Reference model: BI product

Fourteen dimensions, six levels, observable criteria: what a BI product does and guarantees.

≈ 5 min readReading

At a glance

Families: value and usage, data, trust and operations, evolution and delivery.
Level 0: starting point; levels 1 to 5: criteria to meet.
A level is reached only if all its criteria are met.

On this page

Overview
The six levels
The four families of dimensions
Adapting the model

The model describes what a BI product does and guarantees (a decision-support tool or solution), across 14 dimensions grouped into four families. For each dimension, six levels (0 to 5); levels 1 to 5 are defined by observable criteria to be met (174 in total). To place your product, use the inventory tool; to understand the calculation, see the calculation rules.

How to read

Level 0 has no criteria: it is the starting point, a description. A level is reached only when all its criteria (and those of the lower levels) are in place. The criteria are verifiable facts, preferably by a demonstration or a document, not by an impression.

Overview#

Each cell is the signature of the level for the dimension.

Dimension	0 · Handcrafted	1 · Locally tooled	2 · Shared	3 · Industrialized	4 · Integrated and governed	5 · Autonomous
Usage	Dashboards built by hand in local files, used only by their author.	Local dashboards (Excel) shared by file or by email.	Centralized publication: one place, one version, identified users.	Defined path: reference reports, common navigation and filters, guided onboarding.	Supervised customization and creation: views per user or group, creation and sharing by users.	Assisted experience: the product suggests, alerts and guides without the user having to search.
Integration	None: results are copied by hand from one tool to another.	Manual exports (file, screenshot) to other tools.	Scheduled exports and sharing links.	Documented interfaces: standard connectors, API, single sign-on.	Embedded integration in business tools, with the user's context.	Bidirectional, event-driven integration: the product triggers actions and reacts to events.
Business processes	Informal use: decisions are made elsewhere, with no trace.	Decisions made from the dashboards are noted by hand.	Described processes: you know who decides what based on which indicator.	Simple workflows in the tool: thresholds, alerts, approvals.	Workflows customizable by business unit, role or group.	Business processes automated end to end, with human supervision by exception.
Data sources and flows	Files built by hand and kept locally.	Local data extracted from an ERP (or another management system).	Scheduled extraction to a common source.	Industrialized flows: recovery after failure, logging, volumes tracked.	Multiple sources consolidated in a common model.	Continuous flows and data contracts: the product adapts to changes in the sources.
Data quality	No checks: errors are noticed when they become visible.	Manual visual checks by the author.	Defined and documented check rules.	Automated checks and anomaly reports.	Quality measured by dimension, visible to users, with data owners.	Automatic corrections and prevention at the source.
Semantic model and indicators	Each file recalculates its own indicators.	Formulas copied from one file to another.	A shared model: common tables and relationships.	Indicators defined once, glossary published.	Governed semantic layer: reusable, versioned, multilingual.	Indicators “as code”: tested and deployed automatically.
Documentation and traceability	Tacit knowledge, held by the author.	Personal notes.	Single, summary, accessible documentation.	Documentation updated with every version, data dictionary.	Catalog and end-to-end traceability, glossary.	Documentation generated automatically from the model, checked at every delivery.
Access and permissions	The file is copied to whoever asks.	Files shared in a folder or password-protected.	Authenticated access, named accounts.	Groups and roles, rights managed centrally.	Row-level security, inherited scopes, periodic review.	Permissions driven by policy: automatic assignment, temporary rights, continuous audit.
Security and compliance	No specific measures: data lives on uncontrolled workstations or drives.	Sensitive data is identified.	Basic measures: encryption of exchanges, backups, updates.	Security by design: requirements, reviews, penetration tests.	Compliance built into deliveries: check before every version, access logged.	Automated controls in the delivery pipeline, continuous monitoring.
Availability and performance	Available when the author's workstation is.	Files on a workstation or a share, with no guarantee.	Shared server, backups, announced maintenance.	Availability objectives defined and measured, monitoring.	Sized capacity, high availability, managed costs.	Automatic adjustment: start, stop, scale up and down; self-healing.
Support and observability	The author troubleshoots, when available.	Users know whom to contact.	Single entry point and request tracking.	Support process, knowledge base, classified incidents.	Observability: logs, metrics and usage telemetry put to use.	Predictive detection: drifts are anticipated before the incident.
Industrialization and deployment	The file is modified in place then copied.	Manual copy of new versions of the file.	Numbered versions, documented manual deployment.	Build pipeline, separate environments, repeatable deployment.	Controlled version life cycle: multiple versions, migrations, documented deliveries.	Continuous, automatic deployment, with rollback.
Testing and fixes	Problems are noticed in use.	Occasional manual checks.	Test sets and checklists.	Automated regression tests.	Data, load and security tests; automated interface acceptance testing.	Tests triggered on every change, automatic fixes for known anomalies.
Project management and roadmap	Changes follow the demand of the moment.	Requests received by email or verbally.	A single backlog.	Structured planning and tracking (iterations, milestones).	Roadmap based on requests, product maturity and constraints; priorities arbitrated.	Value-driven management: measured usage and value feed the roadmap.

The six levels#

Level	Name	General signature
0	Handcrafted	Files and dashboards built by hand, locally, by a single person.
1	Locally tooled	Data extracted from a management system, local dashboards, shared by file.
2	Shared	A common source and a centralized publication: one version for everyone.
3	Industrialized	Flows, deployments, tests, access and monitoring managed in a repeatable way.
4	Integrated and governed	Multiple sources, customization by group, integration, measured quality, traceability, workflows.
5	Autonomous	Automatic deployment, tests and fixes; automated business processes; the product anticipates.

The four families of dimensions#

The fourteen dimensions are divided into four families, each detailed on its own page.

Value and usageUse, integration and business processes: the value and usage dimensions of a BI product.DataSources, quality, semantic model and documentation: the “data” dimensions of a BI product.Trust and operationsAccess, security, availability and support: the trust and operations dimensions of a BI product.Change and deliveryDeployment, testing and steering: the evolution and delivery dimensions of a BI product.

Adapting the model#

Remove a dimension that does not apply to your product: mark its criteria “not applicable” in the tool.
Add your own criteria: the principle is that a criterion be an observable fact and cumulative (it assumes the lower levels).
Change the order of your priorities with the tool's weights, without changing the levels.