Leveraging Government Data

Governments, funders, and service providers need credible evidence about how employment and training programs are performing. However, rigorous impact evaluations can require considerable time, funding, data, and technical expertise. They cannot feasibly be conducted for every program on an ongoing basis.

The Leveraging Government Data (LGD) initiative is developing a benchmarking model that can complement existing evaluation methods. Led by Blueprint and part of the Future Skills Centre’s Building Data Capacity portfolio, the model uses historical information about employment and training programs to predict the outcomes that could reasonably be expected for different participants, interventions, and labour-market conditions.

Organizations could compare their observed results with these predictions. This could help them identify outcomes that are stronger or weaker than expected, compare possible program designs, set realistic portfolio-level expectations, and explore how changing economic conditions could affect results. Benchmarking cannot provide the same causal evidence as a randomized controlled trial, but it can offer a more informative assessment than examining outcomes without a credible point of comparison.

Blueprint initially tested the concept using data from 8,398 participants across 11 programs in the Future Skills Centre’s Scaling Up Skills Development portfolio. This work established an initial modelling process and clarified the data and outcomes required. It also demonstrated the limitations of building a broadly applicable model from a relatively small portfolio.

In September 2025, Blueprint received secure access to Employment and Social Development Canada’s Labour Market Program Data Platform. This much larger source contains administrative information about participants in provincially and territorially delivered labour-market programming.

Blueprint prepared an analytical dataset that combines participant histories and program information with local labour-market conditions and program costs. We then tested several modelling approaches. A Bayesian regression model predicting average net income over the five years following program participation provided the best balance between predictive performance and practical usefulness.

In July 2026, Blueprint delivered an internal technical report to Employment and Social Development Canada. It describes the data, modelling process, model performance, and potential applications. Blueprint is now responding to feedback, briefing federal officials, and working to ensure future development reflects the needs of government decision-makers.

A final report planned for December 2026 will present further detail, subject to authorization to release analytical outputs. It will also provide a roadmap for refining and applying the model so it can support better planning, evaluation, and continuous improvement across Canada’s skills ecosystem.

Reports and resources

• Leveraging Government Data: Status Update 2 (coming October 2026)
• Leveraging Government Data: Status Update 1 (April 2025)
• Leveraging Government Data: Design Report (June 2024)