at AncestryDNA. There is little understanding of the benefits of having a tree at GEDmatch as well.
    Most matches are very helpful, but some do not understand that their permission settings lock out anyone who does not also have an Ancestry subscription.
    In my experience, this is the flip side of so many people having trees elsewhere. They do not also place one at GEDmatch. I place a tree wherever my DNA is. Because – well I want people to respond, right? Because I want to find cousins, not just plant a flag.

  5. I subscribe to ancestry.
    Personally, I think more people would make their trees more informative if the tree linked to their DNA results was only visible to their matches, rather than everyone subscribed to ancestry. Everyone’s fear is that part of their public tree will be copied and added to an incorrect family. I think this is less likely to happen if only dna matches can view a tree.
    (It seems my matches fall in to the region of private, no trees at all or limited information on the trees)

    1. The problem with making your tree only available to DNA matches is that would exclude cousins who are in fact related, but do not match your DNA. I

      1. You have a good point Julie. I don’t know what the answer is then. But I do know that I’m not comfortable making my whole tree public on ancestry. I have done too much work on it to have parts of it copied on to the wrong trees over and over again which has happened numerous times. It wouldn’t be so bad if people would just say something like ‘hi, do you mind if copy parts of your tree’ and give me a little information to confirm our ancestral link. But as the rules are now, by making my tree public I am agreeing to anyone copying bits of it. Many ancestry subscribers know this and won’t share their tree. I have a ‘public’ skeleton tree of direct ancestors because I know how frustrating it is to see ‘no tree’ or ‘private tree,’ but of course that limits the people who will recognise a familiar ancestor. I very rarely get an email from anyone who shares dna with me and never from a dna match on Gedmatch (which I love never the less)

  6. Interesting post. I see you did two columns for linked and unlinked trees at Ancestry. FTDNA is the only company that allows you to also record specific surnames (for matching purposes) which some people prefer to do instead of adding an actual tree. It is equally as helpful. I understand that you were specifically looking at trees however it would have been interesting to compare statistics if you had included two columns for trees and for surnames and a total for FTDNA like you did with Ancestry, I think it would change the outcome considerably.

    1. I did two columns for AncestryDNA because I had to tally trees there in two separate steps (first, for linked trees, then for unlinked ones).

      23andMe also allows you to list surnames and locations. I disagree that a list of names/places is just as helpful as a tree.

  7. I really like the FTDNA feature of making a tree only available to matches. That way I don’t get emails asking about “Browns” when they aren’t even related to me at all, let alone on my Brown line.

    1. ANYONE can search the trees at FTDNA, including people who aren’t customers there. To see for yourself, log out of your account, go to their web page, and enter MyMomma (no space) in the search field at the top. If you click on the tree icon beside the (fake) name, you can see the whole tree.

      1. You can search find some trees, but only those whose owners have not gone into their account and set their settings to either “matches” and/or “private” for the various categories (Deceased people born 100+ years ago, Deceased people born in the last 100 years, and Living people). If you set any of those for public, then the public can find them in a search.

        1. The point is that I can make my tree public and EVERY living person in that tree will be visible, whether they agreed to it or not. No one should be allowed to share personal family information through FTDNA’s system without the other person’s consent, especially in a world where maiden names are often used for password recovery.

  8. I find the predicted relationship provided by the companies interesting.

    Of my 1200+ FTDNA matches, I know my actual relationship to eight of them. All, except one, are within the predicted range. The exception is a first cousin who is predicted to be a half-sibling. This is because of the very close and unusual relationship of our mutual grandparents. So FTDNA predictions seem reliable

    Of my many Ancestry matches, I also know the actual relationship with eight of them. A couple are in both databases. Ancestry predictions encompass the actual relationship in only one case. A first cousin once removed was predicted as 1st-2nd cousin. In every other case, the actual relationship is closer than the predicted range. So Ancestry predictions seem unreliable.

    Have others found the same thing?

    On the tree question, I have never found a tree on either FTDNA or Ancestry that was of any use at all. They are either private, very small, interested in other lines to my own, or simply and demonstrably incorrect.

    None of us is perfect of course, and mistakes will happen. I have had fruitful responses from some matches that have helped us both. Reply rates are pretty low overall.

    Despite these whinges, for me, the glass is half-full for sure.

    1. FTDNA tends to overestimate relationships (that is, predict them as closer than they truly are) because they include small, false-positive, segments in their total amount of shared DNA. The overestimates aren’t a problem for closer relationships (2nd cousins and closer), but they’re a big problem as you get more distant.

      You are right that Ancestry tends to underestimate. They don’t have separate categories for removed or half, so they “round down” for those centimorgan ranges.

      I get great utility out of the trees, although I put a lot of work into them myself. A small tree can be augmented (in my own account). An incorrect tree can be fixed. Other lines can be built out. And private trees are often searchable, letting you know which surnames are in it. There’s also this trick for identifying the Shared Ancestor Hint in a locked tree:
      https://thednageek.com/pruning-to-find-shared-ancestors-in-private-trees/

  9. I made a tree at 23andMe some years ago, for my father’s account. When 23andMe decided to discontinue the trees on their site, and instead offered customers the ability to link the 23andMe trees to MyHeritage instead, the option was given at the time to keep a tree at 23andMe only (which I did, but I don’t know how many others chose to do the same). My father’s account showed the tree for a long time afterward.

    I thought that there still might be a few trees like that in people’s profiles on 23andMe (see https://customercare.23andme.com/hc/en-us/articles/229598288-23andMe-and-MyHeritage), but I checked my father’s account and his no longer appears (created pre-May 2015 at 23andMe). Perhaps it disappeared once they changed to the New Experience; I never thought to check. I guess they’re trying to say on the linked page that there is no way to have a tree at 23andMe, since they talk about them in the past tense, but they sure don’t make it very clear!

    1. It’s a shame that 23andMe doesn’t emphasize trees. We can now link a tree from any site (MyHeritage, Ancestry, FamilySearch, etc.) now, but it’s pretty hard to find, both in terms of the settings to link at tree and the tree itself from the match’s end.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.