- Home
- Consulting
- Data and BI
Raw data turned into strategic decisions.
Modern data architecture, executive dashboards, predictive models and LGPD governance, delivered by a senior team from assessment to ongoing support.
We don't sell technology: we help you choose it. We design the architecture, build the pipelines and deliver dashboards and models that run in production, with governance from day one.
From source to insight, on a single platform
Four layers, source, ingestion, storage and consumption, with native integration and end-to-end observability.
- ERPOrders, tax and inventory
- CRMCustomers and opportunities
- E-commerceStore, APIs and events
- Ingestion and CDCFivetran and Airbyte
- PipelinesAirflow orchestration
- Data warehouseSnowflake or BigQuery
- Modelingdbt, tests and documentation
- Quality and LGPDLineage and observability
- BI and dashboardsPower BI or Looker
- Analytics and MLForecasting and scoring
- DecisionReliable KPIs by area
Reference architecture. The tools in each layer change according to the stack the company already uses.
What we deliver in data
Four fronts handled by the same squad, with an analytics engineer, a data engineer and a data scientist on the same team.
Data engineering
Reliable pipelines that run on their own, with orchestration, automated tests, observability and CI/CD. No more cron jobs lost on a virtual machine.
- Warehouse on Snowflake, BigQuery or Databricks, with dimensional modeling.
- ELT ingestion with Fivetran, Airbyte and change data capture (CDC).
- Transformations in dbt, versioned, tested and documented.
- Orchestration in Airflow and model deployment through CI/CD.
BI and dashboards
Power BI, Tableau or Looker with solid semantic modeling, row-level security and reliable refreshes. Real adoption, not just good-looking dashboards.
- Star schema modeling (Kimball) and optimized DAX measures.
- Row-level security (RLS) and workspaces by area.
- Incremental refresh and deployment pipelines across Dev, QA and Prod.
- Versioned LookML and analytics embedded in digital products.
Governance and quality
Real governance, not a policy in a PDF: catalog, lineage, data contracts and LGPD from day one.
- Data catalog and versioned business glossary.
- Automated lineage from source to dashboard.
- Dynamic masking, tokenization and data subject rights.
- Alerts on freshness, volume and schema changes.
Advanced analytics
From dashboard to automated decision: predictive models, generative AI and automations that run in production.
- Demand forecasting with time series and neural networks.
- Purchase propensity, churn and LTV per customer.
- Natural-language questions over the warehouse.
- MLOps with Vertex AI, SageMaker and MLflow.
Which tool makes sense for you
We are stack-agnostic. In the assessment, the choice comes from who will use it, analytical maturity and the total cost of operation, not from a fixed package.
How mature is your company's data?
Four stages, from spreadsheet reports to automated decisions. The assessment shows where the operation stands and what comes next.
Reactive
Spreadsheets and manual exports. Each area has its own number.
- Centralize data in a cloud warehouse.
- Replace critical spreadsheets with versioned dashboards.
- Inventory sources and personal data (LGPD).
Repeatable
Operational database and ad hoc reports that take days to produce.
- Versioned, tested transformations in dbt.
- Data catalog with a business glossary.
- Freshness and quality targets per pipeline.
Managed
Centralized warehouse and metrics aligned across areas.
- Data observability.
- A structured analytics engineering team.
- First ML models operated with MLOps.
Optimized
A single source of truth, with lineage and continuous auditing.
- Generative AI and advanced self-service.
- Data products and mesh architecture.
- Data served to business areas through internal APIs.
Data maturity assessment
Five questions, about two minutes. At the end, you see your operation's stage and three next steps to move forward.
Your answers stay in your browser. Nothing is sent until you ask for the assessment by email.
Question 1 of 5
Four steps to a data-driven operation
- 1
Discovery and assessment
We map sources, KPIs, gaps and business goals and deliver the roadmap.
- 2
Architecture and MVP
We design the stack and put the first use case into production.
- 3
Scale
We add new sources, dashboards, models and governance.
- 4
Ongoing support
A team dedicated to evolution, with observability and continuous monitoring.
Frequently asked questions
Where does a data project with CCX start?
With a maturity assessment: we map sources, KPIs by area, bottlenecks and goals. That produces the roadmap and the first use case, which goes into production before any expansion.
Which BI tool do you work with?
Power BI, Tableau, Looker and Metabase. We are stack-agnostic: the choice depends on who consumes the data, analytical maturity and total cost, not on a fixed package.
Do we need a data warehouse?
If your numbers live in spreadsheets and operational databases, centralizing them in a cloud warehouse (Snowflake, BigQuery or Databricks) is usually the first step. It gives dashboards and models a single source of truth.
How does LGPD fit into the project?
From day one: personal data mapping, dynamic masking and tokenization, data subject rights, auditable legal bases and access logs with configurable retention.
Can we use AI on our data?
Yes. With a validated semantic layer, you can ask the warehouse questions in natural language, and run demand forecasting, churn and LTV models in production.
Does the knowledge stay with our team?
It does. Every deliverable is documented in dbt docs, with a searchable catalog and role-based training: executive, analyst and technical.
Does CCX stay on after delivery?
In the ongoing support stage, a dedicated team handles platform evolution, pipeline observability and continuous monitoring.
Enterprise data and BI guide
The modern stack for real-time decisions: architecture, warehouse, BI, governance and applied AI, the way CCX implements them.
Modern architecture: lakehouse, ELT and Medallion layers
The reference architecture combines Bronze (raw data), Silver (cleaned and conformed) and Gold (business-ready) layers on top of a lakehouse or warehouse. This separates responsibilities, allows reprocessing without duplication and provides end-to-end traceability.
Ingestion is ELT, not ETL: Fivetran for SaaS connectors, Airbyte for custom connectors and CDC for transactional databases. Transformations live in dbt, with tests, documentation and CI/CD, and observability alerts the team before a problem reaches the dashboard.
Snowflake, BigQuery, Redshift or Databricks
Snowflake separates compute and storage and runs on AWS, Azure and GCP. BigQuery is serverless and the natural choice for teams already on Google Cloud. Redshift makes sense for operations heavily invested in AWS. Databricks is the choice when ML and Spark engineering become the core of the product.
In most projects, the first stack is Snowflake, dbt and Fivetran with the BI tool the team already uses, evolving to Databricks when ML becomes central.
Power BI, Tableau, Looker or Metabase
Power BI fits Microsoft operations with many casual consumers. Tableau wins at deep visual exploration. Looker is BI as code, with a rigorous semantic layer and embedded analytics. Metabase serves lean teams that want BI without red tape.
CCX implements all four and helps you choose based on user profile, analytical maturity and total cost.
Governance, LGPD and data culture
Most projects fail for lack of ownership and process, not technology. The framework is lean: data owners in the business, stewards on the data team, a catalog, data contracts between producers and consumers, and a committee that approves changes to the Gold layer.
For LGPD: personal data classification, policy-based masking in the warehouse, pseudonymization and the right to be forgotten with an audit trail. Data culture is part of the scope, with documentation the business can actually use and role-based training.
Applied AI: from the warehouse to an agent in production
The data stack feeds three types of AI: predictive analytics (churn, LTV, forecasting), production recommendations through a feature store, and generative AI connected to the warehouse, with agents that answer business questions in natural language.
The semantic layer is a prerequisite: without it, the agent makes up metrics. That is why we recommend starting with natural-language questions over the semantic layer before moving on to complex agents.
Talk to a data architect.
Thirty minutes with a CCX engineer, at no cost, to map your current stack, its bottlenecks and the path to a data-driven operation.