The invisible field methodology

Data Analysed

We brought together publicly available organisational and funding data to map the visible landscape of support for young people, with a focus on boys and young men. This included charity register data, community interest company data, cooperatives data, charity classification tags, organisation descriptions and, for the deep-dive sample, information available on organisations’ own websites. We matched this to 360Giving GrantNav data to analyse funding flows. The analysis does not include individual beneficiary records, private case data, or personal data. It does not capture statutory provision, schools, universities, private providers, informal/unincorporated groups, or funders that do not publish grant data in 360Giving format.

To inform our analytical approach, we reviewed sector documents such as research published by Movember, The Centre for Social Justice, Equimundo, and UK Government’s publication on Men’s Health Strategy for England. Based on this literature, we developed five social-impact domains, what we call our framework, capturing the important outcomes for boys and young men. Our quantitative analysis followed five stages. First, we defined the organisational ecosystem by creating a UK-focused dataset and removing organisations outside scope, including funding bodies, overseas charities and organisations with no usable description. Second, we classified organisations by beneficiary group, identifying whether they primarily served young people and, where relevant, whether they had a specific focus on Boys & Young Men or Young People more generally. Third, we classified each in-scope organisation into one or more of the five social-impact domains identified as part of the literature review. Fourth, we produced cohort-level statistics, including cohort sizes, income estimates, domain shares, confidence intervals. Finally, we matched 360Giving GrantNav funding data to the organisation list and conducted a deep dive into the BYM cohort. We conducted conversations with funders, practitioners, and sector experts to test and interpret the findings emerging from the data.

Gender Categories   

Two primary gender categories were used for the headline analysis:  

• Boys & Young Men (BYM): organisations whose name, beneficiary description or delivery model identifies boys and young men as the primary group.  

• Young People (General): organisations supporting young people without a specific gender restriction.   

• Health Outcomes and Wellbeing: Covers mental health distress and suicide, physical health risks, risky coping behaviours, and barriers to help-seeking or staying engaged with support.
• Relationships and Belonging: Covers social isolation, loneliness, weak support networks, and the role of family relationships, fatherhood, healthy masculinity, mentoring and positive role models in shaping outcomes for boys and young men.
• Education and Economic Pathways: Covers education disadvantage, including school readiness, attainment gaps, exclusions and SEND-related issues, alongside employment and economic insecurity such as NEET risk, lack of meaningful work, retraining needs and uncertainty about future pathways.
• Safety and Healthy Digital Engagement: Covers violence, exploitation and offending-related harms, including knife crime, criminal exploitation, and anti-social behaviour, alongside the risks associated with digital environments such as harmful online content, pornography exposure, violent media, and manosphere or influencer cultures.
• Systems and Infrastructure: Covers the wider conditions needed to strengthen the field, including evidence, advocacy, partnerships, coordination and organisational capacity.

Key limitations  

The analytic approach has several limitations. First, classification depends heavily on the language organisations use in their public descriptions and websites. Organisations with sparse, vague or unusual descriptions may be misclassified or remain unclassified. Website evidence and AI-supported review mitigate this, but do not eliminate the issue.

Second, the gender classification is organisation-level and binary for the purposes of the headline analysis. Many general youth organisations may support boys and young men as a substantial part of their work, but if they are not BYM-specific, they remain in the Young People category.

Third, GrantNav is a strong open source for UK grant funding, but it is not exhaustive. Funders that do not publish grants in 360Giving format are absent, so the funding analysis should be read as a lower-bound estimate of funding flowing to the in-scope organisations.

Fourth, income and funding figures include some estimation. Missing income values were imputed using median values within relevant strata, and grants to multi-domain organisations were split equally across domains. These are transparent and auditable assumptions, but they mean some financial figures should be read as indicative rather than precise.

Finally, our deep-dive on BYM organisations is sample-based. It provides statistically tested estimates of the organisational landscape, not a manual census of every BYM organisation.

