Modern BI: tools, maturity & the semantic layer
Business IntelligenceBusiness IntelligenceTechnologies and processes that turn raw data into actionable insights via reporting, dashboards and analysis, so teams can decide based on facts rather than intuition.View full definition → has been around for thirty years. But the way organizations do BI has changed fundamentally in the last decade, from IT-controlled report factories to self-serve analytics platforms accessible to every business user.
Understanding this evolution, and where your organization sits on the maturity curve, is foundational to making the right BI architecture decisions.
The BI maturity journey
Stage 1, Reactive reporting: IT generates reports on request. Analysts wait days for data. Decision-making is slow because data access is a bottleneck. Most organizations lived here until the early 2010s.
Stage 2, Descriptive analytics: A data team builds dashboards. Business users can view pre-built reports. They can see what happened, but can't explore why.
Stage 3, Self-serve BI: Business users can explore data themselves without writing SQL. Tools like Tableau, Power BI, and Looker enable drag-and-drop analysis. The data team provides the foundation; business users build their own views.
Stage 4, Embedded analytics: BI is integrated directly into operational tools (CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.View full definition →, ERPERPA single integrated software backbone that runs core operations: finance, procurement, supply chain, HR and manufacturing on shared data.View full definition →, product). Users don't go to a separate dashboard, insights appear where decisions are made.
Stage 5, Augmented analytics: AI assists analysis. Natural language queries, automated insight generation, anomaly detection surfaced proactively. This is now real in most major tools: Power BI has Copilot, Tableau has Tableau Pulse and Agent, ThoughtSpot has Spotter.
Most mid-size enterprises are between Stage 2 and 3. Large data-mature organizations are pushing toward Stage 4 and 5.
Modern Business Intelligence Architecture
Knowledge check
1. What fundamentally distinguishes Stage 3 (Self-serve BI) from Stage 2 (Descriptive analytics) in the BI maturity journey?
2. What is the primary purpose of Looker's semantic layer (LookML)?
3. A mid-size company running entirely on Microsoft 365 and Azure wants to adopt a BI tool with strong value for money. Which choice best fits, and why?
4. Select ALL statements that correctly describe the shift toward modern BI over the last decade.
Select all the correct answers.
5. Select ALL characteristics that correctly describe Stage 5 (Augmented analytics).
Select all the correct answers.
The modern BI tool landscape
Tableau, the longtime market leader in data visualization. Exceptional visualization capabilities, large user community, strong for power users. Acquired by Salesforce in 2019. Now bundled with AI features (Tableau Pulse, Tableau Agent). Pricing: roughly $75+/user/month for Creator licenses.
Power BI, Microsoft's offering, now part of the Microsoft Fabric platform. Deep integration with Microsoft 365 and Azure. Best value for Microsoft-heavy organizations. Power BI Pro is around $14/user/month; Premium capacity is priced separately. Market leader by volume of users.
Looker (Google Cloud), model-based approach. LookML defines business logic centrally; all reports use this semantic layersemantic layerA shared business definitions layer that sits between raw data and reports, so everyone uses the same numbers for revenue, churn or margin.View full definition →, ensuring consistent metrics. Google merged its BI products under the Looker brand, and the old Data Studio is now Looker Studio (a separate, free, ad-hoc tool that does not use LookML). Premium pricing for the core Looker platform.
Apache Superset, open-source BI platform. Free, extensible, used by Airbnb, Lyft, and many data-mature organizations. Requires engineering investment to operate. Preset offers a managed version.
Metabase, simpler, more accessible self-serve BI. Good for smaller teams and less technical users. Free open-source version available alongside paid cloud plans.
ThoughtSpot, search-driven analytics with its Spotter AI assistant. Natural language queries against your data. Strong for executive self-serve.
Choosing the right BI tool
The tool choice is less important than the architectural decisions around it:
Semantic layer strategy: Where does business logic live? If it's in individual report definitions, you get metric fragmentation, every dashboard calculates "revenue" slightly differently. A centralized semantic layer (Looker's LookML, the dbt Semantic Layer, or Cube) defines business logic once and exposes it everywhere.
Governance model: Who can create datasets? Who can publish to production? Who can access sensitive data? Without governance, self-serve becomes a governance nightmare.
