Selected Case Studies

Contents

Telegraph Travel

Judgement that survives the second trip

The Opportunity

Shelf space beside Lonely Planet authority and Google-scale convenience

The competitive bar was never 'another travel app.' Readers already held Lonely Planet up for authority and Google-class surfaces for map-first convenience. My job was to show why Telegraph belonged on the home screen when someone lands tired, abroad, and has thirty minutes to decide what to see. My JTBD…

The Solution

Twenty-nine deep guides, not a database pretending to be editorial

The strategy was to put Telegraph Travel judgement into the product: structured picks, itineraries, and maps so the app reads as the travel desk in software, not a thin rebadge of generic listings. Execution focused on reducing cognitive load in transit and keeping tone aligned with the voice travellers already trust.

Waitrose

4.6 million members, clearer offers, twelve per cent more repeat visits

The Opportunity

Seven per cent more members on a food brand that cannot win on points alone

John Lewis Partnership's 2024/25 annual report puts the scale in plain sight: My Waitrose grew 7% to 4.6 million active members [JLP report 2025](https://www.johnlewispartnership.co.uk/~/media/Files/J/john-lewis/corp/documents/john-lewis-plc-ara-24-25-signed.pdf)::pill while more customers shopped with the Partnership's brands. Waitrose sales reached £8.0 billion [JLP report 2025](https://www.johnlewispartnership.co.uk/~/media/Files/J/john-lewis/corp/documents/john-lewis-plc-ara-24-25-signed.pdf)::pill. The business invested £150 million in lower prices [JLP report…

The Solution

Free club for food lovers, not another points race

My Waitrose works when membership connects three loops in one experience: everyday treats and visible savings at the till, food-lover discovery through recipes and seasonal ranges, and personalised offers explained in plain language online, in the app, and in store.

The Impact

4.6 million members, clearer offers, twelve per cent more repeat visits

In 2024/25, My Waitrose grew to 4.6 million active members, up 7%, while Waitrose sales reached £8.0 billion. Loyalty does not own every line of that growth; it makes food inspiration, Partner service, and relevant value easier to feel on each shop.

Apple

One live view replaced eight tabs and cut weekly debates over the numbers

The Opportunity

Eight tools, no shared truth for performance questions

Apple's global sales organisation ran on a patchwork of tools. Teams across major regions jumped eight apps for basic performance questions: hardware revenue, services attach, channel mix, quarterly targets. Figures did not always match, field work slowed, and nobody had one live view that tied hardware, services, and channel together. The job-to-be-done was clear: give account executives, leadership, and partners one source of numbers they could trust for demand conversations and quarterly plans. I led discovery that included…

The Solution

One mobile-first analytics app: same numbers for field, leadership, and partners

I designed a mobile-first sales analytics app so fragmented data lived in one place for field teams. Three options were considered. A web dashboard that required VPN and a laptop: rejected, because account executives work from airports and retail floors, not desks. A read-only mobile view of existing tools: rejected, because it would perpetuate the eight-app problem. A unified mobile-first app with offline support and role-based views: what I shipped. I owned the key UI decisions: dashboards that…

The Impact

One live view replaced eight tabs and cut weekly debates over the numbers

One app replaced the habit of checking eight systems and disagreeing on the figure. Account executives saw the same real-time picture as leadership, so demand conversations and channel plans started from shared data. Partner briefings matched what executives saw without reconciliation drag. Less time matching spreadsheets meant more time on planning and execution; field plans could stay aligned with quarterly goals because everyone started from the same screen. Because I unified hardware, services, and channel into one mobile…

Google

From 22% to 41% activation, with 35% fewer setup tickets

The Opportunity

Two million publishers, one snippet, and a 58% cliff before the first ad shows

Auto Ads used machine learning to place ads and surface inventory owners overlooked, yet adoption lagged behind the product's depth. The integration was a single snippet, but it read opaque if you did not ship code every day. Manual placements stayed familiar; the ML path sounded like a gamble you could not verify from a kitchen table on a Sunday night. My JTBD research surfaced three blockers: unclear value (why trust ML over hand-placed units?), technical confusion (where…

