Methodology

A score without context is difficult to trust.

More data does not automatically create better intelligence.

AtData has spent 25+ years building an email-centered view of identity: collecting signals from across a broad data network, connecting them to persistent identities, observing how those identities change over time, and converting that history into intelligence organizations can use.

Our methodology is built around a simple principle:

Observe what can be known.

Connect what belongs together.

Derive what the evidence supports.

Make the output usable.

---

Start with a persistent identity anchor.

Every intelligence system needs something stable enough to connect observations across time. For AtData, that anchor is email.

Email persists across devices, accounts, transactions, registrations, customer systems, and long periods of a person’s digital life. Unlike a session or device identifier, it is frequently supplied directly as part of an interaction and continues to appear as environments change.

AtData uses that persistence to build relationships among email addresses and supported identity attributes such as names and postal addresses.

The result goes beyond a database of emails. It is a longitudinal identity foundation designed to answer a more useful question:
What do we know about this identity, its relationships, and its behavior over time?

---

How AtData turns data into intelligence.

Source – Begin with observable data from multiple environments.

No single source provides a complete view of identity. AtData incorporates data and signals from a broad partner and data ecosystem, including identity records, email and web activity, demographic and household sources, and behavioral and purchase-related data.

Source diversity matters because intelligence based on a single organization’s interactions can describe only what happened inside that environment.

AtData’s broader network provides context beyond that first-party view.

Standardize – Make disparate data comparable before making it meaningful.

Raw data arrives in different structures, formats, and levels of completeness.

AtData standardizes and normalizes identity information before it is used to create linkages, signals, or model inputs. That process helps resolve differences in how the same underlying identity may appear across contributing records.

This is foundational work, but it matters. If identity inputs are poorly standardized, every downstream match, signal, score, and model inherits that uncertainty.

Better intelligence starts with better inputs.

Resolve – Connect evidence to the right identity.

AtData uses its identity graph, matching logic, algorithms, and machine learning to connect data points with applicable email, name, postal, and other supported identity relationships.

That resolution creates longitudinal context. Instead of evaluating each appearance of an email as an isolated event, AtData can examine the history and relationships accumulated around the identifier.

Across the AtData environment, this foundation includes approximately 2.4 billion email identity linkages.

A larger graph is not inherently a better graph. The objective is to connect the right information to the right identity.

Observe – Let behavior accumulate meaning over time.

A static attribute describes an identity at a moment. Behavior reveals how that identity changes. AtData continuously receives activity signals that provide additional evidence about email identities across time. The network currently processes more than 150 billion deterministic activity signals each month.

Those observations support measures such as:

• Date First Seen When AtData first observed the email identity.
• Longevity How long the identity has been present in AtData’s observed history.
• Velocity The relative amount of activity observed for an email over a defined period.
• Popularity The breadth of sources across which the email has been observed.
• Recent Activity Whether AtData has observed more recent signs of use or engagement.

These signals turn identity from a static record into a history.

Derive – Convert observations into signals without hiding the evidence.

Raw observations are not always the form an enterprise system needs. AtData transforms underlying identity and behavioral information into structured signals, attributes, and scores that can be incorporated into business decisions.

Some outputs are relatively direct:

• Date first seen
• Last observed activity
• Validation status
• Identity linkages
• Demographic attributes

Others are derived:

• Velocity
• Popularity
• Quality scores
• Risk scores
• Behavioral classifications
• Model outputs

A derived score summarizes evidence. It should not be mistaken for the underlying evidence itself.

Where appropriate, AtData provides both individual signals and composite scoring so organizations can determine how much abstraction fits their own decision strategy. AtData’s fraud products, for example, expose underlying behavioral metrics alongside model-driven risk intelligence.

Make intelligence actionable without making it unnecessarily opaque.

Model – Use AI/ML where combination creates more value than any single signal.

Some decisions depend on patterns too complex for an individual attribute to capture. AtData uses machine-learning models to combine multiple observations into decision-oriented outputs.

Quality Score, for example, evaluates billions of behavioral signals across hundreds of sources and considers factors including frequency and type of activity, source diversity, and overall digital footprint.

Fraud intelligence similarly combines historical, behavioral, metadata, and identity signals to help quantify risk.

The purpose of modeling is not to replace customer decision logic. It is to make a large body of identity evidence more useful inside it.

---

Facts, signals, and scores are not the same

This distinction remains visible throughout AtData products and documentation. Keeping these layers distinct gives data scientists, fraud teams, marketers, product owners, and governance teams greater control over how AtData intelligence is interpreted and applied.

Observed data

Information directly associated with an identity or activity AtData has received or observed.
Example: Date first seen.

Derived signal

A structured measure calculated from multiple observations.
Example: Velocity or popularity.

Model score

An output produced by evaluating a combination of relevant features.
Example: Quality Score or Risk Score.

---

Built to remain useful as identity changes.

Identity intelligence has a shelf life. An email that was active last year may be dormant today. A new identity accumulates history. Relationships change. Fraud strategies adapt. Customer behavior moves.

AtData therefore combines historical depth with continuously updated activity rather than treating identity as a static profile.

That combination matters:
History provides context.
Current activity provides relevance.
Relationships provide continuity.

Together, they make it possible to evaluate not only what an identity looks like, but how it has evolved.

---

Designed for decisions, not just data delivery

AtData intelligence can be delivered as individual attributes, signals, scores, linked identities, or larger datasets depending on the use case. Organizations can use those outputs through:

• Real-time APIs
• Batch processing
• Licensed data
• Supported integrations
• Proprietary models and decision engines

Our customers retains control over the ultimate decision.

A fraud team may weight email age differently from velocity. A marketing model may combine activity with purchase propensity. An identity graph may prioritize current linkages differently from historical ones.

AtData provides the evidence and intelligence layer. The organization determines how that intelligence should affect its customers and business.

---

Methodology includes stewardship.

Data quality is only one dimension of responsible identity intelligence.

AtData’s methodology is supported by documented security, privacy, and data-governance practices. Our information-security program is ISO 27001 certified and subject to internal and independent external audits.

AtData also supports privacy-conscious implementations, including hashed-data options for applicable use cases, and maintains defined compliance boundaries around our data products.

Enterprise teams evaluating specific data provenance, permitted uses, retention, security, or compliance requirements should review the policies relevant to their implementation with AtData.

---

The advantage is not one signal. It is the system behind them.

No individual email attribute can explain an identity.

The differentiation comes from combining persistent identity, deterministic relationships, historical depth, continuous activity, signal engineering, and modeling into an intelligence layer that can be interrogated at the level a decision requires.