Match Transform Scores Similarity

[Former Member]

Godo afternoon,

I am new to performing the match transform and have a question about the scoring results. I have noticed that when I run a match on my criteria values seem to be grouped correctly, however the scores all appear to be the same even if there is a difference in the values.

For example if my matching criteria is First Name and Last name, ID's are matched together correctly even if there names were slightly different.

First Name Last Name Score Rank

Ian Scott Null M

Ian Scott 100 S

Ia Scott 100 S

Ian Scot 100 S

Obviously there are difference in the sub records but the score comes back the same. is there a threshold that I am missing?

Thanks,

Ian S

[Former Member]

Match transform configuration is a complex process with many interacting match options. It will be hard to exactly answer your question without the exact match options you are using.

Based on the sample data provided and inferring that each field has its own criteria and this is Data Services 3.x the option your looking for is the Abreviation Adjustment rule on each field.

IAN       IA
SCOTT     SCOT

As shown in the examples, abbreviation means that the first letter of the shorter word matches the first letter of the longer word, and all remaining letters of the shorter word appear in the longer word in the same order as in the shorter word. The value you enter is the score given to the letters that are in the longer word but not the shorter word.

Enter a value of 0 (zero) to disable abbreviation checking.

Enter a value greater than 0 to enable this option.

Enter a value of 100 if you want abbreviations and longer words to be considered a perfect match.

Note: For this option to work, you must set the Compare data using option to Word Similarity.

If the conditions speciied are not the options you are using, then I would need to see the Match transform exported to XML or ATL to see your exact options.

[Former Member]

Brian,

Thanks, I ended up doing as you suggested and got variations in scores. Question though, if this was for numeric values would the same solution work? You had mentioned to change to "word similarity".

Reason I ask I have a good number of fields that have numeric values that I need to compare. I will test, but wanted to run by someone else.

Appreciate your insight.

Ian S