The Solution

Three emails for three stalls: explain, paste, prove

I closed the gap with segmented email so each publisher saw copy matched to where they were stuck, not a generic welcome blast. A single email to all signups could not serve technical and non-technical owners equally; platform diversity across WordPress and custom stacks ruled out an in-product tutorial overlay. I shipped a behaviour-led email journey segmented by signup source and stall point. Each email had a fixed layout hierarchy: subject line named the stall ('Paste your snippet…

The Impact

From 22% to 41% activation, with 35% fewer setup tickets

Activation climbed from 22% to 41% within 30 days of signup after the segmented programme launched, clearing the 35% target I set upfront: segment 2 (paste instructions) recovered the largest stall cohort, and segment 3 (verify impressions) converted publishers who had pasted but never saw revenue move. Support tickets tied to Auto Ads setup dropped 35% quarter-over-quarter because screenshots replaced abstract ML copy at each stall point. Segmenting email by stall point with CMS-specific paste steps and reporting…

Vodafone

Steer while the period is still open

The Opportunity

Operational spend, rear-view visibility

Spikes came from rollouts, misconfigured devices, roaming, firmware shifts, and fleet usage, not one-off bugs. After billing closed you could explain what happened. You still could not reliably forecast the bill, set guardrails early, or catch drift in time to act. Bill shock eroded trust and finance slowed expansion sign-offs. Ops burned time on spikes without clean attribution or a fast path to root cause. When setup and docs were hard to use, automations never shipped and good…

The Solution

Plain-language forecasts, caps with ownership, docs that match the API

I designed forecasting to answer 'What will we likely spend this period?' in language finance teams already use. Three directions were on the table. Exact predictions with confidence intervals: rejected, because false precision eroded trust when reality diverged. Historical average only: rejected, because it missed rollout and configuration changes. Plain-language forecast ('Likely £12K to £15K this month based on current usage and planned rollouts') with breakdowns by fleet and site: what I shipped. I structured breakdowns to follow…

The Impact

Steer while the period is still open

Teams shifted from 'what happened last month' to what's coming, where caps apply, and what changed. Finance could plan earlier. Ops could act before close. Programme owners could grow fleets with fewer invoice surprises. Plain-language forecasts grouped by fleet and site, cap warnings with named owners, and docs aligned to UI vocabulary ('spending cap' in both places) drove billing escalations tagged 'unexpected spend' down 58% in two quarters, clearing the 50% target. Developer onboarding for cap integration fell…

EE

Task success up 26 points average; coverage and help calls down

The Opportunity

Shoppers lost in eight top-level menus with overlapping labels

EE ran eight top-level navigation items with unclear boundaries: 'Why EE', 'Network', 'Help', and five product categories competed for attention. Customers hunting for coverage, upgrade eligibility, or account help often clicked three menus before finding the right path or abandoned to call support. My JTBD research identified three core missions: shopping for devices and plans, managing existing accounts (My EE), and getting help and network status. I led discovery that included clickstream analysis showing high bounce rates from…

The Solution

Tree testing first, then three buckets: Shop, My EE, Help

I ran tree testing with 200+ participants across twenty high-frequency tasks (finding coverage checker, comparing pay monthly plans, checking network status, understanding upgrade eligibility). Three options were considered. Keep eight menus with relabelling: rejected, because tree tests showed structural confusion, not only labelling issues. Mega-menu with all options visible: rejected for mobile complexity. Three clear buckets (Shop, My EE, Help) with supporting sub-navigation: what tested highest for task success. I owned the key IA decisions: 'Shop' carried all…

The Impact

Task success up 26 points average; coverage and help calls down

Tree testing and task success scores gave us clear evidence: the three-bucket structure improved completion rates across all twenty tested tasks. Coverage checker usage increased 34% in the first quarter after launch because I moved it from buried under Why EE / Network to Help › Coverage checker: the same IA change that raised task success on coverage tasks from 52% to 78% in tree tests. Support contacts related to 'cannot find information online' dropped 22% in the…

