My last post about our new search received a number of comments from concerned readers about the new search experience. I’d like to address some of them in this post.
Before I begin, I think that it is important to explain that most people really like the new search interface. According to our research, over 75% of the people who have tried it out think that it is at least as good as the old search experience, and over 50% think that it is significantly better than the old search experience. That’s a great start, but we’d like even more of you to be happy with the new search experience–so we’re going to keep working on improving it.
It’s also important for me to clarify something about the new search. Nearly all of the search changes you see are in the interface, not in the search engine itself. That’s because making a change to the interface is usually more straight-forward than making a change to the search engine. Changing the search engine is more like turning a huge boat–it is a big operation that takes a long time! But that doesn’t mean we’re not working on it. In fact, we’ve been toiling away on some good improvements to the search engine that should make searching Ancestry.com easier. I hope that I’ll be able to announce some of those changes in the next few months. So, while our search engine team is trying to “turn the ship”, we decided to make some changes to the interface to make it easier to use.
Now on to the concerns. Here is the general gist of a few of the recurring concerns…
Concern #1: The search is returning results that aren’t even possibly right (Examples: Searched for someone who died in 1861 and got back a match for the 1930 US Federal Census. Searched for a person born in West Virginia but get back matches from the British Isles).
This has to do with the way the search engine works–it basically looks for any possible match that might be for your ancestor. Now it’s unlikely that a death in 1861 was mistranscribed and should have been 1961, but it is possible. Thus, the engine will return the match if other elements of the record look similar to the search criteria even if something looks wrong, just in case the input data was incorrect. That said, this is something we’re trying to improve to make the search engine stricter when it comes to dates that are clearly outside of the person’s lifespan. As for the location fuzziness, we return these matches because they could possibly be for the right person—we find that often times users inherited incorrect information about their ancestors and/or they didn’t realize their ancestors lived in a different location for a time period. I’ve had this happen to me several times in my research. That said, I know it is aggravating when you know a match is wrong and you still get it. We’re trying to improve the search engine to make it stricter on locations as well.
I tried a ranked search in both the old and the new systems for John Williams b. 1782, d. 1861. The first matches I got were in the US Federal Census Mortality Schedules in the right time frame. The only way a 1930 match would appear first is if the search engine did not find anything that was closer in terms of the names, places and the dates I searched on. Now, if I want to get only results that closely match the death date, for example, I can simply click on “Advanced” in the new search system, open the death information and click the “Exact” checkbox for the death year. (I can make timeframe a little broader by giving it +/- 5 years.) When running this search, I could not possibly get back any 1930 census results because the dates do not exactly match any data in the 1930 census records. Similarly on location, I can simply check the “exact” box next to the location I want to match exactly, and I will receive only matches with that location. Using the advanced features is simple and will weed out the other partial matches.
So the bottom line is that we think we can eventually make the search engine stricter on these types of matches. You can also use the “Advanced” functionality in the new search to limit the results you get back on a particular field to only those that exactly match what you specify–this should also eliminate any of those erroneous 1930 census matches or the British Isles matches.
Concern #2: I can’t find what I’m used to in the new search, but I can in the old search
I’m not sure exactly how to respond to this issue. We’ve done a significant amount of testing, and nearly all of the searches between the new and the old search return the same results when they’re entered the same way. If you find examples where they are materially different, please send VERY SPECIFIC EXAMPLES to firstname.lastname@example.org and I’ll take a look at them and see what we can learn (please include the URLs/addresses to the results so I can see them).
Concern #3: The card catalog is hard to use
I agree that the card catalog isn’t as easy to use as it should be. We’re working on making it a lot easier to search the card catalog, rather than only being able to browse it. We’re also trying to make it more intuitive and use the space better to display more matches in a single screen. Hopefully you’ll begin to see some of those changes in the next couple of weeks.
Concern #4: Ancestry.com indexes are low quality–you should spend time fixing the indexes rather than improving the search
Building indexes from hand-written records is extremely difficult and time-consuming, and is as much an art as it is a science. My first experience indexing old records about people that weren’t in my family lines was a humbling process. We spend six months training each of our indexers to understand nuances of old handwriting in order to bring more content online quickly while still meeting quality standards–we also spend millions of dollars each year making records available. I know that our indices have transcription errors in them–any indexing process does. To help combat this common problem, we encourage anyone who finds an error to provide corrections to the names in our indices. Those name corrections are usually re-indexed as alternates within weeks of being submitted. In order to submit a comment or correction on a record, simply click on the “Comments and Corrections” link on the record page. We’re also working on ways to allow you to correct any information on the records, not just names. Additionally, we have internal maintenance projects to improve the records we already have online.
That said, I think the solution to this issue is really two-pronged: (1) Allow anyone to correct mistranscriptions; (2) Have the search engine find fuzzy matches on names, dates, and places–our name search, for example, searches on exact matches, as well as matches from our name authority (full of alternate spellings), Soundex matches, and common abbreviations and misspellings. This allows you to more easily find transcription errors. Similarly, the fuzziness around dates and places also helps with transcription errors on elements other than names.
Concern #5: The location fields in the new search don’t recognize counties
The location fields in the new search do recognize counties. You can simply begin typing the county name and select it from the type-ahead list. For example, if I start typing “Utah” in the location field, the type-ahead listing gives me an option for “Utah, USA” (this would be the state), and the next option is “Utah, Utah, USA” (this is Utah County in Utah). Now if I type in “Payson” I get a choice for “Payson, Utah, USA” — that doesn’t mean that the county won’t be searched, it is just that we’re not displaying it in the limited space available in the search box. We still search on the county as well as the city. I hope this clarification helps.
Concern #6: I keep getting zero results in the new search
I think I understand what’s happening here–one of the neatest features in the new search is that it remembers the information from your Ancestry.com Member Tree, and when you begin typing names into the search box, you can select a name from your family tree and it will fill out the whole search form for you. Obviously, your tree may contain many details about the person for whom you’re searching, including the names of their parents, siblings, children, their birth and death information, etc. As a result, checking the “Exact matches only” checkbox would return only matches that both have all of that information specified AND match each element exactly–finding such a match would be a truly rare event. Thus, your best bet is to avoid checking the “Exact matches only” checkbox when using this feature. Instead, turn on the “Advanced” options and select a few of the fields that you want to be exact rather than the entire set of fields. For example, you may want to mark only the surname, the birth year, and the birth location as exact–this should give you a small set of good matches.
Thanks for posting your comments about our search experience. Your passion for our products and family history research really come through. That’s what we’re trying to do with the new search–make that experience easier for everyone. We’re dedicated to improving the new search experience on Ancestry.com to make it better and better, and your feedback is valuable in that process.
About Kendall Hulet
Kendall Hulet has served as our Senior Vice President of Product Management at Ancestry since March 2015. He joined the Company in 2003 has held a variety of roles in the product organization including Director of International Product Management and most recently Vice President of Product Management for AncestryDNA. During his tenure, he was deeply involved in some of the most popular innovations at Ancestry, including the “Shaky Leaf” hinting system that has delivered over five billion discoveries; the Ancestry Family Tree system that has led to the creation of over 70 million family trees containing six billion ancestors; and the creation of the award winning Ancestry mobile app, which has been downloaded more than 12 million times.