Use of Artificial Intelligence (AI)

We used Artificial Intelligence (AI), including large language models (LLMs), in supporting roles within the process. Final analytical judgements were reviewed and validated by the research team to ensure accuracy and consistency.

• We used a pre-trained sentence encoder to compare the semantic similarity between organisation descriptions and our social impact domain definitions.
• LLMs (Claude Opus 4.8) were used to support the development of the analytical pipeline, including refining Python code and troubleshooting errors.

Classification method

We classified organisations in stages, beginning with the clearest and most transparent methods, and only using more judgement-based checks where needed.

First, we searched organisation names and descriptions for agreed keywords to identify whether an organisation worked with young people, whether it had a gender focus, and which broad area of work it was most relevant to. We drew on the public charity registers for England and Wales, Northern Ireland and Scotland; the UK Third and Civil Society Sector Database for community interest companies; and Cooperatives UK open data for cooperatives. We also enriched the charity data with charity classification tags, giving more detail on organisations’ activities, beneficiary groups and areas of focus.

We then used semantic matching to classify organisations into the five social impact domains. This went beyond simple keyword searching by comparing the meaning of an organisation’s description with our domain definitions, rather than relying only on exact words or phrases. For example, an organisation might not use the phrase “mental health”, but still clearly describe work around wellbeing, emotional support or counselling. To support this, we used a sentence embedding model, which turns written descriptions into numerical representations of meaning, allowing organisations to be matched based on context as well as wording.

For a small number of harder cases, we used a large language model as a supporting check. This was limited to reviewing unclear beneficiary groups, assigning a best-fit domain to BYM-focused organisations that had not been classified, and checking tags in the boys and young men deep-dive. The LLM was not used as the first-line method for deciding which organisations were included; it was used only to support classification where the primary methods were uncertain.

Boys & Young Men (BYM) deep dive

The Boys & Young Men deep dive was designed to understand what the BYM organisational landscape looks like in more detail. Because reviewing every BYM organisation manually was not feasible, we drew a statistically valid random sample from the estimated BYM population and classified each sampled organisation against a taxonomy with seven domains:

• Life stage: which age/life-stage groups (e.g. adolescence, later adolescence, emerging adulthood and young adulthood) are visible.
• Transition context: whether support appears around key moments such as exclusion, NEET status, justice risk or early fatherhood.
• Intersectional reach: which groups are explicitly named, such as low-income, minority ethnic, LGBTQIA+, rural or SEND groups.
• Delivery setting: where support takes place, such as schools, community spaces, sport, digital settings or family/home.
• Issue area: what needs (e.g. isolation, mental health distress, education, employment, violence/exploitation, physical health and digital harms) are addressed.
• Underlying drivers: what causes or barriers (e.g. life-stage vulnerability, loss of youth/community infrastructure, stigma and masculinity pressures) organisations name.
• Service response: what organisations deliver (e.g. including peer support, mentoring, school-based support, employment pathways, safeguarding, counselling and advocacy).

The deep dive findings should be treated as estimates rather than exact counts. We used a statistically valid sample of organisations focused on boys and young men, and report percentages with confidence intervals to show the likely range around each estimate. To check that the findings were robust, we ran the deep dive twice using two different random samples.

Funding analysis

For the funding analysis, we matched grants from 360Giving’s GrantNav to our organisation list using recipient identifiers. Each matched grant inherited the recipient organisation’s gender category and domain classification. Where an organisation was tagged with more than one domain, we split the grant value equally across those domains. This means domain-level funding figures are indicative estimates of how funding is distributed across areas of work, rather than precise allocations of each grant’s intended use.

We also examined funder portfolios, including funders’ relative focus on boys and young men and funders with significant youth giving but no matched BYM-specific grants in the available data. We aimed to understand not only how much funding reaches BYM-specific organisations, but also which types of funders are, or are not, currently visible in that landscape.