Industry impact
or rigorous research
— pick the right door first.

Reference-letter template (industry / Tech Nation): 'I worked with [DS] at [Project] from [Year]-[Year]. They led the experimentation infrastructure that powered [N experiments] generating [verifiable business outcome — £X ARR, +Y bps retention]. Their [conference talk / blog post / open-source contribution] articulated the work to the wider community. Among data scientists I've worked with at this seniority, they are in the top ~5 in [sub-domain].'

Reference-letter template (academic / British Academy or Royal Society): 'I have known [Researcher] since [Year]. They are PI / co-PI on [grant of size £X], with [N] first-author publications at [AEJ / Lancet / JSM / RSS-A], cumulative citations of [N]. Their methodological contributions to [sub-area] have shaped my thinking. I consider them among the strongest applied econometricians / biostatisticians of their cohort.'

Quantified-impact narrative example for the personal statement: 'As staff data scientist on [Experimentation / Recommendation / Risk] at [Company], built the [system] that generated [£X ARR / +Y% retention / -Z% fraud loss] across [scale]. Documented at [public engineering blog / paper], talked at [PyData / Data Council 2024]. Three [open-source / methodological] contributions adopted by [named users].'

Recognition narrative example: 'Invited keynote at PyData London 2024 (1,800 attendees). Sustained Substack at [verifiable subscriber count] with [name-recognition coverage]. Top contributor on [open-source data tooling] used by [N companies including named users]. Formal advisor at [N funded data startups, verifiable via Companies House].'

The data-science discipline has fragmented into two distinct career trajectories with different evidence catalogues. The industry-applied path is closer to engineering: shipping data products, building experimentation infrastructure, owning ML feature stores, scaling analytics platforms. The research path is closer to academia: publishing in econometrics / health economics / methodology venues, applying for grants, peer-reviewing, supervising students.

Picking the wrong endorsing body for your evidence is the single most common rejection cause for data scientists. Tech Nation reviewers want product-impact narratives — they're less equipped to weight an econometrics-paper publication record. Royal Society / British Academy reviewers want publication / citation / grant evidence — they're less equipped to weight an A/B-testing platform you built at scale.

The decision principle: where would the dominant 60% of your strongest evidence land more accurately? If your CV reads more like a Stripe / Netflix / Airbnb / Booking data team contributor with a few papers, that's Tech Nation. If your CV reads more like a research economist or biostatistician with some industry consulting, that's British Academy or Royal Society.

Hybrid careers — academic-trained data scientists now in industry, or industry-rooted statisticians publishing on the side — should pick the route matching the dominant 60% and use the rest as supporting evidence. Don't try to optimise for what you think the panel wants to see; optimise for honest evidence-route fit.

Strong industry evidence catalogue: top-N maintainer of a major data tooling project (dbt, Airflow, Dagster, Prefect, Polars, DuckDB, Pandas, scikit-learn, Streamlit) with verifiable adoption; quantified business impact narratives ('led the experimentation platform that generated +£12M annualised through 47 experiments'); named industry-conference invited talks; advisory roles at funded data / analytics startups; public technical writing with verifiable audience numbers; participation in standards-track work (Apache Foundation projects, Linux Foundation AI / Data initiatives).

Public technical writing is well-rewarded — sustained writing with audience numbers (Substack subscribers, blog analytics, conference attendance for derivative talks) is concrete evidence. Distill / The Gradient / Towards Data Science (with verified reach) / personal-domain blogs all count.

What doesn't clear the bar: BI dashboards, internal analytics ownership, Kaggle medals alone, internal awards. The recurring rejection-pattern is 'I led a data team at Company X' — internal scope without external recognition.

Strong academic evidence catalogue: 3+ first-author or last-author papers at top venues for your sub-discipline (Journal of Econometrics, AEJ Applied Economics, AEJ Applied, Lancet Global Health, BMJ, Statistical Science, JRSS-A, JASA, Biometrika, NeurIPS Statistics, JSM, NIPS workshop on Causal Inference); citation impact corresponding to your sub-field's senior-IC level; sustained peer-review and Area Chair or Editor duties; PI or co-PI on competitive grants (UKRI ESRC, NIHR, Wellcome, ERC); senior referee letters from named academics.

British Academy is particularly suited to economists, applied microeconomists, public-health analysts, sociologists doing computational work, and quantitative political scientists. Royal Society is suited to computational statisticians, biostatisticians, methodologists in physical / life sciences, and applied probability researchers.

Senior referee letters matter materially. Three letters from full professors at named UK / international universities, Royal Society Fellows, BA Fellows, or senior figures at recognised research institutions. Letters should attest to specific work, not just status.

London senior / staff data-science salaries are £100-180k+ at scaled tech firms, with specialised roles (ML platform leads, head of data at growth-stage startups, principal data scientists at unicorns) reaching £200-300k+ base. Add equity at high-growth companies and total comp can equal mid-tier US Bay Area packages, particularly at UK arms of US public companies.

Founder optionality: Global Talent permits founding companies. The UK data / ML startup ecosystem is dense — funded by Atomico, Index, Notion Capital, Local Globe, Plural, Seedcamp, EF, plus international funds active in the UK (Sequoia, A16Z, Tiger).

Academic / industry hybrid: many UK universities (Imperial Business School, LSE, UCL, Cambridge Judge, Oxford Saïd, Manchester Alliance, Edinburgh Business School) actively recruit applied data scientists for joint industry-academic roles. The visa supports this dual path.

Public sector: ONS Data Science Campus, NHS England Analytics, Office for National Statistics, Behavioural Insights Team, AI Safety Institute, and various government-data initiatives recruit senior data scientists. Civil-service salaries are lower than private but the work is interesting and visible.