Projects with
receipts attached.

Every case below lists the problem we were handed, what we changed, and what measurably happened afterwards — including the ones where the first attempt did not work.

Projects shipped

Industries

Median return on ad spend

Client retention

E-commerceShopify PlusCRO

Saffron & Co. — replatform & checkout rebuild

The problem. A 900-SKU spice retailer on a legacy cart that took four seconds to load a collection page and five steps to complete a purchase. Traffic was healthy; revenue was not.

What we did. Replatformed to Shopify Plus with a custom storefront, cut checkout to two steps with wallet payments, rebuilt faceted search, and re-merchandised collections around contribution margin instead of unit volume.

+64%Revenue / session

-41%Cart abandonment

1.1sLCP

AppReact NativeHealthcare

Lumen Health — patient companion app

The problem. A clinic group across eleven sites losing a significant share of appointments to no-shows, with rescheduling only possible by phone during office hours.

What we did. Cut the original feature list roughly in half, then shipped a focused React Native app: smart reminders timed to each appointment type, one-tap rescheduling into live availability, and secure document access.

-38%No-show rate

4.8★App Store

14 wksTo launch

SEOContentFinance

Vertex Capital — 18-month organic growth programme

The problem. A corporate finance firm invisible in search outside its own brand name, entirely dependent on referrals and an expensive paid search account.

What we did. Six weeks of technical remediation, a topical content programme written with their own partners as subject experts, and digital PR built on original transaction data. The first content cluster underperformed; we said so in month three and reallocated.

+310%Organic traffic

#1–342 target terms

+87Leads / month

AI AutomationLogisticsOps

Orbit Logistics — freight quote automation

The problem. Sales reps spent most of every morning pricing inbound freight enquiries by hand against rate cards held in four spreadsheets. Quotes routinely took a day, and the fastest supplier usually won.

What we did. Recorded a two-week baseline of the manual process, then built an agent that parses inbound requests, prices against consolidated live rate cards, and drafts the reply for approval. Anything outside confidence thresholds escalates to a human.

6 minAvg quote time

92%Auto-handled

31 hrsSaved / week

AI ChatbotEducationSupport

BrightPath Learning — support chatbot

The problem. An online education provider whose support inbox opened at around 400 tickets every Monday, most of them the same twenty questions about access, billing and course transfers.

What we did. Audited and rewrote the help centre first — the bot is only as good as its sources — then built a retrieval-grounded assistant with citations, strict refusal behaviour and context-preserving handover into Zendesk.

54%Resolved by bot

<3sFirst response

4.6/5CSAT on bot chats

Paid MediaRetailAnalytics

Northwind Retail — paid media restructure

The problem. Six figures of annual ad spend reported on clicks and impressions, with no connection between campaigns and the orders they produced. Nobody could say which half worked.

What we did. Rebuilt tracking with server-side tagging and order-level attribution before touching a campaign, then restructured accounts around margin. Two long-running campaigns turned out to be losing money and were cut in month one.

3.8xReturn on ad spend

-34%Cost per order

+22%Revenue, flat budget

Web DesignWeb DevelopmentB2B

Kestrel Studios — brand site rebuild

The problem. A production studio with excellent work and a site that hid it. Usability testing found that three of five participants could not locate pricing information at all.

What we did. Restructured the information architecture around what prospects actually asked in sales calls, rebuilt the case study format, and shipped a headless CMS the team could update without us.

+40%Enquiries

99Lighthouse perf.

-52%Bounce on key pages

AI SolutionsProfessional servicesRAG

Atlas Fitness — member retention scoring

The problem. A gym group with 14,000 members and no way to spot who was about to cancel until the cancellation arrived. Retention offers went out indiscriminately and cost more than they saved.

What we did. Built a churn model on eighteen months of attendance, booking and billing data, surfaced scores directly in the staff dashboard with the reasons attached, and targeted intervention at the top decile only.

-19%Monthly churn

0.81Model AUC

-61%Retention offer cost

In their words

The clients behind those numbers.

The checkout rebuild paid for the whole project inside four months. What surprised me more was the margin reporting — we found out our best-selling collection was our worst earner and changed what we promote.

Ravi Mehta

Founder, Saffron & Co.

🇬🇧 UK

They cut our feature list roughly in half before writing any code, which annoyed me at the time. Six months later I understand it — the app is focused, retention is well above projection, and we shipped a quarter early.

Dr. Sarah Okonkwo

Operations Director, Lumen Health

🇨🇦 CA

Eighteen months in, organic is our largest source of qualified leads. What I valued most was the month they told us a content cluster had failed and explained why, instead of burying it in an appendix.

James Lindqvist

Head of Marketing, Vertex Capital

🇸🇪 SE

The assessment phase saved us more than the build did. Two of the four ideas we brought them were dropped with a clear explanation. That kind of honesty is rare in this space right now.

Adeel Hussain

COO, Orbit Logistics

🇦🇪 AE

Six figures of ad spend and nobody could tell me what it produced. They fixed the measurement before touching a campaign, then cut two campaigns we had run for years. Revenue went up on the same budget.

Daniel Braithwaite

CTO, Northwind Retail