Performance architecture: BI tools can be fast or slow. The difference is usually not the tool, it's query optimization, caching strategy, and data model design.
The semantic layer: the missing piece
The semantic layer is the translation between technical data structures and business concepts. It defines what "revenue" means, how "active customer" is calculated, what "conversion rate" is in each context.
Without a semantic layer, business logic is duplicated across dozens of dashboards. When the definition of "revenue" changes, it changes in 47 different places, or more commonly, it only changes in some places, creating inconsistent reporting.
With a semantic layer, the definition is single-source. Change it once, every downstream report reflects the change. The main options today: LookML (inside Looker), the dbt Semantic Layer powered by MetricFlow (dbt Labs deprecated the older "dbt Metrics" package and replaced it with MetricFlow after acquiring Transform in 2023), and Cube. A newer trend worth watching is universal semantic layers that BI tools query directly, so the same metric definition feeds Tableau, Power BI, and a notebook.
This is one of the highest-leverage investments a data team can make, and one of the most frequently skipped.
Quiz Questions
- Qu'est-ce que la "couche sémantique" (semantic layer) apporte principalement à une architecture BI ?
A) Elle accélère l'exécution des requêtes SQLSQLSales Qualified Lead: a prospect the sales team has validated as ready for direct outreach and a proposal, having passed clear qualification criteria.View full definition →
B) Elle fournit un lieu unique de définition des métriques métier, garantissant la cohérence à travers tous les rapports
C) Elle remplace la nécessité d'un data warehousedata warehouseA central repository that consolidates data from many source systems into a structured, query-optimized store designed for analytics, reporting, and business intelligence.View full definition →
D) Elle sécurise l'accès aux données
Réponse: B
- Quelle est la principale différence entre Looker et Tableau dans leur approche du BI ?
A) Looker est gratuit, Tableau est payant
B) Looker utilise LookML pour définir la logique métier centralement, Tableau est orienté visualisation avec des définitions par rapport
C) Tableau supporte plus de sources de données
D) Looker est uniquement pour les données Google
Réponse: B
- À quel stade de maturité BI se situe une organisation où les utilisateurs métier peuvent explorer les données eux-mêmes sans SQL ?
A) Stage 1
B) Stage 2
C) Stage 3
D) Stage 5
Réponse: C
Key Takeaways
- BI maturity moves from IT-controlled reporting toward self-serve, embedded, and AI-augmented analytics. Most mid-size firms sit between Stage 2 and 3.
- The main tools have shifted: Power BI is now part of Microsoft Fabric (Pro around $14/user/month), Tableau ships AI features like Pulse and Agent, and Google's old Data Studio is now Looker Studio (distinct from LookML-based Looker).
- Augmented analytics is no longer just "emerging": Copilot, Tableau Pulse, and ThoughtSpot Spotter put natural language querying in production tools.
- The semantic layer is the highest-leverage, most-skipped decision. Current options are LookML, the dbt Semantic Layer (MetricFlow, which replaced the deprecated dbt Metrics), and Cube.
- Pick the tool second. Get semantic definitions, governance, and query performance right first.
What to do, from this lesson
These actions are compiled in the role's Playbook.
- Deploy a single semantic layer where every tool resolves metric definitions
Related articles
Recent articles from the blog that build on this lesson.
- DataAirbus lost a quarter to misaligned revenue metrics: here is the playbook that prevents itWhen finance, sales, and product each calculate "revenue" differently, the damage shows up in board decks, budget fights, and delayed decisions. This playbook walks CDOs through building a semantic layer that makes metric definitions a shared organizational fact, not a tribal negotiation.
- DataDuckDB just made the analyst's job description obsolete, here is what CDOs should do about itNatural-language BI is collapsing the distance between a business question and a working query. CDOs who treat this as a tooling upgrade rather than a workforce redesign will find themselves managing a team built for problems that no longer exist.
- DataNatural-language BI and the analyst's new role: why the "democratisation" story is only half trueNatural-language query tools promise to put business intelligence in everyone's hands, removing the analyst bottleneck. The reality is more complicated, and CDOs who act on the simple version of this story will make costly structural mistakes.