Website Edit Plan

Generated from strategic planning session, April 2026

Context and Goals

Two sites, two distinct roles:

  • inferenceworks.co.uk — the company site. Formal, commercial, buyer-facing. This is where a corporate L&D manager, marketing director, or data team lead lands and decides whether to enquire. It needs to sell.
  • drbenvincent.github.io — the personal reputation site. Technical, credibility-building, community-facing. This is where a practitioner, conference organiser, or potential collaborator goes to understand who you are in depth.

The immediate commercial goal is to establish a corporate workshop business alongside consulting — modelled on the observation that subject-matter experts with organic content audiences can convert that audience into B2B training revenue without paid marketing. Your niche is causal inference and Bayesian analytics. Your sharpest commercial wedge is marketing measurement (MMM / attribution).

Site 1: inferenceworks.co.uk (Astro / Netlify / GitHub)

Current state

Preliminary Astro content, not yet strategically structured. Previously a WordPress placeholder.

Overall tone and audience

Write for buyers, not practitioners. The reader is a data science manager, marketing analytics lead, or L&D professional. They don’t know what PyMC is. They do care about whether their teams are making good decisions from data.

1.1 Homepage

Headline — replace whatever is there with something buyer-oriented. Examples: - “Causal thinking for data and analytics teams” - “Better decisions through causal data science” - “Your data team can reason causally. Most don’t yet.”

Sub-headline / intro paragraph — 2–3 sentences max. Answer: who do you help, what problem, what do you offer. Example:

InferenceWorks helps data teams, marketing analysts, and organisations move beyond correlation-based thinking. We offer expert consulting and practical workshops in causal inference, Bayesian analytics, and marketing measurement. Based in Scotland, working internationally.

Credibility signals block — logos or brief text mentions: - PyMC Labs (subcontractor / collaborator) - Named clients: HelloFresh, Bill & Melinda Gates Foundation, Colgate-Palmolive - CausalPy — open source package, 1,100+ GitHub stars - Book (forthcoming — can say “co-author of upcoming book on [topic]”)

Services summary — two clear tracks, each with a one-line description and a link: - Consulting → /consulting (or link to relevant page) - Workshops → /workshops

CTA — one clear action at the bottom of the hero / above the fold: - “Enquire about a workshop” → links to /workshops or scrolls to contact

1.2 New Page: /workshops

This page is the most commercially important thing to build. It is the destination for any inbound interest.

Page structure:

Intro (3–4 sentences)

Explain what the workshops are, who they’re for, and the general format (delivered to teams, half-day or full-day, remote or in-person, no minimum technical background required for some tracks).

Workshop 1: Marketing Measurement — Beyond Attribution

  • Audience: Marketing directors, marketing analytics leads, agencies, CMOs
  • Format: One day, remote or in-person, no coding required
  • The problem it solves: Multi-touch attribution gives systematically wrong answers. Most marketing teams are optimising budgets based on numbers that don’t reflect reality.
  • What we cover: Why attribution models mislead; what Media Mix Modelling is and how it works conceptually; incrementality and experiment design for marketing; how to brief your data team or agency on measurement
  • Outcome: Attendees leave able to critically evaluate measurement approaches and ask the right questions of their analytics function
  • Why this is your sharpest offering: The buyer (marketing director) has budget authority. The pain is concrete and expensive. It connects naturally to Pangolin as a follow-on product.

Workshop 2: Causal Inference for Data Teams

  • Audience: Data scientists, analysts, ML engineers
  • Format: One or two days, remote or in-person, Python-based practical elements
  • The problem it solves: Most data teams can identify correlations but can’t reliably answer causal questions (“did this intervention work?”, “what would have happened without this change?”)
  • What we cover: Confounding and why it matters; directed acyclic graphs (DAGs) as a thinking tool; quasi-experimental methods (diff-in-diff, regression discontinuity, synthetic control); practical implementation in Python with CausalPy
  • Outcome: Teams can design and evaluate natural experiments, interpret results causally, and flag when causal claims are unjustified
  • Note: Can be tailored for more or less technical audiences

Workshop 3: Experiment Design and Causal Evaluation

  • Audience: Product teams, growth analysts, data scientists running A/B tests
  • Format: One day, remote or in-person, optional Python elements
  • The problem it solves: A/B tests are routinely misdesigned, misanalysed, or misinterpreted. Statistical significance is not the same as a reliable causal estimate.
  • What we cover: What makes a valid experiment; power analysis and sample size; common failure modes (SUTVA violations, novelty effects, peeking); moving beyond p-values; Bayesian approaches to experiment analysis
  • Outcome: Teams run more valid experiments and make fewer bad decisions from misleading results

Pricing note

No need to publish prices. Use: > Pricing depends on team size, format, and duration. Get in touch to discuss your requirements.

