Business Intelligence Enterprise turns Dolibarr into a full enterprise decision-support platform — an integrated alternative to Microsoft Power BI, Tableau, Qlik Sense, SAP Analytics Cloud, IBM Cognos, Oracle Analytics, Looker, Metabase, Apache Superset, Grafana, Pentaho and Sisense — without leaving the ERP.

The module covers the entire Business Intelligence lifecycle:

Collection → Centralisation → Transformation (ETL) → Modeling → Multidimensional analysis (OLAP) → Dashboarding → Data visualization → Reporting → Forecasting → Alerting → Decision AI.

The executive cockpit above aggregates the whole platform on a single page: 16 premium KPI cards, RAG target-achievement bullets per business domain, revenue trend against target, forecasts, active alerts, AI insights, ESG indicators, a 12-month activity heatmap, geographic breakdown, ETL pipeline health and an audited event timeline.

1.1 Steering scope

·       Strategic steering — revenue, result, margin, profitability, forecasts, strategic KPIs.

·       Operational steering — sales, purchasing, inventory, production, logistics, quality, maintenance, projects.

·       Financial steering — cash, accounting, receivables, payables, cash flow, budget.

·       Sales & CRM analysis — targets, conversion, pipeline, customer 360, rep performance.

·       HR analysis — headcount, leave, recruitment, training, performance.

·       ESG / CSR analysis — CO2 scopes 1/2/3, energy, water, waste, safety, diversity.

 

The module follows native Dolibarr architecture end to end: CommonObject-based business classes, a descriptor-driven module definition, standard permissions, MultiCompany/MultiEntity awareness through getEntity(), and the native translation system.

2.1 Layers

 

2.2 Physical model

·       25 tables prefixed llx_bie_*, each carrying entity, status, date_creation, tms, fk_user_creat, fk_user_modif and import_key.

·       Every foreign key is indexed; index names are kept short to stay under the 64-character limit.

·       A schema self-heal (bieEnsureSchema) runs at activation and ALTERs in any column added by a later version — Dolibarr only ever runs CREATE TABLE, never ALTER.

·       table_element carries no llx_ prefix; the SQL files keep the literal llx_bie_* names.

 

3.1 Data sources

Native: Dolibarr database, MultiCompany, MultiEntity. External (architecture in place): MySQL, MariaDB, PostgreSQL, SQL Server, Oracle, CSV, Excel, REST API, JSON, XML, Google Sheets. Each source declares its connection, authentication, refresh mode (real-time, hourly, daily, weekly, monthly, manual), cron expression, sync history and owner.

3.2 Data Warehouse and ETL

The warehouse layer persists fact tables, dimension tables and aggregates with grain, partition key, retention, size, index count and incremental mode. The ETL engine runs Extract → Transform → Load with ordered, typed steps (cleansing, deduplication, mapping, normalisation, calculations) and tracks rows in / out / rejected, duration, run count, error count and last-run status.

 

3.3 Semantic modeling

Star and snowflake models relate data sets through typed relations (cardinality, join type). Dimensions carry drill hierarchies (e.g. Year > Quarter > Month > Day); measures carry aggregation, formula, unit, format and additivity.

 

The builder composes dashboards by drag & drop on a 12-column grid. Widgets can be resized, duplicated, deleted and bound to a KPI; the layout is persisted as bie_widget rows.

4.1 Widget catalogue (22 types)

KPI Card · Gauge · Thermometer · Donut · Pie · Bar · Line · Area · Radar · Waterfall · Funnel · Sankey · Treemap · Heatmap · Scatter · Bubble · Geographic map · Timeline · Data table · Pivot table · Calendar · LED indicator.

4.2 Chart catalogue (20 families)

Line · Bar · Donut · Radar · Treemap · Sankey · Waterfall · Funnel · Heatmap · Scatter · Bubble · Analytical Gantt · KPI card · Bullet graph · Pareto · Sunburst · Gauge · Timeline · World map · Regional map. All charts are interactive, zoomable, filterable and exportable.