John Lewis

35% faster discovery, £8.2M incremental revenue, attribution you could defend

The Opportunity

A £15B catalogue where loyalty met filter fatigue

John Lewis runs more than 400,000 SKUs online. Loyal shoppers still stalled in category depth: filters failed to surface the right stock quickly, and Quick View stopped short of a confident buy. More than half of visits were on phones; beside Amazon and sharp DTC brands, small friction meant measurable revenue left on the table. My JTBD research identified three pain points: slow filtering on deep pages (smart TVs returning 200+ results with unclear differentiation), Quick View needing…

The Solution

Filters with memory, Quick View that finishes the job, PDPs that front-load trust

I ran three workstreams in parallel: filtering with live counts and saved combinations, Quick View with size, delivery, and add to basket in the overlay, and product page hierarchy that surfaced key specs earlier. Each workstream had rejected paths. Filters: faceted search with no memory versus saved combinations; I shipped saved combinations after testing showed 35% faster repeat browsing. Quick View: lightweight preview versus full overlay with checkout actions; I shipped the full overlay after A/B tests showed…

The Impact

35% faster discovery, £8.2M incremental revenue, attribution you could defend

Filtering and Quick View cut time-to-product by about 35%, measured as median clicks from category entry to add-to-basket. Listing and product page work lifted engagement and basket size. Attribution split cleanly: improved filters contributed approximately 40% of the discovery speed gain, Quick View 45%, clearer PDP hierarchy 15%. The programme landed around £8.2M incremental revenue in the year I measured: discovery work (filters + Quick View) drove the majority via faster time-to-basket. Because I shipped logged-in prefill and…

Aviva

Downloads above plan, measurable savings, brand lift on invention

The Opportunity

Demographic buckets masked real driving risk on the road

UK car insurance still leaned on broad demographic buckets: safer drivers often paid like risky ones because price was not tied to trips on the road. Most people think they drive better than average, so flat stereotypes felt unfair and missed real risk. Aviva needed a signal from the vehicle, not only the postcode on the form. My JTBD research showed two core needs: young drivers wanted a fair chance to prove safe behaviour and lower premiums; families…

The Solution

UK-first smartphone telematics: score the trip, quote from behaviour

I led design for the UK's first smartphone-based telematics app, pricing cover from how people actually drive. Three options were considered. Black-box hardware installed in vehicles: rejected for installation friction and cost. OBD-II dongles: rejected for limited vehicle compatibility. Smartphone GPS and sensors: what I shipped. Any driver with a smartphone could participate without hardware installation or vehicle compatibility limits. I owned the key UI decisions: the app used GPS to log acceleration, braking, cornering, speed, and phone…

The Impact

Downloads above plan, measurable savings, brand lift on invention

Aviva Drive passed 1M downloads in about six weeks, well above the annual plan, picked up Apple's 'App of the Week', and held around 4.5/5 in the store. Roughly four in ten users scored 7.1 or above and saved about £101 on average; most people in the programme qualified for some discount. Because I designed trip breakdowns with star ratings per behaviour and visible badges for milestone progress, drivers corrected braking within three trips and 18% more completed…

Lloyds Banking Group

75% cut in resolution time, 60% digital self-serve in six months

The Opportunity

Three call types, 250 FTEs, and resolution stretching past fifteen minutes

The UK's largest retail bank faced acute pressure in home insurance contact centres: high volume, rising cost, frustrated customers. Three call types (amendments, cancellations, policy or cover queries) consumed nearly 75% of total handling time and over 250 FTEs. Average resolution ran from about six minutes to well over fifteen, leaving little slack for good service. My JTBD research identified three needs: customers wanted to amend cover (add contents, change excess) without 20-minute waits; agents needed faster access…

The Solution

One design system, rule-based flows, volume shifted to digital

As design director I led strategy and team coordination and owned hands-on design of progressive disclosure on amendment flows, asking for information only when the customer's path required it. Bespoke flows per brand would have put Halifax, Lloyds, Bank of Scotland, and MBNA on diverging maintenance paths with inconsistent experiences. Generic CRM forms with agent scripting would have perpetuated hold times. I built one design system with rule-based flows for self-serve and agent use across all four brands.…

