Vaultline came to us with a clear problem: retail investors had no good option between a basic spreadsheet and an expensive Bloomberg terminal. They needed a fast, clean, genuinely useful portfolio tracker — built for people who take their investments seriously but don't work in finance full-time.

We started with three weeks of user research — interviewing 40 retail investors about their current workflow and pain points. The consistent feedback was that existing tools were either too simple or too complex. Our design principle became: surface the right data at the right time, hide everything else.

The dashboard was built around a single performance chart with contextual overlays for news events, dividends, and economic data. Every data decision was tested with real users before being built. The tech stack was chosen for speed above all — Next.js with server components, PostgreSQL for complex portfolio queries, and Recharts rendered on the server to eliminate layout shift.

The beta launched to an invite-only list of 500 users. Word of mouth drove 4,200 sign-ups in the first month before any paid marketing. The Core Web Vitals score of 94 was achieved through aggressive server-side rendering and edge caching — LCP dropped from 2.8s to 0.5s on a standard 4G connection.

Lumé had a beautiful product and a brand that resonated. What they didn't have was a store that matched. Their existing Shopify theme was a generic template that fought against conversions — slow to load, cluttered on mobile, and with a checkout flow that lost 60% of customers before purchase.

We rebuilt the store from scratch in Liquid — no theme templates, no compromises. Every design decision was rooted in conversion data. The product page was redesigned around a single clear CTA, with social proof, ingredient breakdowns, and a subscription upsell placed at psychologically tested scroll depths.

The mobile checkout was rebuilt as a single-page flow, reducing the number of screens from 4 to 2. We integrated Klaviyo for abandoned cart sequences with product-specific copy rather than generic reminder emails.

Within 60 days of launch, conversion rate increased by 41% and cart abandonment dropped by 28%. The subscription upsell — presented after the add-to-cart action — converted at 22%, adding meaningful recurring revenue to what had previously been a purely transactional business.

Fieldbook's founding team understood that the existing farm management software market was designed by software companies, not farmers. The apps assumed reliable internet, desktop usage, and users with time to learn complex interfaces. None of those assumptions hold in a field in Yorkshire at 6am.

We built offline-first from day one — SQLite for local storage, with a custom sync engine that queues changes and pushes them to the server when connectivity returns. Every action the user takes is stored locally first. The server is a sync target, not a dependency.

The UI was designed with one principle: every common action must be reachable in two taps or fewer. We tested paper prototypes with 12 farmers before writing a single line of code. The map view — used to log crop data by field — was built using a custom SVG renderer rather than a heavy mapping library, keeping the app under 15MB installed.

Fieldbook launched on the App Store and Google Play simultaneously. Within 6 months, 850 farms had onboarded — growing entirely by word of mouth within farming communities. The 4.8-star rating and 11-minute average session time reflect an app that people actually want to use.

Archon Media were spending £180k per year on external content agencies to feed their clients' publishing schedules. Their internal team was spending 80% of their time on briefing, chasing, and editing — not strategy. They needed a way to reclaim that time without sacrificing content quality.

We built a custom LLM pipeline using GPT-4o with a structured prompt chain: brief → research → outline → draft → meta → social variants. Each step was tuned with client-specific style guides, tone of voice documents, and brand glossaries fed via a RAG system built on their existing content library.

The React dashboard was designed around a review-and-approve workflow rather than a generation interface — the assumption being that editors are skilled at reviewing and refining, not at prompt engineering. One button generates a full content package; the editor's job is to make it better.

Archon replaced their external agency contracts within 90 days of launch. The 10× content output figure reflects not just speed but coverage — they were able to serve three times as many client accounts with the same headcount. The brief-to-publish time dropped from 4 hours to 22 minutes on average.

Kroft Interiors were getting referral business but struggling to convert website visitors into enquiries. Their previous site looked like a template — generic stock photos, a standard contact form, and nothing that communicated the quality and specificity of their work. They were losing high-value clients to competitors with better digital presence.

We started with a brand identity refresh before touching the website — establishing a typographic system, a neutral colour palette that let the photography lead, and a set of editorial layouts inspired by architecture and design publications. The logo mark was a refined geometric form that scaled from favicon to signage.

The website was built to feel like a portfolio publication, not a service business. The project gallery used a fullscreen lightbox with context notes for each image. The enquiry form was deliberately longer than standard — asking detailed questions about project type, timeline, and budget — to qualify leads before they reached the client.

Organic enquiries tripled within 90 days, driven by the SEO improvements and the increased time-on-site (4 minutes 12 seconds average). The enquiry form filtering meant the quality of leads improved significantly — Kroft reported that more than 80% of enquiries from the new site converted to paid consultations, up from under 30%.

Nodi's founding team had a strong idea — a platform where people could exchange skills without money changing hands — but nothing visual to show investors. They had 8 weeks before a seed round meeting. They needed a brand, a landing page to capture waitlist sign-ups, and a clickable prototype credible enough to anchor an investor pitch.

We compressed a typical 12-week brand and build process into 8 weeks by working in parallel — branding and product design running simultaneously. The brand identity was built around the concept of connection: an organic node-and-line system that referenced both human networks and the app's peer-to-peer mechanic.

The landing page was designed as a conversion machine — one clear value proposition, a frictionless email capture, and a social proof counter that updated in real time as sign-ups accumulated. The Figma prototype covered the core user journey: sign up, create a skill listing, browse matches, make a request, and complete an exchange.

1,840 people joined the waitlist before the app existed, built entirely through organic social sharing. The investors funded the round at £320k — citing the quality of the product vision as communicated through the prototype as a key factor. Nodi was featured in TechCrunch UK two weeks after the round closed.