Impact analysis methodology
The platform runs two complementary analyses: one estimates what each mobility measure contributed to observed KPI changes, the other ranks options across competing criteria.
Impact analysis model
The impact analysis tool estimates the contribution of each mobility measure to observed KPI changes across Living Labs. It uses ridge regression, a variant of linear regression designed to handle situations where the number of predictors (measures) is large relative to the number of observations (cities).
Use the impact analysis tool → Understanding the methodology
Regression analysis approach
Understanding the methodology
Regression analysis approach
The impact analysis assesses the effect of push and pull measures implemented across multiple Living Labs as part of Sustainable Urban Mobility Plan (SUMP) interventions. Policy measures are combinations of restrictions on private car use (push) and incentives for shared mobility adoption (pull).
The analysis follows a three-step process:
- Baseline data collection : KPIs are measured before the interventions.
- Impact estimation : KPIs are recalculated after implementation and compared to the baseline.
- Interactive interface : Results are communicated through a dedicated interface to support stakeholder exploration.
To quantify the contribution of each measure, a Ridge regression model (linear regression with L2 regularization) is used. Each measure is encoded as a binary variable (implemented = 1, not implemented = 0) per Living Lab. The model associates a regression coefficient to each measure, estimating its individual contribution to the observed change in each KPI. Ridge regularization is chosen because the number of measures exceeds the number of Living Labs, helping prevent overfitting and address multicollinearity.
Both univariate (per KPI) and multivariate (grouped KPIs of the same type, ie. environmental, societal, economic) analyses are conducted to ensure robustness, especially when data points are limited for local KPIs.
Data collection
What data is this analysis based on?
Data collection
What data is this analysis based on?
KPIs tracking
Collection of KPI data before and after the implementation of push and pull measures across all Living Labs
Push and pull measures implementation
Binary records (0/1) per Living Lab indicating which push and/or pull measures were applied (e.g. parking charging, vehicle sharing)
Participating cities
SUM Living Labs and Contributing cities participating in the SUM Open Data Platform, providing a diverse dataset for analysis
The analysis draws on data collected across all Living Labs, combining intervention records and performance measurements (KPIs). These inputs feed into the regression model to estimate the impact of each measure on mobility outcomes.
How to interpret the results ?
Understanding regression coefficients and model accuracy
How to interpret the results ?
Understanding regression coefficients and model accuracy
The model outputs a regression coefficient (β) for each measure, representing its estimated contribution to the observed change in a given KPI:
- A positive coefficient indicates the measure contributed to an increase in that KPI.
- A negative coefficient indicates a decrease.
- A coefficient close to zero suggests the measure had little to no isolated impact on that KPI.
Results should be interpreted with the following in mind:
- If a measure was implemented uniformly across all Living Labs, its individual effect may be harder to isolate, and results should be read with caution.
- The mean squared error (MSE) of the model provides an indication of estimation accuracy. A lower MSE means the model fits the observed data more closely.
Multi-Criteria Decision Analysis (MCDA)
The multi-criteria decision analysis (MCDA) tool uses the PROMETHEE II (Preference Ranking Organisation METHod for Enrichment Evaluations) method to rank business activities for New Shared Mobility according to multiple criteria simultaneously.
Use the decision tool → PROMETHEE-GAIA Multi-Criteria Analysis
Multi-criteria decision analysis approach with PROMETHEE and GAIA visualization
PROMETHEE-GAIA Multi-Criteria Analysis
Multi-criteria decision analysis approach with PROMETHEE and GAIA visualization
What is PROMETHEE?
What is GAIA?
How to read the PROMETHEE-GAIA results?
Pro tip for interpretation
Compare results across different perspectives to identify measures that work well for multiple stakeholders.
How to interpret the results ?
Understanding Multi-Criteria Analysis outputs and model accuracy
How to interpret the results ?
Understanding Multi-Criteria Analysis outputs and model accuracy
PROMETHEE-GAIA results combine two complementary outputs that work together to support decision-making. PROMETHEE produces a ranked order of the transport policy alternatives, reflecting how each city's interventions perform across all weighted criteria and stakeholder perspectives. A higher rank indicates a policy that better satisfies the overall set of criteria as valued by the stakeholders involved. The GAIA plane then brings a visual dimension to these rankings, mapping both the policy alternatives and the evaluation criteria onto a two-dimensional graph. On this plane, criteria pointing in similar directions indicate aligned objectives, while criteria pointing in opposite directions reveal trade-offs. Policies positioned close to a criterion's axis perform well on that specific criterion. A decision axis summarizes the overall PROMETHEE ranking direction, helping users quickly identify the most globally preferred alternatives.
Together, these two outputs allow users to:
Rank Alternatives
Compare transport policies based on their overall performance across all criteria and stakeholder groups
Identify Trade-offs
Spot tensions between competing objectives (e.g. cost reduction vs. environmental impact) that cannot be simultaneously optimized
Explore Consensus
Identify areas where different stakeholders' preferences align, supporting more inclusive and robust decision-making
Data context and limitations
The results presented on this platform are based on data collected from SUM Living Labs and Contributing cities. They reflect patterns across cities, but do not include local indicators, context, or the local knowledge each city holds about why something worked.
You can use it to
- See which measures tend to co-occur with improvements in a KPI across cities
- Compare your city against peers on a common, validated indicator set
- See how a ranking shifts when different stakeholder priorities are applied
- Judge how much data sits behind any given result, before you rely on it
- Build a shortlist of measures worth a local feasibility study
You cannot use it to
- Conclude that a measure caused an observed outcome
- Assume a result from one set of cities transfers to yours unchanged
- Read an MCDA ranking as an objective ordering independent of whose priorities were used
- Treat a missing value as a zero, or as evidence that nothing happened
- Replace local expertise, stakeholder engagement or regulatory assessment