Understanding Accuracy


We’ve received a lot of feedback regarding accuracy and wanted to let you know a little more about why we started showing the accuracy rating and about how accuracy is counted.

As more contributors joined the project we were asked many times for an indicator of how well they were doing – since we all want to know if our contributions are making a difference.  Based on this feedback we looked for a way to show you how well you were doing with your keying efforts and this is where the accuracy rating came in.  We are grateful for your feedback as we are always looking at ways we can clarify and improve both this feature and other features.

In regard to whether your contributions are making a difference that cannot be measured.  During this stage of beta testing your contributions have been incredibly beneficial – we would not be where we are today without you, the many contributors who give of your time keying, providing feedback, assisting us with bug testing, testing arbitration, and the list goes on.  Your accuracy rating does not reflect your value to the project and the value that you are providing to people who will view the indexes free because of your efforts.

There have been some speculations on how accuracy is counted and what types of records are counted and we want to clarify this for you.  Accuracy is counted based on the number of fields that were accepted in the arbitration process. 

  • If an image set has not reached arbitration it does not reflect in your accuracy. 
  • If all of the fields for Keyer A and Keyer B match in an image set that image set is counted as 100% accurate.  (Each image set goes through arbitration whether the fields all match or not.)
  • When Keyer A and Keyer B have fields that do not match the arbitrator chooses which, or neither, record is correctly recorded from the image.  Each of the fields that are accepted as correct positively impact accuracy and each field that is not accepted negatively affects accuracy.   With the accuracy rating being based on the field level it is clear to see how small mistypes or not keying all of the available fields will affect your accuracy negatively.

Our hope is that you won’t be discouraged by your accuracy score.   The beta process has been a learning process for everyone and we acknowledge that at times the instructions and directions have not been as detailed as we need them to be.  We appreciate your contributions to the project and for being patient and willing to continue on as we learn the best ways to set up the projects and communicate with you. 

As you see your accuracy rating going up or down – at least mine does – we want you to be aware of the resources that are available to assist you in keying and I hope that you will recognize your very important role as a contributor no matter what your accuracy rating is. 

Help articles

Tutorial (English Version)

Project pages (access these via your dashboard or by choosing “View More About This Project”)

Information and Links

Join the fray by commenting, tracking what others have to say, or linking to it from your blog.


Other Posts

Write a Comment

Take a moment to comment and tell us what you think. Some basic HTML is allowed for formatting.

Reader Comments

I have been Keying and Abitrating for a while now. Lately my Accuracy score dropped to 63% and then to Nill% Quite alarming really with no feed back as to why. I wonder if this has happened to any one else. or if anyone knows the reason behind this.

I just committed an error with the PA petitions. I was mentally gathering data on Jacob Enders, awaiting the Intention or Petition form to appear to enter it when the image set completed. He was not in the next set and therefore will not appear. Is there any way to recall an image set to rectify an error?

I do not mind an accuracy rating, but feedback is needed on corrections made to help improve input. Why is there no feedback???

Given the nature of this work and knowing that the accuracy of records can make or break finding ancestors, I feel the accuracy rating is even more important than the number of records keyed.

I’ve been a keyer for 4 weeks now, maintaining a 98-99% accuracy (21,000+ records keyed to date). When I started arbitrating 2 weeks ago, I was amazed by the poor keying of some records I’ve seen, and I’m not talking about typing in all caps or entering a leading zero.

I’m talking about the keyer who keys the surname & given name and leaves other fields blank, even when the information is CLEARLY WRITTEN on the record; OR repeatedly enters the given name as a surname and the surname as the given name; OR repeatedly omits the middle name(s) from the given name even though this information is on the record. The unwritten message I get from these keyers is “I’m only here for the discount.”

If completed projects were to go live without any accountability for accuracy, the discount wouldn’t be worth a hill of beans if the keyer can’t find the records he/she is looking for because he/she neglected to key all available required informtion.

My first 800 or so records had an accuracy score of 0% which perhaps could be dismissed as “no records checked yet”. It then went up to 36% and stayed there without change through 1,500; 2,000 and now 2,180 records.

I’m not convinced that the accuracy score is in anyway meanful. On the basis of transparency, if a score is to be given, the score at least aught to provide that date that it was last calculated and the number of records it was based on.

Phil

It just seems to me that the accuracy rating would mean a lot more if we had some sort of handle on just exactly what it is we are doing wrong.
I have run on to a few things that could have been entered more than one way. If I had some sort of feedback on what it is that I am doing incorrectly the accuracy would improve dramatically.
So far I am holding at 87%, but the perfectionist in me is screaming constantly. I know that this will eventually lead to discouragement and abandonment of the entire project on my part.

Oh Well.

I did my first 20 records and received 0% accuracy! I double checked all entries so it can’t be the typing. Did I not follow directions correctly? Did I use the wrong form? Please help me understand.
Audrey (audstew1)

I am unable to download the new keying tool. Help!

Cat

My problem with accuracy checks and using the drop down lists forces the keyer to at times follow misspellings started at the beginning of keying (other peoples). An example is BUSMHAN rather than BUSHMAN. BUSMHAN is shown in the drop down menu; the listing, however shows BUSHMAN, John (Ida). Am I wrong in presuming to ignore the drop down list?

I have the same accuracy rating for close to a month. I am doing the PA and MD project. I am at 84% and haven’t changed since then. I realize there are alot of projects to arbitrate, but I have only did this project since it started. Could someone please give me my correct accuracy?

Thank you
Barbara

As a new keyer, one thing that would be tremendously helpful is the ability to see which records were inaccurate. When I saw how low my initial accuracy was, I really went and read the instructions closer (which I should have done from the start). I’d like to continue getting better and seeing my mistakes would help.

In gouing over the naturalization record I find that the typed pages are often different from the handwritten pages, especially the signatures by the person in question. In keying the correct info I regularly input info different from the Typed page. Also there are documents that have a small piece of paper covering allowing for all required iformation to still be seen. When I input these pages although I am supposed to skip them, my rating goes up. When I input the handwritten info over the typed my rating goes down. Doesnt make sense. Are the arbitrators looking at the same records or just making assumptions? It is disconcerting to make a big drop when you work so hard. I guess however you rate, it would be difficult.Ever onward.

I just did my first project, and got back 46% accuracy which was very disheartening since I was so meticulous, and I scanned all of the help pages and was not able to find anything that addressed my questions!!! Since I couln’t get help I had to make an educated guess. It would have been so much more helpful if I could have seen what mistakes I had made (probably when added in Proprietor as it was in the doc even though there was no field for it) which would have helped in knowing how to key next time…

I agree with everyone who has requested feedback on errors. I have been over and over the instructions and tips for the New York U.S. Naturalization records and sometimes my accuracy goes up, sometimes down. If I knew what I was doing incorrectly, it would be a great help. I’m a little reluctant to keep going “blind”. Hopefully I’ll get a response from Support soon.

I downloaded two sets that I really didn’t understand how to key, so I cancelled them. I was using Firefox and it wouldn’t let me explain why in the box. I am much more comfortable with the NY immigration lists and there are plenty of them. Do those 2 cancellations affect my score?

I just started keying last week, maybe some of the pros can give me some insight as to why my accuracy needs improvement. It is not the keying, as I am a typist. It is not the decipering as they are typed cards. I believe it is the fields, the example given does not even closely resemble the card in the image set I am given. I spend so much time trying to decipher what they want on these cards that I could be keying if only the examples and instructions matched.
The other issue is how about showing us an example already keyed and more importantly where our errors were? If we don’t know where we erred it is a bit difficult to fix.