MD101 Glossary

Master Data Management (MDM) is how your data product keeps the core entities, like customers, products, or locations, consistent, accurate, and reusable across systems. In a data product context, MDM shifts from being a centralised, IT-driven initiative to a modular, API-first service that supports trusted data use across evolving workflows and features.

Model Drift occurs when a machine learning model's performance degrades over time due to changes in input data or real-world conditions.

Near Real-Time Processing refers to the rapid handling of data with minimal delay, not instantly though but fast enough for users or systems to react quickly. It strikes a fine balance between speed and efficiency, useful in use cases like fraud detection or user activity tracking.

A Non-Relational Data Model organises data in flexible formats like key-value pairs, documents, graphs, or wide-columns, allowing teams to structure data around how it’s actually used. It’s ideal for applications that require speed, scale, and adaptability to varied or evolving data, such as content platforms, recommendation systems, or IoT solutions. This model prioritizes performance and fit over rigid structure, making it easier to build responsive, user-aligned data products.

An Observability Dashboard visualises the health, performance, and reliability of data systems in real time. It helps teams spot issues, track metrics, and stay ahead of potential failures.

On-Demand Data Processing runs data tasks only when requested, rather than on a fixed schedule. It allows teams to run resource-intensive jobs, generate custom reports, or refresh specific outputs as needed.

Operational Analytics uses current data from core systems to monitor and optimise day-to-day business operations. It powers informed actions in areas like inventory, logistics, or customer support.

PII Handling is the process of managing, storing, and transmitting sensitive personal data in compliance with privacy laws and user expectations. It includes practices like minimisation, encryption, masking, and auditability, protecting individuals and maintaining trust.

Predictive Analytics uses historical data and statistical models to forecast future outcomes. It supports proactive decisions by giving users early signals based on data patterns.

Query Federation allows users to run a single query across multiple, diverse data sources, without needing to move or duplicate data. It enables unified access to distributed systems (like cloud warehouses, databases, and lakes), so data products can pull in the most current information from where it lives. This reduces complexity while not compromising on governance across the stack.

Query Optimisation improves how queries are executed so they run faster and consume fewer resources. Despite natural language querying, backend optimisation is essential to improve context and processing for both machine and human users. Optimisation is crucial to ensure data products remain responsive and scalable, especially as adoption grows. The outcome is more reliable user experience with accurate or fresh results.

Real-Time Data Processing is the continuous ingestion, transformation, and delivery of data the moment it’s generated, so it can power decisions, alerts, and experiences without delay. It enables data products to respond to events as they happen (whether updating dashboards, triggering workflows, or personalising user interactions) making the product feel immediate, intelligent, and trustworthy.

A Recommendation Engine suggests relevant content, products, or actions to users based on their preferences, behaviour, or context.

Role-Based Access Control or RBAC is a security model where access permissions are granted based on a user’s role within an organisation. This simplifies access management and aligns security with organisational structure and responsibilities.

SQL Exploration is the practice of using SQL to query, slice, and examine data directly. It enables fast, flexible access to raw or modeled data without needing a prebuilt dashboard.

Schema Evolution is the ability of a data product to adapt its structure such as adding, removing, or modifying fields, without breaking downstream systems or user expectations. It supports agility in how data changes over time, enabling teams to respond to new requirements while preserving continuity, trust, and interoperability across the product’s lifecycle.

A Service Level Agreement (SLA) is a formalised contract that sets the expected standards of service between a provider and a consumer. It defines what users can count on (such as uptime, performance, or support response times) and helps build trust through clear accountability. Within an SLA, a Service Level Objective (SLO) is the specific, measurable target the system aims to meet, like 99.9% availability or a 2-hour response window.

A Service Level Indicator or SLI is a measurable metric like the uptime, latency, or error rate. It reflects the actual performance of a service.

A Service Level Objective (SLO) is a specific, measurable target for the level of service a system should provide. Examples can be uptime, response time, or error rate. It helps teams prioritise reliability and performance, giving both users and engineers a clear standard to aim for and track over time.

Smart Data Caching stores frequently used or expensive-to-compute data in a way that accelerates access without redundant processing. It ensures a fast user experience while managing resource overhead.