Wednesday, March 29, 2017

Elusive Bug in Ancestry Search

Elusive bug in Ancestry searchFor years people have told me that there are bugs in Ancestry’s search. It was returning something it shouldn’t. But each time they sent me an example, I found the search system was working as intended. It was displaying results that partially mismatched because other parts matched so strongly. There was a distinct possibility it really was the person you were looking for, albeit you might have provided a little bit of incorrect information, or the record itself contained a little bit of erroneous information. Now whether you agree or disagree with that approach, the fact remains, the system was working as Ancestry intended it to work.

Another scenario users experience is that the library version returned different results than the home edition. I haven’t bothered to track that one down, but I don’t believe it. It would require adding special code to handicap one or the other. There is no incentive for them to add extra code that would lead library users to believe Ancestry works poorly. If you experience this phenomena, chances are more likely that you have the filters set differently between the two.

Another phenomena users experience is that they perform the same search twice in a row and once it returns less results. When I instruct them to try it several more times, it always returns the greater number of results. This is not a bug in their code, although I suppose it should be considered a bug in their architecture. Years ago Ancestry met with bloggers and explained that their search system divided up global searches among several computers. The search system took the results from the several computers and assembled them for presentation to you. This architecture had two ramifications. The several computers didn’t all finish at the same time and the order they completed was not fixed, resulting in the search results being in a different order each time. I don’t know if they still have that architecture, but if they do, it explains why once in a great while a search returns fewer results than it does every other time you run it. What happens is that one of the computers assigned part of the global search completely failed. It’s results were never returned to you.

Still, the specter of an elusive bug persists.

Well, I think I finally have evidence that the bug exists.

If you perform this search, the only result is Thelma I Raymond (shown below). 

Thelma I Raymond is the only result.

One of the search parameters specifies she was born in 1909. If I remove that constraint, I should get additional results—and I do. But she still matches, so she should be included in those results—and she isn’t (shown below). I get six results and she isn’t one of them. She matches better than the other six results so she should be included.

Thelma I Raymond is not among the results.

I don’t believe this is the intended behavior. That makes it a bug. So if you have long believed you’ve experienced erroneous results, you may be right.

13 comments:

  1. Don't know if it is a bug or an errant omission of a field, but since Ancestry redid the HeritageQuest site, the books there--some if not all don't have author attributions and one result is no author search

    ReplyDelete
  2. When I teach my beginners class, I always tell them to do the search more than once and with different parameters. I have known for years that the result you demonstrated is exactly what I had been getting. I reported it ions ago. Now, I just work around it. I never trust online searches. Pam

    ReplyDelete
  3. I have gone in done a regular search and found something. Then I would go in with Family Tree Marker to use it as a source and search the same way and would not be able to have it pull up. I don't sync but do source through the program.

    Very frustrating and time consuming.

    ReplyDelete
  4. For YEARS I have said the results I get are different from the results my cousin gets using the exact same search terms. I think it was last year, sitting in the chair not 3 feet from me in my own house using my computer signed in to her account while I used a different computer signed in to my account I showed her the result I got from a search and told her to look at that one. She didn't have it. She couldn't even get it doing multiple searches. And that is another reason to save everything you find because it's a crap shoot whether you will ever find it again.

    ReplyDelete
  5. Similarly, If I inadvertently X out of a Results screen and try to restore the same results, I cannot reproduce the same results screen. One or two of the results that came up earlier do not show at all.

    ReplyDelete
  6. Funny that you blogged about this Ancestry "feature" today. Lately Ancestry seems to be choking on itself. The biggest problems for me are incomplete searches and trees that won't update. Searches are returning less useful results than usual. I've been told that searching individual databases and avoiding global searches (the default) give better search results. This makes some sense to me if Ancestry searches are so massive that they are divided across servers on different continents. On the other hand, splitting databases and calculations among servers isn't new technology on the "cloud."

    Thanks for providing an concrete example of something all Ancestry users know to be true, if only anecdotally. The problem is that it happens often enough to be aggravating but not enough to drive users away.

    ReplyDelete
  7. I have seen this behavior forever. I also reported it and offered up examples. First I was told I was mistaken, and then it was written off to "search term particularities", whatever that means, and lastly "we are in the process of improving our search, results may vary". I have just accepted it now as just how it works. I even had examples where I loosened the search and got less matches. Try explaining this to someone new to Ancestry.com.

    ReplyDelete
  8. Just why I prefer searching for records on FamilySearch, then searching Ancestry for extra records that are missing or indexed differently on FS.

    ReplyDelete
  9. The Ancestry Insider,

    You have a great example, for me at least, as to why I search broadly, then add pieces of information to reduce the results. As a colleague and friend says, search for failure. That is adding bits and pieces, move those wonderful sliders until you get a failure, then back that list slider or bit of information back one notch.

    Not saying you haven't spotted a bug, as you may well have, but my approach for Ancestry or any other website, "can I find my person", get many results, then add additional information until I get a failure.

    That process has worked for me many times.

    Thanks for the post.

    Russ

    ReplyDelete
  10. Dear Ancestry Insider

    Although it has been known to many of us for a long time that there are errors in the Ancestry search engine, it follows for all search engines.

    There is no one perfect algorithm for this as anyone should know that it the search engine is trying to also incorporate old and inefficient OCR text readings in the results.

    We as researchers have become all to complacent and too reliant on this.

    Resesrch is work, hard work and there is no substitute for this.

    As for a bug in the system, so can se said for Google, Bing, My Heritage, Family Search and on an on.

    Thanks for making us feel better with this post.

    ReplyDelete
  11. One year ago I retired from a library system that offered Ancestry's Library database. I regularly got fewer results in the library than at home. I don't know if there are filters set on libraries according to how much they pay Ancestry or what, but there is most definitely a difference.

    ReplyDelete
    Replies
    1. I agree, Coral. When I meet a client at a library, even if I'm signed on to my Ancestry account through my device but am using the library's wireless connection, I get different results than what I found at home on the same device. I always thought that the library edition not only had different filters but access restricted to some Ancestry records. But if I'm signed on to my account, I should be able to see databases that might not be included in the library edition so I could never understand the results.

      Delete
  12. The Library Edition doesn't have the same content as the home version, so the results that you get may be different. The library version has the foreign content, which home users may not subscribe too. Also, certain content is not available in the library version, but it is available through the home version. One example of this is the newspaper content. Here is some info: http://www.vims.edu/library/_docs/anc-comvsanc-lib022013.pdf.

    ReplyDelete