Use Of Predictive Analytics In Swiss-Seated Arbitrations
I. What “Predictive Analytics” Means in Swiss Arbitration
In Swiss-seated arbitration, predictive analytics typically refers to:
statistical forecasting models (e.g., damages projections),
probabilistic risk assessments,
machine-learning-assisted valuation,
scenario simulations (Monte Carlo, regression),
algorithmic analysis of large datasets (pricing, performance, delays).
Swiss law does not regulate predictive analytics as a category. Such material is treated as:
expert evidence, or
technical demonstrative evidence, depending on its function.
II. Admissibility of Predictive Analytics Evidence
1. No Exclusionary Rules of Evidence
Swiss arbitration follows the principle of free evaluation of evidence (freie Beweiswürdigung).
Predictive analytics is admissible if:
relevant,
intelligible to the tribunal,
procedurally fair.
SFT Decision 4A_150/2012
Confirmed that:
arbitral tribunals are free to admit any type of evidence
No prohibition on novel or computational methods
Swiss tribunals therefore admit predictive models even if:
methodologies are contested,
assumptions are probabilistic rather than deterministic.
III. Predictive Analytics as Expert Evidence
1. Qualification and Methodology
Predictive analytics is usually introduced through:
party-appointed experts,
tribunal-appointed experts,
joint expert reports using models and forecasts.
Swiss tribunals assess:
methodological coherence,
transparency of assumptions,
reproducibility of results.
SFT Decision 4A_488/2011
Held that:
evaluation of expert methodology
lies entirely within the tribunal’s discretion
Courts do not assess scientific correctness
Thus, even controversial or cutting-edge analytics can be relied upon.
IV. Burden of Proof and Predictive Evidence
1. Probability vs Certainty
Swiss law does not require absolute certainty of proof in civil matters.
Predictive analytics may satisfy the burden of proof where:
exact quantification is impossible,
approximation is justified by circumstances.
SFT Decision 4A_240/2009
Accepted that:
damages may be estimated
using reasonable methods
Mathematical precision is not required
This is particularly relevant in:
lost profits claims,
future cash-flow disputes,
long-term project arbitrations.
V. Due Process Constraints
1. Transparency and Right to Be Heard
Predictive analytics must comply with:
the right to be heard,
equality of arms.
This requires:
disclosure of models,
access to underlying data (where feasible),
opportunity to challenge assumptions.
SFT Decision 4A_46/2011
Clarified that:
due process requires opportunity to comment,
not equal technical sophistication
Asymmetry in analytical capacity is not per se unfair
Swiss tribunals are not required to:
explain models line by line,
ensure parties fully master complex algorithms.
VI. Tribunal’s Use of Predictive Analytics Sua Sponte
1. Limits on Tribunal Initiative
Swiss tribunals may:
rely on predictive models proposed by experts,
draw inferences from statistical material.
They may not:
introduce undisclosed analytics on their own,
rely on secret computational tools.
SFT Decision 4A_360/2011
Held that:
tribunals must not base decisions
on facts or analyses not discussed with parties
Surprise analytics violate due process
Predictive analytics must therefore be procedurally integrated.
VII. Standard of Judicial Review
1. No Review of Analytical Correctness
Swiss courts will not review:
whether a predictive model was accurate,
whether assumptions were realistic,
whether a better model existed.
SFT Decision 4A_558/2011
Reaffirmed that:
incorrect economic reasoning
does not violate public policy
Even manifestly debatable analytics are immune from review
Only procedural violations or ultra petita reasoning may justify annulment.
VIII. Predictive Analytics and Public Policy
1. No Public Policy Objection
The use of algorithms or AI does not offend Swiss international public policy.
SFT Decision 4A_124/2014
Confirmed that:
technological complexity
does not implicate ordre public
Party autonomy includes evidentiary methods
Swiss law views predictive analytics as:
a neutral technical tool,
not a delegation of adjudicatory power.
IX. Cost Allocation and Predictive Evidence
1. Cost Consequences
Swiss tribunals may:
allocate costs based on usefulness of predictive analytics,
penalise excessive or opaque modelling.
SFT Decision 4A_232/2015
Confirmed that:
tribunals may reflect evidentiary efficiency
in cost awards
Over-engineering can backfire financially
X. Consolidated Case Law Table
| SFT Decision | Principle Relevant to Predictive Analytics |
|---|---|
| 4A_150/2012 | Free admissibility of evidence |
| 4A_488/2011 | Tribunal discretion on expert methods |
| 4A_240/2009 | Approximation of damages allowed |
| 4A_46/2011 | Due process ≠ equal technical skill |
| 4A_360/2011 | No reliance on undisclosed analysis |
| 4A_558/2011 | No merits review of economic reasoning |
| 4A_124/2014 | Technology neutral under public policy |
| 4A_232/2015 | Cost consequences of evidentiary conduct |
XI. Practical Implications for Arbitration Users
Predictive analytics is fully admissible in Swiss-seated arbitration.
Transparency of assumptions is more important than sophistication.
Opposing parties must be given a fair opportunity to challenge models.
Courts will not second-guess analytics post-award.
Excessive or opaque modelling may affect costs, not validity.
XII. Conclusion
Swiss-seated arbitration offers a technology-neutral and evidence-friendly environment for predictive analytics. Swiss tribunals and the SFT:
permit advanced statistical and algorithmic evidence,
treat it as expert material subject to procedural fairness,
refuse to review correctness or reliability on the merits,
protect finality and party autonomy.
As a result, Switzerland is a highly attractive seat for arbitrations involving:
complex damages modelling,
energy and infrastructure forecasting,
financial and commodities analytics,
data-intensive commercial disputes.
RELATED Blog
comments