Value driver networks · KPI ontologies · Causal analysis
Value driver networks, tested against your data.
Cosmos Group helps companies build and maintain value driver networks – the cause-and-effect links from operational drivers to financial KPIs – and the KPI ontologies that define them. Language models draft candidate drivers and KPI definitions. Causal analysis of your own data shows which links the data supports. Every change to a driver or a KPI definition is approved by a named owner and recorded with its reason. We apply the same approach in a second field: case files in law firms.
Main focus
Value driver networks and KPI ontologies
Value driver network review
We make the links between operational drivers and financial KPIs explicit and test them against your data – in retail logistics, for example, from picks per hour, dock-to-stock time and OTIF to logistics cost per order, inventory turnover and EBIT.
KPI ontologies
A KPI ontology records, for each KPI, its definition, its place in the hierarchy, its calculation rule and the source of its data – in a form that people, systems and AI tools can read. Language models prepare the draft; your controllers check and approve it.
Included in every engagement
Approval points for AI-drafted proposals, agreed roles and a record of each decision – and training for your team in what language models can and cannot do.
Second field
Case file pipelines for law firms
Pipelines that turn case files into structured, searchable information – with lawyers reviewing the results.

Approach
Our approach in one sentence
We make assumptions visible, test them against data, present well-founded changes for your decision and document every step.
Why now
Assumptions go out of date
Every value driver network rests on assumptions about which operational drivers move which financial KPIs, and how strongly. After a crisis, a market shift or a new regulation, these assumptions may no longer hold – yet budgets and forecasts are still built on them.
Until now, testing such assumptions regularly took considerable effort. Today, language models propose candidate drivers and KPI definitions, and causal analysis of your data shows which of them the data supports, which it contradicts and which remain open.
Which of your planning assumptions does your data still support?
We look forward to an initial conversation.
