Working in the charity sector, and particularly in impact, we all know that rigorous monitoring and evaluation matters.

In a climate of tightening public finances, increasing accountability and growing expectations from funders, demonstrating impact has never been more important.

The rise of agentic AI has transformed what's possible in analytics. It can surface patterns, process volumes of data, and generate insight that would once have been unimaginable.

Used well, it has enormous potential to strengthen how charities learn and improve.

But they also present a new challenge. AI can construct persuasive narratives from
incomplete evidence. It can identify correlations that look compelling, generate confident explanations and produce polished reports in minutes. What it cannot do is establish causality where the evidence doesn't exist.

For charities, the question is no longer simply whether we can tell an engaging story about our impact. It is whether our claims withstand independent scrutiny, whether we achieve the outcomes we set out to deliver, and whether those outcomes genuinely change people's lives.

That was the challenge we set ourselves at social mobility charity upReach.

Looking beyond our own data

Every year, we collect more information about who we support, their challenges and their aspirations, and then data on how they engage with the programme, and the outcomes they go on to achieve. This data helps us continually refine our programmes, respond to the changing labour market conditions and improve the experience of the people we work with.

Like many charities, we have developed increasingly sophisticated operational reporting. We know our programmes well and deepen this knowledge daily.

But operational data has its limits. It tells you what happened, but it cannot tell you what would have happened anyway.

If a participant secures highly skilled employment after graduating was that because of our programme? Would they have achieved the same outcome regardless? How much of the difference can genuinely be attributed to our intervention? Higher education outcomes are subject to a number of factors, with each starting point and set of circumstances faced fundamentally unique.

Answering these questions requires more than high-quality operational data. It requires independent evaluation designed to establish whether the intervention itself made a measurable difference. That became our ambition.

Journal-worthy, not just report-worthy

When we began developing our Request for Proposals (RFP), we were clear about the standard of evidence we wanted to achieve.

We weren’t looking for another consultancy report or a descriptive evaluation. Those have an important place, particularly in learning and refinement through reflective story-telling, but we wanted something capable of standing up to academic and policy scrutiny.

In short, we wanted research that was closer to journal-worthy.

Specifically, we commissioned three complementary pieces of work.

First, a summative impact evaluation using a quasi-experimental design, comparing participants' outcomes against a rigorously matched comparison group.

Second, a process evaluation to understand not only whether the programme works, but how, why and for whom.

Finally, we wanted an honest assessment of what would be required to progress towards a randomised controlled trial in the future, and whether that would even be appropriate given the complexity of our operating environment.

It was an ambitious brief.

The transition from higher education into graduate employment is shaped by countless interacting factors, from university attended and subject studied to geography, labour market conditions and socio-economic background. Isolating the contribution of any single intervention is inherently difficult, which is precisely why we believed rigorous research was necessary.

Building on previous evidence

This was not the first study of its kind conducted by upReach. Our earlier benchmarking work, based on our 2019 graduating cohort, informed our 2022 Impact Report and represented an important step in our evaluation journey. Alongside outcome studies with university partners, it helped to strengthen our understanding of participant outcomes and continuously improve our approach to impact measurement.

But this next stage, only made possible by our committed funders, allowed us to take this work to the next level.

Lessons from the journey

Looking back, three lessons stand out for organisations considering similar work.

Finding the right partner takes time.

Identifying a research partner capable of delivering this level of evaluation was harder than we anticipated.

Higher education, social mobility and graduate employment are incredibly complex systems. Any organisation promising to demonstrate simple cause and effect should probably raise more questions than confidence.

The strongest research partners were those who were candid about the limitations of the evidence, transparent about the methodological challenges and willing to invest significant time in designing an approach robust enough to address them.

That honesty proved invaluable throughout the project.

Data governance matters as much as methodology

One of the biggest surprises was how much of the work had nothing to do with statistical modelling.

Preparing data for independent evaluation required us to revisit governance, documentation, information sharing, and data protection in considerable detail. It involved multiple conversations with universities, legal teams and data protection specialists, alongside careful consideration of GDPR and our responsibilities to participants.

Far from being an administrative exercise, this strengthened our organisation, enhanced all of our data processing and future-proofed our research systems. Long before the final report was published, the evaluation had already made us a better organisation.

Rigorous Evaluation cannot be rushed

Like many charities embarking on major evaluation projects, we underestimated the time involved. Design refinement, stakeholder engagement, ethics, governance, data preparation and technical review all required considerably more iteration than expected.

At times, progress felt slow.

In reality, those iterations strengthened the quality of the final evaluation. Every challenge forced us to ask better questions, improve our processes or increase confidence in the eventual findings.

Good evidence takes time.

A learning journey

One of the most valuable outcomes from the project wasn’t actually about our impact. It was about research design, and committing to the next step.

Our academic partners and advisors were clear that moving directly to a randomised controlled trial (RCT) would be premature. Before considering an RCT, they recommended building a deeper understanding of the many interacting factors influencing graduate outcomes, including regional labour markets, university effects and socio-economic context. This advice guided our process and decision-making.

There is often an assumption that an RCT represents the inevitable destination for impact evaluation. In reality, rigorous evidence is built progressively. Each stage should answer the questions necessary to make the next stage worthwhile. Rigorous is not a single summit. It is a staircase, and every step needs to bear weight before you continue to climb.

The result

The result is upReach's first independent external impact evaluation, delivered by the National Centre for Social Research (NatCen) and funded by The Hg Foundation. Using a rigorously matched comparison group, the evaluation examined outcomes directly linked to our Theory of Change, including highly skilled employment, full-time employment, graduate destinations and earnings.

The findings were both encouraging and significant.

Compared with similar graduates who had not participated in upReach, our participants were:

● 44% more likely to secure highly skilled full-time employment
● 30% more likely to be in full-time employment
● 31% more likely to achieve a positive graduate destination, defined as either highly skilled employment or progression to postgraduate study
● 62% more likely to earn above the UK median graduate salary.

The evaluation also found even stronger outcomes for participants in our Tech500 programme, with differences of 65% in highly skilled employment, 45% in full-time employment, 48% in positive graduate destinations and 80% in earning above the UK median graduate salary compared with matched non-participants.

For us, however, the importance of the evaluation extends beyond the numbers themselves. For the first time, we have an externally verified, rigorously matched benchmark of our impact, not an in-context reflection, but a result that can stand in the room with policymakers, funders and academics alike.

The findings are already hard at work, informing the upReach 2030 Strategy, and building insights to improve our day-to-day work by enhancing programmatic design. We continue to analyse the data, to dig deeper and to optimise our impact.

Why this matters now more than ever

This evaluation represents another step in upReach’s own research journey, but I believe it also reflects something much stronger for the sector.

AI is transforming how we collect, analyse and interpret data, and charities should embrace the opportunities it presents. It has the potential to accelerate analysis, reveal patterns at scale and support more informed decision-making across the sector.

But technology cannot replace robust evaluation. If anything the opposite is true. The easier it becomes to generate plausible narratives, the greater the value of evidence that has earned the right to be believed.

Independent, rigorous evaluation is no longer simply about accountability to funders. It is about credibility. It is about giving policymakers, partners and the public confidence that our interventions genuinely change lives and understanding not only whether they work, but why, for whom and under what conditions.

For charities serious about demonstrating impact, investing in robust evaluation is no longer a ‘nice-to-have’, it is foundational. In a world increasingly shaped by intelligent machines, good, old-fashioned evidence may prove to be one of the sector's most valuable assets.

Angela Biden is head of data and impact at social mobility charity upReach.

---