Reltio Connect

What is the difference between relevance-based match and Match IQ and what are the best practices

Relevance-based matching, let's talk about how it is different from the automatic and the suspect rule matches that we have, or match rule types that we have. Automatic and suspect match rule types are more of a binary output rules. When you put some instruction in one of these rules type, the result will always be either true or false, whether it matched or it did not match. If it did match, then it basically goes and perform the action that is tied to that rule type. For example, in automatic, it always goes and automatically merge, with the suspect if the bad is evaluated then you will create the potential match. With the relevance-based matching, you have the option to basically use the score in the matched pair to determine what action you want to take on that.

That's really the difference between the binary output match rule type versus the relevance-based match rule type. Every match pair that is evaluated using relevance-based match rule, you will have a score and you will get to basically define the range of score and what action match engine your platform should pull from on that match pair. That's the difference between the relevance based and the other type of match rule. The difference between relevance based and the match IQ, match IQ is basically a completely different matching engine in that trilogy of support, in addition to the rule-based matching. Match IQ is ML-based, or machine learning powered matching. There is no rule configuration that user must do, we train the model, and we have provided a tool which will allow you to train that model by answering the question to the sample pairs that we present.

And based on the answers that you provide, the match IQ module basically derive the matching requirement and build all listing, whatever comparator, tokenize, whatever we need to build, based on those answers, we will build a model, machine learning model, which is then used for running the matching. Relevance based matching, automatic mode match rules and the suspect rule type, they all fall under the rule-based matching, where users put specific instructions, how the matching should perform. And match IQ is machine learning power-based matching, where you don't have to write any of these rules, all you do is train the model based on your matching requirement and match IQ will basically run the matching for you or do whatever it has to do in the background for you.

If you have not watched the Community webinar on Matching and Merging, take a look below: