Understanding which innovation projects are most likely to succeed is a long-standing challenge for governments, funders, and R&D‑intensive organisations. Despite the widespread use of Technology Readiness Levels (TRLs) as a benchmark for maturity, few organisations have a reliable, data‑driven way to predict how projects will progress through these levels or what proportion of a portfolio will ultimately reach critical milestones.
Our recent work tries to close this gap by developing a structured, statistical methodology to estimate the likelihood of TRL progression at both the project and portfolio levels. The approach, tested using large datasets from energy innovation programmes, offers a practical and transferable approach that other organisations can readily adopt to move from descriptive reporting towards forward-looking decision making.
A robust, evidence‑based approach
The methodology combines three core elements:
- (1) A rapid evidence review. We synthesised insights from more than 20 studies exploring TRL progression, innovation risk, and project delivery (including work conducted by DARPA and NASA). From this, we identified recurring drivers of success – including initial TRL position, project complexity, team capabilities and broader framework conditions. This ensured the modelling was not only data‑driven but also grounded in existing theory.
- (2) Statistical modelling using real project data. The core of the work consisted of ordinal and logistic regression models estimated on cleaned and harmonised project data. Following extensive exploratory analysis, including correlation screening, we tested a wide set of potential explanatory variables and identified the strongest predictors of reaching higher TRL thresholds, particularly TRL 7+. Four factors consistently stood out:
- Initial TRL,
- Expected TRL after project completion,
- Project duration (non‑linear effect),
- Lead organisation size (weak but directionally informative).
By contrast, commonly assumed drivers such as funding amounts showed no statistically significant effect once maturity, expectations and timing were accounted for.
- (3) Portfolio‑level probability modelling. To translate project‑level predictions into portfolio‑level insights, we used a Poisson binomial distribution, allowing each project to retain its unique probability of success. This enables organisations to answer practical questions such as:
“What is the probability that at least 20% of our portfolio will reach TRL 7+?”
“How many projects do we need in the pipeline to ensure a minimum number reach TRL 8 or TRL 9?”
Why this matters for other funders and organisations
Although developed in a public innovation funding context, the approach is highly transferable. Any organisation managing a pipeline of research, product development or complex technological innovation can benefit from moving beyond descriptive TRL reporting towards probabilistic decision-support.
The framework supports more strategic portfolio design by identifying which combinations of project characteristics – such as starting TRL, expected progress and duration – are associated with higher success probabilities. It also allows organisations to stress-test assumptions by constructing synthetic portfolios (for example, exploring outcomes if all projects enter at a higher TRL), thereby improving transparency and risk awareness in investment decisions.
Looking ahead
The predictive power of the model is strong but can be improved further. In particular, gathering data on the complexity of the integration of the technology developed under a specific project, as well as information on team capabilities, may improve its predictive power.
Ultimately, this work demonstrates that TRL progression can be predicted, and that a thoughtful combination of evidence review, statistical modelling and probability theory can help funders and other organisations make smarter, more transparent decisions about their innovation investments.
For more information contact: Cristina Rosemberg and Iakov Frizis.