To set up identity resolution rules, you must first configure your input models and select the appropriate identifiers.
This page goes through process of creating models, selecting identifiers, and customizing your resolution operations:
Model selection
Hightouch supports two types of models:
- Profile models
- Event models
Profile models typically represent users (for example, contacts or people). However, Hightouch's Identity Resolution is configurable to any granularity of entity, whether that be households, accounts, organizations, or whatever else your business requires.
Event models, conversely, are for timestamp ordered events (for example, transactions or clicks).
You may use as many models as required in your identity graph, and depending on your use case, you may need a combination of profile and event models. For example, you may use a users table, a Salesforce contacts table, a transactions table of events, and a table of "identify" events.
Model order columns
You'll notice that event models require you to select a timestamp column to determine the ordering of rows in the model.
The timestamp column is used for incremental processing. Hightouch will only pull in new events on subsequent runs to prevent re-processing rows that have already been processed. This significantly improves performance of the identity resolution process and reduces cost on your warehouse.
Identifier selection
After you select your model, you need to configure which identifiers are present on this model. You can choose from Hightouch's default options, or you can create your own by typing anything into the identifier text box.
You must map the identifier to a column in one or more models to use it in a merge or limit rule.
Sequencing
The order in which you sequence your models may be important. For example, if you have a limit rule that prevents more than 3 users merging together, Hightouch merges the first two users based on the sequence provided here. More users won't merge in to the same record.
If you don't have any limit rules, then the sequence doesn't have any impact on your identity graph.
Sequence vs model order
Models are ordered via the model sequence and rows ordered within models based on the model order by column.
Hightouch imposes a "total ordering" on all rows in the identity resolution process by using the model order, then the order by column within each model.
Based on the data and limit rules you configure, rows of higher priority models may merge together while rows of lower priority models may not merge.