Rendering is native by default (no external dependency). The architecture also accepts Apache ECharts, Chart.js, Highcharts and D3.js, selectable in the setup page.

 

The module ships a library of 284 ready-to-use indicators across 12 business domains. Each KPI carries a name, description, category, formula, unit, target, minimum and maximum thresholds, owner, calculation frequency, current and previous value, variation, achievement rate and an improvement direction that drives all RAG colouring.

 

 

The OLAP page builds a cube from the KPI value history joined to the KPI dimensions, and exposes the five canonical operations.

 

Cells are shaded proportionally to their value, so the pivot reads as a heat map; row totals, column totals and the grand total are computed in the same single grouped query. The whole table exports to CSV exactly as displayed.

Security: row/column dimensions, measures and aggregations are resolved through a server-side whitelist — no user input ever reaches the SQL string.

 

The forecasting engine projects any indicator forward over a configurable horizon.

 

Forecast targets include revenue, sales, cash, orders, expenses, margin, production, payroll, purchases and stock value.

 

The decision AI explains indicators, detects anomalies, identifies trends, compares periods, summarises performance and suggests actions — for example answering "Why did sales drop this month?".

8.1 Local statistical engine

The analysis is computed by a deterministic statistical engine — least-squares linear trend plus Z-score anomaly detection — so the page is fully functional with NO external AI provider configured. Every finding and every recommendation is derived from a figure the module actually computes and displays: mean, standard deviation, Z-score, R squared, period-over-period variation and target achievement. The module never states a number it cannot establish.

8.2 Provider architecture

When a provider is configured (OpenAI, Ollama local, Azure OpenAI, Anthropic, Mistral), the same computed findings are what would be handed to it as grounding context. Stored insights record provider, model, question, finding, recommendation, confidence, estimated impact and token usage.

 

The reporting page presents the KPI scorecard and drives the report engine.

·       Scorecard — reference, label, category, current value, target, variation, RAG achievement bar and an inline SVG sparkline per indicator.

·       Category filter — instantly yields the dashboard of a given direction (finance, sales, HR, inventory, production, ESG …).

·       CSV export of the whole scorecard, formula included.

·       Report designer — custom reports, templates, filters, parameters, calculations, charts and groupings.

·       Output formats — PDF, Excel, CSV, Word, PowerPoint, HTML.

·       Scheduling — automatic delivery (daily, weekly, monthly, quarterly) to a recipient list.

Performance note: the scorecard loads the full KPI history in a single grouped query and buckets it in memory, so 284 sparklines cost one query rather than 284.

 

The BI API exposes the platform to external consumers.

 

10.1 Dolibarr integration

The module reads the native Dolibarr modules through its data sets: CRM, third parties, products, stock, warehouses, purchasing, sales, quotations, orders, invoices, accounting, bank, HR, projects, interventions, production, and integrates alongside the DoliResources GMAO, Facility Management, QHSE, ESG, OCR, e-signature, API Manager, Workflow, GED and AI Assistant modules.

 

 

 

 

12.1 Engineering standards

·       Object-oriented, CommonObject-derived classes with fetch/fetchAll/create/update/delete and getNomUrl.

·       MultiCompany / MultiEntity: every query filters through getEntity().

·       Multi-language via the native Dolibarr translation system; the module re-asserts its own labels at render time so that, on an instance running many modules, a shared translation key can never surface another module's wording on a BI screen.

·       Multi-currency through the native price() formatter.

·       Responsive design; CSS served through a module_parts stylesheet.

·       Intelligent caching: configurable dataset cache TTL.

·       SQL optimisation: grouped single-pass queries for the pivot, the heatmap and the scorecard sparklines (no N+1).

·       Interactive JavaScript is guarded against double-binding (Dolibarr may inject module JS more than once).

·       Documented code: file, class and method docblocks throughout.

12.2 Extensibility

The architecture is designed to accept future Big Data connectors, external Data Lake / Data Warehouse targets, advanced machine learning and third-party BI platform integration (including Power BI export), without disturbing the existing model.

 

© 2026 DoliResources — www.doliresources.com