Data analytics for Swiss companies: platforms, pipelines and reporting that decisions can rest on. The question first, then the platform.
Almost never on missing data. Usually on data that is scattered, inconsistently defined, and unresolved from a regulatory standpoint.
The monthly report is assembled by hand from four systems. It's two weeks old when it arrives, and only one person can produce it.
Sales and finance quote different revenue for the same month. The meeting is about the data source instead of the decision.
ERP, CRM, webshop and production don't talk to each other. Every cross-system analysis becomes its own project with manual exports.
Personal data is scattered across systems with no documented lineage and no deletion concept. Under revDSG and GDPR that is an open risk.
There are dashboards, but they answer no question anyone actually asks. Decisions are still made on experience.
Demand, capacity and failures could be forecast from the data you already have — but nobody has ever put the models into production.
Four focus areas that build on each other. Applied generative AI and agents are deliberately not here — you'll find them under AI Solutions.
The foundation without which every analysis stays an opinion. We bring the data from ERP, CRM and line-of-business systems together in one place — with documented lineage, so every figure can be traced back to its source.
Snowflake · BigQuery · Databricks · Airflow · dbt · Kafka · PostgreSQL
Reports that prepare a decision instead of displaying a number. We start from the questions your leadership actually asks and build the dashboard backwards from them — not forwards from whatever the tool offers.
Power BI · Looker · Tableau · Metabase · dbt Semantic Layer
Classic and predictive models on your structured data — forecasting, anomaly detection, classification. Deliberately the proven part of the discipline: measurable, explainable, and therefore auditable.
Python · scikit-learn · MLflow · Feature Store · Airflow
Who may use which data, for what — and for how long. Under revDSG and GDPR, that question decides what you may analyze at all. We answer it before the platform is built, not after.
Data Catalog · Collibra · OpenMetadata · dbt docs · Great Expectations
Data projects rarely fail on the technology. They fail when the end result is a dashboard that answers no question anyone actually asks. So we start from the decision.
We start from the decisions you want to make better — not from an inventory of your systems. That determines which data is needed at all.
Connecting the source systems and settling lineage, ownership and retention. The revDSG and GDPR frame is set here, not checked at the end.
Building pipelines, defining KPIs, and delivering the results where decisions happen — in the dashboard, in the business system, or as an API.
A dashboard nobody uses is not a delivery. We train the business units, measure usage, and adjust what doesn't hold up in practice.
Four things that are different once the data foundation is clean — in the meeting, in the close, and in planning.
KPIs are defined once and computed the same way everywhere. The meeting is about the decision again — not about whose analysis is right.
The monthly report builds itself instead of being assembled by hand. What used to be days of manual work is ready on the morning of the first working day.
Documented lineage, a deletion concept and access logs. You can prove which data is used for what — to customers, auditors and regulators.
Forecasting models for demand, capacity and failures run in production and are monitored — not as a one-off analysis that goes stale in three months.
We'll cut the first case together — one data source, one business unit, one decision. The value starts before the platform is complete.
Discuss a use caseTwo projects where scattered sources became a foundation you can analyze.
A unified clinical ecosystem: Meddilink EMR, HL7/FHIR device integration and a clinical data lake with analytics for European fertility markets.
Computing the shortest viable cable route between turbine and grid, routing around eight obstacle categories, with geospatial export.
In data work, governance decides what you may analyze at all. That's why it sits at the start, not in the acceptance review.
Platform and analytics run in Swiss or EU regions. For particularly sensitive holdings, operation inside your own environment is possible too.
Processing records, a deletion concept and test data anonymization are part of the delivery — not a later expansion phase.
Documented lineage from KPI back to source system. When a number is disputed, the derivation is verifiable instead of a matter of interpretation.
Open formats, versioned transformation logic in your repository. No model and no pipeline that only we can operate.
We're tied to no platform vendor. If your team already works with a tool, that is usually the right tool.
The points that come up in every procurement process — answered up front.
A clear boundary: this pillar covers the data foundation — data platform, pipelines, business intelligence, governance and classic predictive models on structured data. Applied generative AI, agents and LLM applications sit under AI Solutions. In practice the sequence is almost always: foundation first, then the generative use cases.
No. Assessing data quality is part of the first phase — and in most projects it's the very reason for the engagement. We capture the current state, name the gaps, and factor them into a realistic plan rather than handing them back to you as a prerequisite.
In Swiss or EU regions, at your direction; for particularly sensitive holdings, also inside your own environment. Data processing is governed under revDSG and GDPR and access is logged. Where FINMA requirements apply, we align the controls accordingly.
No. If the reporting tool is in place and used, the problem usually sits one level deeper — data quality, inconsistent KPI definitions, or manual steps in between. We start there and keep the tool your people can already operate.
We deliberately cut the work to a first use case that is typically usable in production after 8–12 weeks — one data source, one business unit, one decision. The full platform grows from there, but the value doesn't wait for the end.
