Some people eat, sleep and chew gum, I do genealogy and write...

Sunday, May 5, 2013

Comparing the search engines: FamilySearch, Ancestry and MyHeritage

I thought it might me interesting to do some controlled searches with all three of the US-centric online databases to see how they stack up. What I mean by "controlled searches" is that I will be looking for sources I know are already there in all three databases. I decided to use names from my family lines where I have already gathered quite a few sources. I have done this before to compare online search engines such as Google and Bing, but I don't recall making the same comparison with these big data providers. The comparison is now possible because of the recent addition by MyHeritage.com of all of the U.S. Census records. Now all three search all of the records, even though FamilySearch is missing some of the images.

Of course, you all probably understand that this type of test is far from scientific and results will vary widely depending on the family selected. One issue in considering the number of results is that a large number of false positives, although initially impressive, is really meaningless. That is why I am using searches where I already know the content of the database. Another factor that cannot be taken into consideration is my personal ability to make appropriate searches. You may have a much greater ability to search than I do and your searches may show different results.

At least with the US Census Records, all three large databases have the same records, so if they search engine finds the name in the Census, it should be a good indication of the search engine and not the content.

Here it goes. I am going to use my perennial example, my Great-grandfather Henry Martin Tanner. I have found him in all of the available US Census records from 1860 to 1940. So how do they do on a first pass?

I will start with MyHeritage.com. I can search directly for records for an ancestor from his personal page or profile in my family tree. OK, the Record Match program of MyHeritage, already showed 2 records; neither of these were US Census records. So I clicked on Research this Person. There were 743,982 results. Somewhat overwhelming. I refined the search by looking for only Census and Voter Lists. The number dropped to 238,886 but the results showed Henry in the following US Census Records:

1930
1920
1910
1900
1880
1870

All of these on the first page of the search. Let's see how the others do.

Ancestry.com also can do a search from the person in your family tree. Of course, to duplicate this type of test, you would have to have your family tree in both programs which I happen to have. OK, so how did Ancestry.com do? Here are the results:

1930
1920
1910
1900
1880

That's all I found on the first page of the search. Looks pretty much of a draw so far, except Ancestry.com did not seem to pick up the 1870 US Census. Oh, by the way, Ancestry.com had 490,459 results. I am not sure what the programmers think the users are going to do with all those results? But the numbers look impressive. When I filtered the results to just Census and Voter Lists, the number dropped to 71,442. But the search engine still did not find the 1870 US Census Record.

So far, neither search has found the 1860 US Census.

Now what about FamilySearch.org? In this case, the program does not do a search right out of the Family Tree program so that may put the program at a disadvantage. But I will put in more information than I would usually do for a search to see if that evens things out. Here it goes:

None of the US Census Records show up on the first page, which has, by the way, only 1,894 results. I  try to filter the results by Census and Voter Records and find that I cannot do so. Hmm. What is going on here? I decide to go to the collections directly. I can filter the collections by name or by other criteria, but when I filter the collections to show only Census & Lists, I cannot search the list of filtered results which clearly contain the Census records. So, I have to search each year's census records individually.

I decide to try. I first do a search for Henry Martin Tanner in the 1860 US Census directly by searching that particular collection. The results for Henry Martin Tanner with birth place, year and residence is zero. I decide to jump to the 1880, both MyHeritage and Ancestry found him in this year on the first page of results. Again, the results are zero. No records found. So, I decide to search for just "Henry Tanner" in the same year. Maybe the "Martin" middle name is creating a problem. No records found.

It seems to me that FamilySearch is out of the running on this one. I try taking out the birth year to see what happens. Oh, that's it. I had the birth year off by one year. I put 1852 instead of 1853, but guess what? He was actually born in 1852, not 1853. So I had to take out the right birth date to find him? That is really strange. Ok, so now I go back and do the search without a birth year. Guess what? All the Census records show up.

1930
1920
1910
1880

But of course, I have the residence as Arizona so that is as far back as the search takes me. What if I take out the residence? Well there go all the results. Now I don't find anything again. What if I add back in the residence as Utah? Nope, that doesn't work.

Well, both Ancestry.com and MyHeritage.com do a credible job of finding my ancestor. But FamilySearch.org appears to have a ways to go before it catches up with the other two. Guess what? None of them found Henry in the 1940 US Census.






2 comments:

  1. I would really like this comparison repeated using a female. Pick someone who has been found under multiple surnames, and even as a lone widow. Although I don't have an on-line tree to use as a starting point, I have found that global searches tend to have trouble with changing surnames.

    ReplyDelete
  2. It would be helpful if they would let you do an exact match. My Heritage gives that many hits because it is searching many variants with different middle names, different time periods, etc. If you could customize the search it would help me.

    ReplyDelete