Executive summary

The Consultative Group to Assist the Poor (CGAP) introduces precision analytics and Precision Causal Modelling (PCM) as methodologies to bridge a significant knowledge gap within financial service providers (FSPs). While FSPs typically track repayment, they often lack insight into why similar clients experience products differently. PCM is a “target-first” discovery method that combines causal inference with machine learning to identify which product and delivery combinations are associated with better outcomes for specific client segments using existing administrative data. For investors, PCM provides a more credible basis for due diligence by moving beyond cross-sectional averages to understand which client groups are experiencing positive outcomes and identifying the contributing factors.

Why precision matters: The case for a different kind of evidence

Research indicates that while the average impacts of inclusive credit are often modest, these averages mask significant heterogeneity. Subgroup signals suggest that certain borrowers, such as those with prior business experience, see larger effects. Conventional evaluative methods often struggle to provide the granular, institution-specific insights needed for day-to-day operational decisions. PCM addresses this by helping FSPs identify how products should be priced, how loan ceilings should be set, and how repayments should be structured based on finer-grained variation in client response.

What the pilots revealed: Findings from PCM as an evaluative methodology

CGAP piloted PCM with five institutions: AMK Cambodia, Asociación Grameen Costa Rica, BBVA Microfinance Foundation, FINCA International, and the Banco Central do Brasil. The pilots confirmed that “one size does not fit all”. When clients received the combination of product features and services identified as effective for their segment, they experienced significantly higher rates of successful outcomes. Delivery gaps—the difference in outcome achievement rates between clients who received optimal configurations and those who did not—ranged from 32 to 84 percentage points within the same portfolios.

Four substantive patterns emerged with enough consistency to serve as a diagnostic framework for FSPs:
1. Right-sizing loans is critical; under-lending was the most consistent predictor of failure.
2. The trajectory of delinquency (whether a client is recovering or deteriorating) is a more reliable signal than a single missed payment.
3. Structured loan growth of 30 to 50 per cent per cycle was associated with better outcomes than stagnant or rapid increases.
4. The quality of the client relationship and service delivery, such as loan officer tenure, significantly influences success. In one pilot, officers with nine or more years of tenure improved outcomes by up to 44 percentage points for vulnerable segments.

How PCM works: A deeper look at the methodology

PCM proceeds through six phases, beginning with agreeing on a theory of change and target outcome. It uses matching methods to construct comparable “Block Matched Groups” (BMGs) from within an institution’s existing client base, allowing for a credible approximation of causal effect. Findings are categorised into three types: actionable levers (design features the institution can modify), assessment signals (client characteristics that help identify capacity or risk), and milestone indicators (variables that track successful trajectories but should not be treated as levers).

Is your institution ready for PCM? A practical self-assessment

Readiness is determined by five conditions: data volume and quality, meaningful outcome measurement, sufficient treatment variation, analytical capacity, and institutional change management capability. Preparing data for analysis is demanding; in CGAP’s pilots, data preparation alone consumed between one-third and one-half of the total analytical effort. The methodology requires high-quality, longitudinal, client-level records linking baseline characteristics to product exposure and welfare outcomes.

From evidence to action

To close the gap between analysis and action, FSPs should start with specific analytical questions and focus on delivery consistency before product redesign. The report recommends measuring client trajectories rather than snapshots. For investors, PCM should be framed as an investment in institutional learning and internal capacity building rather than as a tool for accountability. Common errors to avoid include confusing predictive models with causal ones and overinterpreting noisy estimates for small client segments.

Conclusion: The case for precision

The central lesson for the sector is that the information needed to improve client outcomes is often already embedded in an institution’s own data. PCM makes these insights visible, connecting them to product, policy, and operational decisions. The move from retrospective verdicts to ongoing decision-support processes offers a more direct path from evidence to impact, provided institutions build the necessary analytical culture and data foundations.