MD101 Glossary

A generalised data model is a reusable, modular structure your data product uses to represent entities like users, assets, or transactions, regardless of source system. It replaces rigid, source-specific schemas with flexible, product-friendly models that scale across use cases and enable faster feature development.

Granular access controls let your data product define who can see or do what at a very detailed level, down to specific columns, records, or actions. It enables fine-tuned user experiences where security, compliance, and personalisation work together, without locking down innovation.

A graph database enables your data product to model relationships between data points like people, devices, or events, so that connections become first-class citizens. This powers smarter features such as recommendations, fraud detection, or network analysis by mapping how entities are linked in real time.

Growth metrics are the signals your data product tracks to show how usage, adoption, and business value are trending over time. These aren’t just vanity KPIs but are actionable insights that guide roadmap decisions, optimise features, and highlight what's truly driving impact.

Headless BI decouples the data layer from the presentation layer, so your data product can serve insights directly into user experiences without relying on traditional dashboards.This helps power personalised, contextual insights right where users need them, in apps, workflows, or notifications, without forcing them to “go look at the data.”

A hybrid cloud strategy is how your data product balances workloads across both on-premise and cloud environments, so it can meet performance, privacy, or compliance needs without locking users into a single infrastructure. This revolves around building flexibility into how data is processed and stored, so different use cases, from real-time analytics to archival access, can coexist seamlessly.

Identity resolution is the process your data product uses to stitch together different data points, like emails, device IDs, transactions, into a single, unified view of a user. This powers features like personalisation, customer segmentation, and consent management with greater precision and context, while avoiding duplicates and fragmented experiences.

Incident Management is the structured process for identifying, responding to, and resolving issues that disrupt data systems or pipelines. It ensures timely recovery, reduces impact on users, and helps prevent repeat failures by capturing learnings and improving system resilience.

Incremental data processing is how your data product updates only what has changed, rather than reprocessing everything, so it’s faster, cheaper, and more scalable. It’s the shift from traditional batch-heavy workflows to a more efficient, event-aware model where pipelines only touch new or modified data that maps directly to current feature needs.

Interactive Data Exploration is how users engage with data inside your data product; slicing, drilling, and filtering in real time, without writing queries or waiting on scheduled reports. It transforms dashboards from static snapshots into dynamic spaces for discovery, helping users get to insights faster and feel in control of their decisions.

Interoperability refers to the ability of your data product to seamlessly connect with other tools, systems, stacks, or data ,models, without requiring complex conversions or manual workarounds. This helps drive capabilites like federated search, cross-platform API integrations, and composable data experiences to run smoothly across teams and environments, while preserving consistency and control.

JSON Schema is a blueprint your data product uses to validate the structure and format of incoming data, so that only clean, predictable payloads enter your system. This enables reliable APIs, stronger contracts between teams, and fewer downstream failures, by making structure part of the interface, not just an assumption.

KPI standardisation ensures that everyone in your data product sees and uses metrics the same way, no matter the dashboard, domain, or team. From a product point of view, it reduces misalignment and builds trust by turning KPIs into governed, reusable components, often defined once and surfaced consistently across all tools.

A Knowledge Graph helps your data product organise complex relationships between entities, like products, people, or concepts, into a connected, queryable structure. This supports richer experiences like semantic search, relationship-aware recommendations, and decision intelligence that understands context, not just keywords.

Lakehouse Architecture (2.0) is the evolution of unified data platforms combining the flexibility of data lakes with the reliability/performance of data warehouses, while adding native support for governance, real-time processing, and data interoperability. It’s designed not just to store and query data efficiently, but to support the creation, discovery, and scaling of data products across teams. Lakehouse 2.0 treats data as a strategic asset: governed by design, modular by architecture, and product-ready by default.

Last-Mile Data Transformation refers to the final layer of data shaping that happens closest to the end user (often in dashboards, reports, or operational tools) to adapt standardised data for specific decisions, teams, or contexts. It bridges the gap between centralized data models and actual usage, enabling flexibility without compromising governance. This step makes data truly usable, relevant, and actionable at the moment it’s needed.

Latency optimisation is how your data product reduces the delay between action and feedback, so users experience fast, responsive interactions. Whether it’s live dashboards, real-time recommendations, or quick filter changes, optimised latency turns performance into a product feature, not just a backend metric.

Lineage Visualisation is the graphical representation of how data flows through a system: from source to transformation to consumption. It helps teams understand dependencies, trace errors, assess impact of changes, and build trust in data products. By making upstream and downstream relationships visible, it turns complex pipelines into transparent, navigable maps of how data moves and evolves.

A Low-Code Data Pipeline allows users to build, modify, and deploy data workflows with minimal hand-written code, often through user interfaces, drag-and-drop elements, or prebuilt logic bytes/rules. Low code pipelines imply quick iterations, enables access for a wider range of users who aren't pipeline-savvy, and reduces engineering overhead to support scalable and useful data movement. This approach helps teams deliver and evolve data products more quickly, even with limited technical resources.

MLOps governs how machine learning models are developed, deployed, monitored, and maintained. The objective is to enable teams to scale experiments without risk, track and optimise model performance, and ensure a stable and flexible ML lifecycle. Data products play a critical role by providing clean, versioned, and trustworthy data pipelines that models depend on, making ML systems reproducible, auditable, and easier to evolve.