Never before have we seen a faster or greater shift to online transactions as we have since the start of the pandemic. Overnight, most everyone was shopping, ordering and communicating online — creating digital footprints.
The data found in these footprints is valuable. It can inform better experiences and new processes that capture customers’ hearts and minds. The companies with their data ducks in a row have gained the most from this mass migration to digital channels.
For decades, organizations have strived to make sense of customer data. The dream of a 360-degree customer view isn’t new: it has always meant understanding “everything about the customer.” With increased digital transactions and interactions, more data, and new types of data, what “everything” means continually changes. Successfully meeting today’s definition of a customer 360 requires a variety of strategies to connect more data points and deliver context about customers and their relationships, their transactions and interactions.
Combining First- and Next-Generation Techniques for Data Matching
New data types have brought new opportunities to serve customers with increased relevance, better timing and personalized interactions that make transactions easier.
Collecting the data is challenging. But being unable to make sense of and match dynamic, duplicated and fragmented data across business silos is where most customer 360 efforts fail.
First generation matching techniques resolve identities to consolidate disparate customer records into unique customer profiles. Connecting transaction and interaction data with them maximizes the opportunity for delivering well-timed, relevant and informed interactions. Doing that requires a combination of traditional and next generation matching techniques.
If first-generation matching techniques resolve identities, next-generation techniques use artificial intelligence to connect sparse and difficult data types.
Related Article: Data Ingestion Best Practices
An Overview of First-Generation Data Matching Techniques
Whether it is matching people to people or matching organizations to organizations, the two most common first-generation techniques to create unique entities are rules-based and probabilistic.
A rules-based approach is simple enough to understand. This approach is used to compare attribute values of records and applies a set of conditions to determine when the records relate to the same entity. Through established rules, marketers can declare that John Q Adams and J Adams are a match, as long as the birth dates and addresses of two records are also the same.
The probabilistic approach assigns a weight when comparing attribute values and aggregates the results using a formula that yields an overall confidence score. If the aggregated score calculated is above a set threshold, it is then considered a match.
Here, marketers and data experts can assign more importance to certain attributes found within the data. They may give name and tax ID a high importance while more common attributes such as birthdate or address are provided with a lower importance.
When probabilistic matching algorithms are unable to satisfy all scenarios, those scenarios can be addressed through declarative rules.
What’s exciting is that probabilistic matching algorithms can also be learned through supervised training. In this scenario, data stewards and experts who best understand the data can review a subset of customer records pre-selected as a set of match pairs and then label them as either matches or non-matches. The labeled matched pairs form a training set that produces a matching algorithm. The combination of human matching and machine learning delivers greater accuracy and greater efficiency when working with larger and larger data sets. These techniques also set the stage for the next generation of matching.
Related Article: Why Match Rates Tell You Everything and Nothing at the Same Time
Adding Context to a 360-Degree View
Much of today’s data is complex. It is diverse, sparse and unstructured. Notwithstanding the constraints of regulatory compliance, privacy and customer consents, tapping into this data allows marketers to extract valuable insights about customers. Marketers can then use the data to derive or infer indicators of purchase or churn, sentiment, life events and more. But to be actionable and useful, these insights must be associated to a specific customer.
Through machine learning and natural language processing, next generation techniques tie the wealth of new data types to customer and prospect records. For marketers, the attributes in web chats, social media posts, service notes, and other types of unstructured or secondary data reveal a deep understanding of what is happening in, and around, each customer touchpoint. And, by including graph technologies, discovery of non-obvious relationships is supported.
Next generation techniques also use intelligence to derive “contextual attributes” from complex data such as, when did an interaction occur? What was it about? Where did it take place? Who was referenced in it? And what products and services were mentioned? This context is then associated with customers and stored as part of the 360-degree customer view.
Related Article: Sometimes a 270-Degree View of Customers Is Enough
Gold, Silver and Bronze Customer Records
If traditional matching creates a golden record of a customer, contextual attributes contribute to bronze and silver records that can be used by marketers for directional insights.
When you consider that not every business need requires 100% match accuracy, a confidence level can be assigned to match contextual attributes with the “most likely” customer (e.g. 80% confident it belongs with a certain customer). These “most likely” contextual attributes can then be applied for multiple views or “perspectives” of a customer for different business purposes.
For example, the team investigating fraud may need 100% confidence in the matching process, but marketing may be comfortable casting a wider net and utilize contextual or derived attributes with 80% confidence to develop targeted campaigns, find similar customers and prospects, or build personalized marketing offers.
Same Definition, Different 360
A solid data strategy and data management foundation has propelled many digital transformations and equipped teams to meet customers’ needs and lock in loyalty through contactless sales and service, demand forecasting, more targeted and personalized marketing, and stronger customer communities.
As transactions become more digital, the requirements of a 360-degree customer view have expanded. The matching techniques to create a 360-degree customer view have evolved to address those needs.
In creating this strategic data foundation, we gain agility to be responsive and move more quickly because we are now armed with new insights into customers’ life events, household, journey, influencer network and overall sentiment.
Monica Mullen is Solutions Marketing Manager at Informatica®, responsible for Informatica’s customer centricity and customer data strategy solutions. As an experienced tech marketer who’s lived in the data world since 1999, she believes that contextual, trusted data resides at the intersection of the familiar people—process—technology trio.