The Impact

75% cut in resolution time, 60% digital self-serve in six months

Lloyds moved the heaviest insurance jobs into digital self-serve: the same routes that had consumed three-quarters of handle time. Average resolution time fell by about 75% for targeted call types. Digital self-serve crossed 60% within six months, exceeding the 50% target. Because I designed progressive disclosure on amendment flows (one starter question, then only relevant fields), customers completed amendments in under four minutes on mobile instead of fifteen on phone, and agent satisfaction improved as teams handled judgement-heavy…

Eutelsat OneWeb

Less reconciliation, more engineering judgement

The Opportunity

Strong engineering, fuzzy version truth

Eutelsat OneWeb's published fleet counts run to [650 satellites](https://www.eutelsat.com/system/files/2025-10/DOCGroupURD-Eutelsat-2024-25EN301025.pdf)::pill in low Earth orbit, [70 teleports](https://www.eutelsat.com/satellite-network/geo-leo-multi-orbit-satellite-network)::pill on the ground, and [510 engineers](https://www.eutelsat.com/system/files/2025-10/DOCGroupURD-Eutelsat-2024-25EN301025.pdf)::pill in the workforce figures Eutelsat reports for FY 2024–25. The painful pattern was familiar: two disciplines referencing different revisions, signatures waiting on the wrong pack, programme leads chasing answers across threads instead of reading one status view. My JTBD research identified three workflow gaps: new specs and change packs started from blank folders with no standard structure;…

The Solution

Templates, routing, and live status beside the repositories teams already used

I shaped automation around real programme rhythms: structured starting points, routing that respected roles and boundaries, review checkpoints only where policy required them. Three directions were on the table. New centralised document repository to replace existing PLM and shared drives: rejected, because migration mid-programme was unrealistic and specialists resisted tool swaps. Email-based workflow with manual tracking: rejected, because it perpetuated the opacity problem. Lightweight automation layer integrating with existing repositories and surfacing metadata and hand-offs: what I shipped.…

The Impact

Less reconciliation, more engineering judgement

Teams spent less time debating which file counted and more time on the technical decision itself. Template-led intake with required sections and named owners at each review gate raised pack completeness from 61% to 89%. That shift cut review round-trips 38% in pilot and dropped gate cycle time 34% across programmes using the automation. Programme leads could scan document state at a glance instead of chasing answers across threads. Post-pilot surveys showed 82% of engineering leads reporting 'workflow…

Welsh Water

Callbacks down from forty per cent to eleven, crews dispatched twenty-eight per cent faster

The Opportunity

Three million served, one queue for meter reads and mains bursts

Welsh Water already operated at full civic scale: storms, bursts on large mains, and heavy capital work were normal operating noise for a supplier serving around three million people [Welsh Water](https://corporate.dwrcymru.com/en/about-us)::pill. Contact centres carried that load alongside everyday customer work. My JTBD research surfaced three triage problems. Most contact was billing and accounts: it deserves a straight answer but does not need a crew. Leaks and supply hits showed up less often in totals but felt urgent when…

The Solution

Guided intake and RPA, not a chatbot pretending to understand

I ran a design sprint with product, operations, and content to agree what to build first: guided flows that walk someone through reporting a leak or another service request, capturing location, severity, anything unsafe on site, and photos when useful, with plain-language next steps on screen. Three options were considered. Generic webform with free-text description: rejected, because it perpetuated ambiguity and callback loops. Chatbot-style conversation: rejected, because the programme predated reliable LLMs and customers needed fast triage, not…

The Impact

Callbacks down from forty per cent to eleven, crews dispatched twenty-eight per cent faster

The aim was practical: less ambiguity when a request arrives, fewer callbacks to fill gaps, and dispatch seeing what crews need on the first screen. Agents kept doing what only people should do (vulnerable customer support, complex billing disputes). Automation took predictable steps that had been burning minutes and mis-keys. Because I designed branching intake with structured location, severity, and risk fields on first submit, callback rates on digital leak reports dropped from 40% to 11% and average…