Profile Olly Paton.
Digital analytics, experimentation and growth specialist. Twenty years building the systems — and the teams — that move commercial numbers.
What I actually do.
I work across digital analytics, experimentation and product analytics — leading the thinking and shipping the thing that makes it real. The point is always the same: growth. Cleaner numbers, sharper experiments and AI-grade data are means; more revenue, better retention and lower CAC are the ends. I'm comfortable in a meeting with the CFO about marketing ROAS, equally comfortable in a Cloudflare Worker that runs an experiment at the edge, and just as comfortable writing the Python that joins the two.
I started in web analytics in the early 2000s — back when getting a clean number out of Urchin was a small victory — and the work has kept evolving since: Universal Analytics, GA4, Adobe Analytics, tag managers, data layers, server-side tagging, CDPs, product analytics, experimentation platforms, and now AI agents that need clean, well-modelled events to be useful at all.
How I think about the work.
Most analytics problems aren't analytics problems. They're codification problems — processes, definitions and decisions living in heads, spreadsheets and Slack threads instead of structured systems that other people (and increasingly, agents) can use.
AI makes this worse before it makes it better. An LLM querying a dirty event stream gives you a confident, well-written wrong answer. The fix isn't more dashboards; it's better instrumentation, contracts on the events themselves, and workflow systems that encode how analytics work is actually done. That's the thread running through Conform, Conversion Workers, and most of the consulting I take on.
Where my expertise sits.
- Growth strategy: Connecting analytics and experimentation work directly to revenue, retention and CAC — so the technical work demonstrably moves the business, not just the dashboard.
- Analytics platforms: GA4, Universal Analytics, Adobe Analytics, server-side tagging, BigQuery, data warehouse integration.
- Experimentation: A/B and multivariate testing programmes, Optimizely, VWO, custom edge experimentation on Cloudflare Workers, Bayesian and frequentist analysis.
- Product analytics: Event modelling, retention and funnel measurement, integration with engineering and PM workflows.
- Marketing measurement: Total marketing ROAS, attribution modelling, incrementality testing, consent-aware tracking.
- Tooling & infrastructure: Tag management (GTM, Tealium), data layers, event schemas, CI/CD for analytics, AI tooling and agentic workflows.
- AI-enabled workflows: Designing systems that let LLMs and agents actually do useful analytics, experimentation and ops work — without hallucinating on broken data.
- Operator experience: I built and own radio-controlled.co.uk, an ecommerce business at meaningful scale (see below) — so I know what it's like on the receiving end of a "go-live with the new tracking" Friday afternoon.
The operator side: radio-controlled.co.uk.
Alongside Conversion Works, I built and own radio-controlled.co.uk — a UK ecommerce business in the RC hobby space, now trading at around half a million in turnover across roughly 20,000 SKUs. It's a deceptively complicated business: hundreds of brands, supplier-driven catalogue churn, variable stock signals, multi-warehouse fulfilment, returns, pre-orders, bundle pricing, edge-case shipping rules, and a customer base that knows the products better than most operators ever will.
Building it forced me to live inside many of the same problems my clients face. Tagging that breaks at midnight has revenue consequences by breakfast. Attribution gaps aren't theoretical; they're the difference between scaling a working channel and turning it off. Catalogue-quality issues become AI-quality issues the moment you point an agent at the data. Most of the patterns that ended up in Conform, Conversion Workers and the SEO Dashboard came directly out of solving these problems for myself first.
Today the business runs largely on heavy automation I built — using the same patterns and tools I bring to client work — so the operational day-to-day doesn't dominate my week. The hard-won lessons stay; the operational grind has been engineered out. Whatever I recommend to clients, I've already had to ship on a live store with real customers.
What I build.
Conversion Works is the umbrella for the practical tools I make. Conform is an edge-native event quality gateway — it sits between your event sources and your analytics collectors and refuses to let broken data through. Conversion Workers is an edge-first experimentation platform built on Cloudflare Workers — fast, no flicker, and integrated with whatever analytics you already use.
Both came directly from work I was doing for clients. Both are designed to make AI-era analytics actually trustworthy.
Working together.
Engagements usually take one of three shapes: a focused diagnostic on something that's not working, a longer growth programme to design and run analytics or experimentation at scale, or a build engagement where I deliver a tool or workflow that becomes part of your stack. Most of the time it's some mix of all three — anchored to a clear commercial outcome.
If that sounds like something you need, get in touch — or have a look at the articles first to see how I think about the work.
Brands I've delivered for.
Across analytics, experimentation, product and growth — sometimes as a consultant, sometimes embedded, sometimes building the thing that ended up running in production.