Enquiry CTA

Simple contact section at the bottom. Options: - Email link (inferencelab@gmail.com or a contact@ address if you set one up) - A simple contact form (Netlify Forms handles this natively with zero backend — just add netlify attribute to the form tag)

1.3 New Page: /consulting (or update existing if present)

Shorter than workshops. Covers: - What consulting engagements look like (project-based, embedded with a team, advisory) - Types of problems: causal analysis, MMM, experiment design, Bayesian modelling - Engagement via InferenceWorks directly or through PyMC Labs - CTA to get in touch

Site 2: drbenvincent.github.io (Quarto / GitHub Pages)

Current state

A well-populated technical site with Blog, Portfolio, and CV. Reads as an academic CV with a consulting note. Strong credibility content already exists — it just needs reorienting.

Overall tone and audience

This site serves practitioners, collaborators, conference organisers, and technically-minded buyers doing due diligence. It can go deeper than InferenceWorks. But it currently undersells the commercial and applied dimension of your work.

2.1 Homepage bio rewrite

Current: > Dr Benjamin T. Vincent is a consultant and educator in the Bayesian and causal data analysis space. He is the director of InferenceWorks Ltd. Previously, he held a permanent faculty position for 15 years using Bayesian data analysis methods to research human decision making. Originally from Cambridge (UK), he is currently based in Scotland.

Problems with current: - Leads with the academic framing (“previously held a faculty position”) - No mention of workshops or training - No mention of CausalPy or open source work - Doesn’t tell a practitioner or buyer what you’re currently focused on

Suggested rewrite: > Dr Benjamin T. Vincent is a Bayesian data scientist and causal inference specialist. He is the director of InferenceWorks Ltd, where he offers consulting and workshops to data teams and organisations. He is the lead developer of CausalPy, an open source Python package for Bayesian causal inference, and a contributor to PyMC-Marketing. He works with clients via PyMC Labs, and has delivered projects for organisations including HelloFresh, the Bill & Melinda Gates Foundation, and Colgate-Palmolive. He holds a PhD and previously held a faculty position for 15 years applying Bayesian methods to human decision-making research.

(Adjust to taste — but note how this version leads with what he does now, treats the academic background as supporting evidence rather than the headline, and explicitly names workshops.)

2.3 Portfolio page — minor additions

The portfolio is already strong. Two small additions:

  • Add a “Workshops & Training” section near the top listing the three workshop offerings with a link back to inferenceworks.co.uk/workshops
  • Add the MMM calibration webinar to the Talks section (you appeared as expert speaker — worth listing)

2.4 Future consideration: blog post framing

The existing blog posts are technically excellent but primarily practitioner-facing. Over time, add a second category of post: shorter, less code-heavy, aimed at a business reader. These would live here and serve as the upper funnel content that eventually drives workshop enquiries.

Topics to consider: - “Why your A/B test results probably aren’t what you think” - “What is Media Mix Modelling and why does it matter?” - “How we measured the causal effect of X without running an experiment” - “Three things your analytics team wishes you understood about causation”

Cross-site Linking

Once both sites are updated, make sure the links are consistent:

From To Link text
inferenceworks.co.uk homepage drbenvincent.github.io “About Dr Ben Vincent” or “Meet the founder”
drbenvincent.github.io homepage inferenceworks.co.uk “InferenceWorks Ltd” (already present)
drbenvincent.github.io nav inferenceworks.co.uk/workshops “Workshops ↗”
drbenvincent.github.io portfolio inferenceworks.co.uk/workshops “Workshop offerings”

Sequenced To-Do List

Ordered by commercial impact:

Priority Task Site Effort
1 Build /workshops page InferenceWorks 2–3 sessions
2 Rewrite homepage hero and add credibility block InferenceWorks 1 session
3 Add Workshops to nav InferenceWorks 10 mins
4 Add Netlify contact form to workshops page InferenceWorks 30 mins
5 Rewrite homepage bio drbenvincent 20 mins
6 Add Workshops link to nav drbenvincent 10 mins
7 Add workshops section to portfolio page drbenvincent 20 mins
8 Add webinar to Talks section drbenvincent 10 mins
9 Build /consulting page InferenceWorks 1 session
10 Write first “business reader” blog post drbenvincent 1–2 sessions

Notes on Tooling

  • InferenceWorks edits: Astro components in VSCode, commit and push to GitHub, Netlify auto-deploys. Use Claude Code or Copilot for agentic multi-file edits. For contact forms, Netlify Forms requires no backend — add netlify to the <form> tag and a hidden form-name input.
  • drbenvincent edits: Quarto .qmd files in VSCode, push to GitHub, GitHub Actions deploys to Pages. Small targeted edits are easy; nav changes may require editing a shared layout or _quarto.yml.

End of plan