CDP Identity Resolution: A Critical Step to Build a Unified Customer Profile
Published On -
The Unified Customer Profile is a core function of Quaero’s CDP solution – a fully comprehensive matching process that goes beyond standard identity resolution methods based on deterministic stitching, and provides the agility in addressing data complexity and nuance for each client.
Behind the Unified Customer Profile is a process called “Signal Matching”; which ingests and synthesizes the following data signals:
- Personally Identifiable information (PII)
- Cookies (first and third party)
- IP addresses
- Device graphs (multiple)
- User agents
- Device IDs
- Account/customer IDs
We assimilate and apply our algorithms to these data signals in the following way:
- Data Hygiene: Cleansing and preparing a standardized data set to avoid false positives
- Deterministic Matching: Exact matches using email address, 1st party cookie, account numbers, and source IDs
- Probabilistic Matching: Creating probable matches using proximity, scores, thresholds, Device IDs, and fingerprinting
- Multi-pass Matching: Matching on a combination of signals such as email + Device ID, email and 1st party cookie to feed into algorithms, sort on composition keys, and run them in parallel to get maximum linkage within and across data sources. This step also handles configurations of suspicious match keys to avoid people using common devices in a household and pruning recency to improve matching performance
- Transitive Closure: Stitching results together from step #4 match passes to achieve “transitive closure;” also known as “chaining”
Most MDM solutions do a good job in solving identity problems for offline data that handles PII signals (name, address, phone and email), but those solutions don’t work as well with digital data or linking offline to digital consumers. In addition, very often these solutions lack data signals such as device graphs to improve matching accuracy.
The following steps and approach provide a guide to enhanced identity solutions to support various marketing use cases.
It is important to note that no one solution provides 100% match accuracy. However, the incremental lift in match accuracy provides sufficient added return in the above use cases to justify the investment. This is the appropriate metric to use rather than perfect match accuracy. Increased Match accuracy leads to 1) higher levels of enriched user profiles that a brand can then design and enable personalization for, and 2) increased ‘addressability’, allowing brands the means in which to deliver those personalized experiences across various media and channels
In order for an identity solution to work well, it is imperative to have a comprehensive data strategy and data collection process across systems to ensure critical data elements are captured at the source. Algorithms are data hungry, so the more data and better data they are fed, the better and more accurate the outcomes.
Want to learn more? Please contact us anytime for a demo.