Tuesday, September 20, 2016

How Accurate Are Ancestry.com Hints?

This is a question that I am asked occasionally by genealogy colleagues and blog readers.  I had a little free time today (hey, I'm retired, I can do anything I want all day long - um, except for when I can't, of course), so I decided to take a look at my Hints in my "big" tree with over 42,900 persons in it.

Here is the top of my Hints page for this Ancestry Member Tree:


Today, I have:

*  47,874 ALL Hints
*  33,625 RECORDS Hints
*  5,198 PHOTOS Hints
*  547 STORIES Hints
*  8,504 MEMBER TREES Hints

I usually just look at the RECORDS hints.  I get somewhere between 20 and 50 RECORDS Hints every day on this tree - for seemingly random persons in this tree.  I try to keep track of how many are added each day or so.  For instance, the last time I checked on 18 September, I had 33,523 RECORDS Hints, so I'm up 102 in the two days since then.  That is pretty typical.  When I checked the list on 31 August, I had 33,033 RECORDS Hints (but I have resolved 445 Hints since then), so that is 937 RECORDS Hints added since 31 August, or about 47 per day.

Today, I methodically went through the first 20 pages of the RECORDS Hints - that's 400 Hints.  I tried to judge if the Hint was for the specific person in my database, or was not, or might have been but I can't draw a conclusion.  For these 400 Hints, my judgment said that:

*  347 - YES, they were for the specific person (86.75%)
*  39 - NO, they were not for the specific person (9.75%)
*  21 - MAYBE, perhaps the specific person (5.25%).

This is a pretty small sample, but the results indicate that over 85% of the RECORDS Hints in my Ancestry Member Tree are for the specific person cited.  If you had asked me yesterday, I would have said 80% to 90% were for the specific tree persons.

Sometimes, the Hints are very helpful.  for instance, the first person at the top of the screen above is Sarah E. Holbrook (1839-????), who married Joseph Warren Dill (1836-????) in 1872 in Massachusetts.  The 15 Hints for her include records for her birth, her three marriages, and some census records.  I did not have her first marriage in 1855 in my database - now I do!  I did not have her third marriage in my database - now I do.  How do I know these records apply to the same person?  All three of the marriage records name her parents as Enos and Matilda Holbrook.

I am often amazed that the Ancestry Hints system can find some of the females with multiple marriages, especially when I have only one marriage in my database.  Other amazements are how the Ancestry Hints can find interesting name spellings and attribute them to the right tree person.

It is important to note that these Ancestry Member Tree Hints are derived from a limited set of all of the Ancestry.com databases - I've heard the top 10% of all databases listed by record count.  There are 32,720 databases in the Ancestry Card Catalog - so 10% is 3,272 databases.  #3272 today has 17,517 records, so the 10% covers a lot of records.  Researchers should always do a search for a person to find the matches that are not picked up by the Hints engine.

I will do this study again when more Hints are piled up in this tree.  Only by expanding the number of Hints evaluated can a better accuracy percentage be calculated.  I hope to get up to 2,000 Hints the next time I do this study.

Are you surprised by the (YES) accuracy percentage, or by the NO percentage?  What percentage do you think are accurate in your Ancestry Member Tree?  What about on MyHeritage and FamilySearch?

=============================================

The URL for this post is:  http://www.geneamusings.com/2016/09/how-accurate-are-ancestrycom-hints.html

Copyright (c) 2016, Randall J. Seaver

Please comment on this post on the website by clicking the URL above and then the "Comments" link at the bottom of each post.  Share it on Twitter, Facebook, Google+ or Pinterest using the icons below.  Or contact me by email at randy.seaver@gmail.com.


5 comments:

Cousin Russ said...

Randy,

I have done two studies on this topic. My view, however, are those hints that are presented in FTM2014.

In both studies, I ran about 95% of Good / Yes hints. I do NOT include Hints from Ancestry Member Trees.

My observation is, that the more information about the person, with the hint, that is in the database the better the hint. That means Dates and Places, but more important is Relationships.

The example I have: I enter a new person, with dates and places. Adding a relationship, that 'new' person, may reflect a hint. But, if I were to add children, as soon as I hit enter for the Name, I get hints and they are most likely to be accurate hints.

I have done this now on three different Families, with very similar results. I let a new file grow a bit before I look at the hints.

I consider the Ancestry Hints, both in my Ancestry Member Trees and within FTM2014 as my first level of research. I let their search engines find RECORDS for me. Its then up to me to Evaluate those hints.

I rarely Ignore any Hints. I may hold off on dealing with them, as I may not have enough information in my database to make a determination. I make a note, in my research notes / and ToDo list to go back and check later.

Russ

Geolover said...

I get a lot of hints that are for a parent of a subject in a database, but not for the subject -- such as WV births or marriages, and the social-security applications. I find this to be a bit strange. I also get a lot of hints to index databases that reflect data already entered in a tree. I don't find these items to be especially helpful.

Marcia Crawford Philbrick said...

Thanks for the statistical study. I don't have statistics but my experience would mirror the results with one exception: same names in a time/place. Two of my lines in particular suffer from what I would call 'same name syndrome'. With those two lines, I have to take extra care/caution in reviewing the data before attaching it to a specific person. For example, I recently found a newspaper article for letters at a post office (Indianapolis 1827) that listed Hiram M Currey. This could be the Hiram M. Currey that was treasurer of Ohio and likely moved to Indiana after 1821 or it could be the Hiram M. Currey who was a lawyer in Indiana before moving to Peoria, Illinois -- OR it could be a third Hiram M. Currey that I don't know about.

An enhancement to Ancestry would be the ability to 'discuss' the attachment of a record to a person in instances where two people of the same name get mixed together. (Perhaps that capability already exists and I just need to figure out how to do it.)

Cousin Russ said...

Marcia,

I have my share of same names. I think the issue is how much information, Dates, Places, and Relationships, do you have on those same names persons. I don't look at hints if I don't have that much information to be able to determine if the Record Hint is for my person. If I can't tell, how will the search engine find "my person"?

That is not to say I don't get those "bad" / wrong hints. When I look at the data I have in the profile, I move on. I do not Ignore that hint, just move on. I have had a number of cases where they disappear after I have additional information for that person.

For example: I just looked at one person's Hints (from my ToDo list). Two Ignored Hints. Both hints were for the same named person, located in a different state, but those two hints were for the same person, just not mine. When I looked in my database, I didn't have any Dates nor Place in the Profile.

I would not count that as a Bad Hint, but as an Ignored Hint, based on my database.

My way of "discussing" that hint, as I have NOT attached any record, is to make a note for that person on my ToDo List. When I go back to review hints for that person, I'll check my ToDo List for my notes for the Hint, check to see if I have updated the profile, and just leave the Hint as a hint and move on. I don't Ignore it, just move on. In many cases, and I guess I need to track this, when I just leave that hint, and it later goes away. I do have a number of hints that do Go Away, over time.

Only my view on Hints for "not my person".

Russ

June said...

Yesterday I was adding Family Bible entries to my tree. While checking the hints for verification all but one hint was person Specific, I thought. I clicked the link for the one I didn't think was mine to see why it was added. It turns out it was my person. The transcriber used the surname from the family above for my person. Upon viewing the record it gave me 100% accuracy for 200 records viewed yesterday. I would say my average is in the 90% range most days.

I don't usually use the Family Tree hints unless I'm stumped. I view them for record sources I may have missed only.