case library
15 Active cases across the category taxonomy.// Work
14 years of engagements, grouped by the shape of the problem.
Seven categories, fifteen active cases. Ordered to lead with the loudest evidence: AI workflow design and data engineering first, methodology underneath. Each case is its own page; click in for the full anatomy. Names anonymised where engagements are under NDA.
categories
07 Problem-shape taxonomy used to organise the work.experience
14 yrs Analytics experience across consultancy and in-house.credentials
11 Amplitude credentials including two Expert tracks.alert noise
83% Reduction in V1 production AI alert triage.BQ cost
~50% BigQuery cost reduction from pipeline redesign.Directory logic
The work is grouped by problem shape, not employer chronology.
Use the category index to jump by hiring question: AI workflows, data engineering, instrumentation, adoption, marketing analytics, experimentation, or question discovery. The page stays static and manually maintainable; no auto-updating timestamps, charts, or invented operational data. Category numbers reflect the problem taxonomy, not the display order — the page leads with the highest-impact evidence.
Category 04
1 case
AI workflows
Where AI assists, where humans still decide.
Category 07
2 cases
Data engineering
Pipeline architecture, cost-thinking, schema design, and the patient work of stepping back to look at the whole system.
Category 02
5 cases
Instrumentation
The technical layer the measurement stack runs on: SDKs, event design, QA, schemas, and cross-platform parity.
Category 05
1 case
Adoption and enablement
The final measure of an engagement is whether the team changed how they work.
Category 06
1 case
Marketing analytics
CDP architecture, lifecycle automation, attribution, and the operational data layer behind campaigns.
Category 03
2 cases
Experimentation
Designing the measurement framework before the change ships, so product decisions can be evaluated rather than narrated afterward.
Category 01
3 cases
Question discovery
The methodology underneath everything else: identifying the right question before optimising the wrong system.
Category 04
AI workflows
Where AI assists, where humans still decide.
For: Product Analytics / UX Telemetry / AI workflow design
Case
Major quick-service restaurant brand / analytics architecture engagement
Designing AI workflows that don't outsource the thinking
A production AI alert-triage system with explicit human-judgement gates.
Category 07
Data engineering
Pipeline architecture, cost-thinking, schema design, and the patient work of stepping back to look at the whole system.
For: Analytics Engineering / Data Architecture / Platform Analytics
Case
State-regulated retail platform / analytics architecture engagement
Three raw scans to one. Fifteen CTEs to zero. ~50% cost reduction.
GA4-to-BigQuery pipeline redesign cutting scheduled-query cost roughly in half.
Case
Fuel iX / TELUS Digital enterprise GenAI platform / agency-client engagement
Product analytics for an enterprise GenAI platform
GA4-to-BigQuery pipeline for enterprise GenAI platform analytics.
Category 02
Instrumentation
The technical layer the measurement stack runs on: SDKs, event design, QA, schemas, and cross-platform parity.
For: Product Analytics / Analytics Engineering / UX Telemetry
Case
State-regulated retail platform / analytics architecture engagement
From what should we measure to 30+ shipped tickets
Greenfield mobile analytics layer across iOS and Android: strategy to production.
Case
Global AI services firm / candidate sourcing platform
A 50%+ candidate drop-off, decomposed step by step
Candidate journey instrumentation that turned vague drop-off into step-level evidence.
Case
Major quick-service restaurant brand / Amplitude monitors / checkout telemetry
Clean taxonomy, trustworthy alerts, enriched context
Three layers of remediation on a noisy alert channel.
Case
Major quick-service restaurant brand / Segment CDP / mobile app telemetry
A silent iOS data-loss bug, traced to the initialisation sequence
Human-led anomaly detection plus AI-assisted SDK source review for a silent iOS undercount.
Case
B2B/B2C hardware authentication brand / e-commerce instrumentation / agency engagement
All analytics in the tag manager. Not one line in the site code.
GTM-only analytics architecture on a Magento/SPA e-commerce storefront, driven by a security requirement that banned analytics code from the codebase.
Category 05
Adoption and enablement
The final measure of an engagement is whether the team changed how they work.
For: Insight roles / Product Analytics / Consulting
Case
State-regulated retail platform / analytics architecture engagement
The reporting layer survived the people who built it
BigQuery-to-Looker Studio reporting infrastructure plus a structured handoff designed for independence.
Category 06
Marketing analytics
CDP architecture, lifecycle automation, attribution, and the operational data layer behind campaigns.
For: Marketing Analytics / CRM Analytics / Lifecycle
Case
Consumer electronics retailer / lifecycle marketing and CDP migration
Replatformed the CDP. The campaigns did not notice.
Oracle Responsys to Bloomreach CDP migration with the customer model redesigned, not ported.
Category 03
Experimentation
Designing the measurement framework before the change ships, so product decisions can be evaluated rather than narrated afterward.
For: Product Analytics / Marketing Analytics / UX Telemetry
Case
Stanley Steemer / home services franchise / 280 locations
Introducing A/B testing as a leadership practice
Five years building marketing analytics from scratch at 280 franchise locations.
Case
B2B/B2C hardware authentication brand / conversion optimisation / agency engagement
The quiz worked. The question was where to put it.
GA instrumentation confirmed a product quiz's conversion signal; VWO A/B testing found the store-page placement.
Category 01
Question discovery
The methodology underneath everything else: identifying the right question before optimising the wrong system.
For: UX Research / Insight / Analytics Strategy
Case
Consumer electronics retailer / lifecycle marketing engagement
The workaround broke. The data showed the gap. Frontline staff knew why.
A shadow IT discovery that became a core product capability.
Case
Tattoo studio / small business / freelance engagement
The CTA worked. The question it answered was too shallow.
A failure case: accepted the wrong brief, the fix worked, but the understanding didn't transfer.
Case
Stanley Steemer / national home services franchise / 280+ locations
Telemetry locates. Method diagnoses.
Quantitative telemetry paired with Sense-Making-informed qualitative research at a national home services franchise.