In previous blog posts I talked about how the ability of Novetta Entity Analytics to understand relationships between entities allows it to deliver more accurate entity resolution and analytics results. Today I’ll explain in more detail how Novetta Entity Analytics detects relationships within and between entities from many data attributes and sources, and why this awareness yields greater insights.
Payload Attributes Provide Relationship Insights
To understand relationships, Novetta Entity Analytics analyzes both the attributes used for entity resolution, and additional payload attributes that exist within the data but are not used during the entity resolution process. When analyzed in different combinations and joined back to resolved entities, payload attributes can identify hidden relationships and patterns within combined data sets. For example, payload attributes can define behavioral patterns within transactional data, such as recent purchases or those made at a particular location at a point in time. These records on their own can’t help resolve a specific entity, but they can detect relationships between resolved entities. Novetta Entity Analytics uses processes similar to those it uses for entity resolution to identify relationships.
If Novetta Entity Analytics can access resolved data about a person, a location type and an event’s location from A, B and C sources, it can derive relationships between entities from the data. The relationships can be simple ones, such as a house holding relationship where two resolved entities share the same house at the same time, which means they are either married or roommates. Or, the relationships can be more complex, such as identifying a patient’s employer from information contained in their hospital inpatient records, or a customer’s transactional behavioral patterns from information within transactional records. These examples show how both disclosed and undisclosed relationships within data may contain interesting information about hidden connections between entities.
To identify relationships using Novetta Entity Analytics, analysts map the output of entity resolved results, which contain the attributes used for resolution, to common relationship resolution rules such as house holding and hierarchies. Analysts review payload attributes within the combined data sources to identify attributes, patterns and relationships that may help them find specific complex relationships types useful for fraud detection, campaign analysis, market segmentation or other activities. The analysts then write rules that Novetta Entity Analytics can use to uncover these relationships within the data. This process requires that analysts understand the data sources they are working with, and how to use the software to detect relationships that are otherwise obscured by the volume, variety and complexity of enterprise data.
Relationship Awareness Depends on Available Data
The types of relationships Novetta Entity Analytics can uncover depend on available data sources. Analysts who are looking for familial relationships among customers, but only have access to data from a master data management system containing customer information, can most likely only detect house-holding relationships that may or may not be familial. Analysts who are looking for business hierarchies within commercial customers, but can only access data from internal systems, are limited to identifying titles used by a customer and the level of each of their employees at a certain point in time. Their analysis could be enriched by adding public data sources, such as business or SEC filings, to confirm or deny titles, and help put actual hierarchies in place.
Let’s look at how Novetta Entity Analytics can leverage relationship knowledge to gain greater insights from data generated when an individual makes a purchase from a retailer. The transaction creates data about the items purchased, the customer, the location, the salesperson, the method of payment, and the relationships between these entities. Novetta Entity Analytics can analyze the data from many viewpoints, such as customer, salesperson, product, location or transaction, to provide many new insights about entities involved in the transaction. From a product perspective, the software might look at how often a specific product was sold over time at the specific location, or how many people bought the product there and their similarities and differences. From a customer perspective, the software could identify whether a customer most frequently interacts with the retailer through its website or the retail location, or if a website interaction led to the visit to the retail location.
Relationship Knowledge Leads to Better Insights, More Accurate Data
When analysts can gain a better understanding of all attributes, entities and potential relationships involved in events, they can ask more interesting questions of the data. In addition, relationships uncovered within the data can help identify entity resolution errors that should be corrected, such as detecting two extremely different transaction patterns from a resolved entity, which may indicate two real-world entities, instead of one, or highlight potential fraud.
Gaining an understanding of relationships between entities in multiple data sources requires a combination of human intelligence and iterative software that can learn from human input. Analysts must know what types of relationships they are looking for and the questions to ask to find them. Novetta Entity Analytics will intelligently and rapidly process large amounts of data to distill it into the answers they need.