The target is not the conclusion.
A savings target or growth ambition may be legitimate. It is still not an analytical result. The analysis must establish what the data actually supports.
About RÄVIR · method before narrative
Small company.
Clear accountability.
RÄVIR is an independent analytical practice focused on operational data, forecasting and decision support. RÄVIR is founded and operated by Tony Weberg.
Founder · operations & analytics
Tony Weberg
Operations and data analyst with years of experience across logistics, transport, energy and operations.
Work has included forecasting and capacity planning, KPIs and performance analysis, data quality, transport and supplier performance, telemetry data and the development of analytical and planning tools.
Why RÄVIR?
I have spent much of my working life close to day-to-day operations, where an answer is only useful if it can be used and defended. RÄVIR is built on that experience: start with the question, examine the data, test the explanations and document both what we know and what we do not.
Start with the question.
What are you trying to understand, predict or improve — and how would a better answer be used?
Working principles
Open to scrutiny rather than claiming certainty.
RÄVIR does not claim to be free of bias. We therefore aim to make assumptions visible and produce analyses that others can understand, check and challenge.
Uncertainty must be visible.
We distinguish observation, calculation and interpretation. Limitations in the data remain attached to the result.
The result is not predetermined.
Scope and fixed fee are agreed before work starts. Payment does not depend on a particular conclusion.
We do not seek the desired conclusion.
We seek the best-supported one.
Empirical systems · operational intelligence
AI in the working model
A tool for speed.
Not a shortcut to truth.
AI allows a small practice to work quickly, investigate more hypotheses and reduce repetitive manual work. The client pays for the analysis and decision support — not for the size of the team producing it.
Research, code and model exploration.
AI can support research, code, data work, documentation and critical review.
Output is not evidence by itself.
Suggestions, calculations and code must be checked. RÄVIR retains human responsibility for method, conclusions and delivery.
Less repetitive work.
Automation and AI should make the working model more efficient without lowering the bar for documentation and validation.
Working method & client data
Open to scrutiny.
After delivery, too.
How is client data handled?
When AI is used directly on client data, RÄVIR's principle is to use its own self-hosted models that have been tested for the task before use. Client data is not sent to external AI services. External models may support coding and research without client data. Purpose, access, retention and deletion are agreed in advance.
What if the data cannot answer the question?
We describe the limitation and what is missing. A data-quality assessment or a recommendation to wait may be the appropriate professional result.
How does RÄVIR address bias?
We make important assumptions explicit, examine relevant alternative explanations and try to describe what new evidence would cause us to revise the conclusion.
A simple starting point
Start with the problem.
An initial conversation requires neither confidential data nor a finished specification.