[This topic is pre-release documentation and is subject to change.]
What is predictive matching?
Predictive matching is the process of combining data from multiple sources. Data coming from various sources often use different fields to uniquely identify records, and contain different values.
For example, consider the following customer records.
In one company's records, customers are identified by First Name and Last Name. In another company, Name is used. The name given is Mike at one company and Michael at another. Yet, these two records represent the same customer. Predictive matching in Dynamics 365 for Customer Insights uses context-based matching of entities from diverse sources to merge the data.
Create a Predictive Match policy
Open your Customer Insights Customer 360 application.
Select Show Menu .
Select All Options > Predictive Match.
Select Add and fill in the values.
Item Description Policy Name The name of the Predictive Match policy. Policy Description A description of the policy. Relationship Threshold Default value: 80. Only matches with a score equal or higher than this value will be displayed. Save Results Default value: Top 1. Use to select the number of matches to save. Policy Type Profile to Profile:
Interaction to Profile:
Item Description Profile 1 Profile 2 Profile Property Match Type Default value: Default
Select Save to save your new policy. It will automatically run in the background and be available for your Customer 360 analysis.
Sample Predictive Match Policy settings
The following values were used to create a sample Predictive Match policy:
|Policy Description||Matching web registration to contact info|
|Save Results||Top 1|
|Policy Type||Profile to Profile|
|Profile Property||Firstname; LastName|
The new policy shows up as a tile when saved.
You can see the policy in action in Customer 360.
For example, you can see the predictive matching inferences for Karen Johnson.
Click the Show Predictive Matches in her Info card to display the predictive matches for her contact information.
In this sample, phone numbers appear to match as well as last names despite the differences. We therefore have a matching score of 99%.