Work · Evidence Analytics

Evidence Analytics engagements

Continuing MEL support to a long-standing client, an international agricultural development implementer, from 2026.

South Africa · Smallholder farmer development

A management information system the programme team can run and extend

Challenge: Show the funder that farmer yields and incomes rise over time, with data collected offline across dispersed rural communities, several crops and changing farmer roles.
Approach: An ActivityInfo MIS with registries and validation rules; annual outcome surveys with stratified sampling and design weights; Power BI dashboards connected through the API; plausibility checks on every refresh; and an AI-assisted field-visit advisory the agronomy team reviews before acting.
Result: An integrated pipeline from field capture to funder-ready indicators, with automated refresh and clear notes on how current each indicator is.

MIS
Sampling
Dashboards
Data quality
AI-supported

India · Smallholder sourcing

Evaluating a WhatsApp advisory pilot

Challenge: Find out whether advisory videos reach farmers, are watched and change practice, where phones are shared and in-person extension runs alongside.
Approach: A bilingual English–Hindi mixed-methods evaluation anchored to a theory of change, with women’s access as a primary lens. Messaging records were rebuilt, linked to farmer registries and sampled at handset level.
Result: The reach gap was reframed: most farmers never reached were on numbers that could not receive WhatsApp, concentrated among women and less-literate farmers. That led to specific, low-cost fixes.

Evaluation
Qualitative
Platform analytics
AI-supported

Staff role at the same client, 2020–2025

Director, then Senior Director, of Data and Learning. The first phase of the same client relationship, before it continued as consultancy.

Burkina Faso · India · South Africa · Portfolio

A single MEL backbone across a programme portfolio

Challenge: Each programme kept data its own way, so results could not be added up and staff waited on specialists for basic numbers.
Approach: A standard relational database structure; platform selection, pilot and migration of every programme onto ActivityInfo; Power BI KPI dashboards; theory of change as the standard design method; and the organization’s first Do No Harm, consent and personal-data standards.
Result: Portfolio-wide results became possible, and a funder chose to let the organization evaluate its next phase itself.

MIS
Migration
Dashboards
Governance

Burkina Faso · Household nutrition

Evidence that holds in an insecure setting

Challenge: Measure change for farm households where insecurity disrupted both production and fieldwork.
Approach: A MEL plan rewritten around flexible sentinel data collection; sampling and forms for the baseline, midline and endline of a nutrition model; a rapid A/B test with an external research institute; and an in-country review of the cleaned midline data.
Result: The analysis showed real progress on nutrition even where production targets were missed, evidence the programme could stand behind in its funder report.

Evaluation
Sampling
Sentinel monitoring

South Africa · Funder reporting

Correcting the record

Challenge: Preliminary outcome analysis containing errors reached the programme’s funder.
Approach: A full review and re-analysis of the survey data, and a message to the funder explaining the problem, giving corrected results and setting out how a repeat would be prevented. The following year’s analysis separated the effect of extreme weather.
Result: The funder invited a second-phase proposal, and the weather analysis helped secure the support the extension needed.

Data quality
Funder reporting

Earlier career, 2008–2019

Staff roles at a global anti-trafficking and justice INGO and a social-sector data-systems consultancy. These built the foundations of the practice. No AI was used in this period.

Cambodia · Philippines · Justice

Measuring a hidden crime

Challenge: Test whether stronger public justice systems reduce child sex trafficking, where the population is hidden and those affected are minors.
Approach: Prevalence studies using time-space sampling, with validation at the point of entry; research on justice system performance; and end-of-programme evaluations with external research experts.
Result: Externally reviewed studies, published by the organization, documented a substantial fall in trafficking in the programme areas studied.

Prevalence research
Sampling
Evaluation

Global · Justice data systems

From legacy databases to data staff can use

Challenge: Programme data sat in legacy case-management databases that few people could query.
Approach: Tableau introduced with a parameter-driven control panel for non-technical staff; ETL from legacy SQL Server databases into a standard reporting structure; and selection of a new organization-wide data platform.
Result: Programme data reached staff who could not query it before, and data quality improved at entry.

Dashboards
ETL
Platform selection

Case studies