Case studies.
Real systems, real numbers. The receipts behind the philosophy.
Nolvo — a client-intelligence system that briefs itself
CS-2026-01 · July 2026 · Brand-identity studio · France
How a six-day branding studio replaced ad-hoc form reading with an automated pipeline that turns every client questionnaire into a sourced market analysis, a bilingual creative brief and an interactive dashboard — before a single pixel is designed.
The client
Nolvo (nolvo.co) is a French brand-identity studio with an unusual promise: a complete, custom premium brand in six business days, from 790 € — roughly ten times cheaper and ten times faster than a traditional agency. The model works because of radical process discipline: one 20-minute onboarding form, zero meetings, validation at key milestones only. Speed is the product. Anything that slows the first 48 hours — misread briefs, missing context, back-and-forth — attacks the business model directly.
The problem
Nolvo's entire creative process starts from a Typeform: positioning, ideal client, brand personality, visual references, channels, deadlines. But the raw export is a wall of CSV. Three failure modes kept recurring:
Manual reading — insight quality depended on who read the form and when. Cost: inconsistent briefs; details lost between form and designer.
No market context — designers worked from the client's self-description only. Cost: concepts disconnected from the client's real competitive landscape.
Silent gaps — questions the client skipped or answered vaguely surfaced mid-project. Cost: revision loops that threatened the 6-day promise.
The pilot client made the stakes concrete: SOYVA, a social-media agency for building trades, answered the form in good faith but left the editorial tone as "no idea," the launch message blank, and visual direction as "honestly, I don't know — if you have ideas, why not." A typical case: an ambitious founder who knows what he wants to achieve, not how to translate it. The form alone was not a brief.
The principle applied
We do not sell heroics; we design compounding systems. The rule applied here: AI is a lever, not a shortcut. The system does not replace Nolvo's creative judgment — it removes the repetitive extraction work so judgment is applied to a complete, sourced picture instead of a raw CSV. Small, deliberate, repeatable — every client processed makes the templates sharper.
The system
A six-step pipeline (Typeform CSV → Claude → deliverables) documented in a one-page SOP, powered by a reusable master prompt and two living templates. Per client, it produces:
Creative brief (DOCX, FR + EN) — client card, business & vision, ideal client, brand personality, visual references, gaps to clarify, recommended creative direction.
Interactive dashboard (HTML, FR/EN toggle) — live-researched market analysis: market size, digital-adoption charts, buyer-behavior data, competitor table with exploitable weaknesses, SWOT, project timeline — every figure linked to its source.
Scoping-question list — the 3–5 questions the form did not answer, each with "why it matters" — resolved in one 15-minute call instead of mid-project surprises.
Pilot results — SOYVA
Time from form to usable brief: 2–4 hours of variable quality → under 1 hour, ~15 minutes of human time.
Market research per project: ad hoc or skipped → sourced analysis: 621,803 artisan firms, 90% of buyers searching online, 6 competitors profiled.
Gap detection: discovered mid-project → 4 scoping questions flagged before design started.
Client experience: form → silence → concepts, replaced by form → an intelligence report the client can explore, in his language.
The dashboard also surfaced a positioning insight the form alone could not: the 24-point gap between consumer demand (90% of buyers search for tradesmen online) and tradesmen's social-media presence (66%) — which became the quantified argument for SOYVA's entire niche.
Why it compounds
Each run improves the asset base: the master prompt encodes what a good brief looks like; the checklist encodes what "done" means; the templates carry Nolvo's brand system. The next iteration wires the Typeform webhook so the pipeline triggers itself on submission, and adds a feedback loop comparing each brief's recommended direction with the concept the client ultimately chose. The system gets smarter with every client — the work compounds.
Takeaway
Most small studios treat client intake as admin. Treated as a system, it becomes a moat: faster starts, fewer revisions, and a client experience — an interactive, sourced intelligence report within hours of a 20-minute form — that most 10,000 € agencies never deliver.
Build the systems that build you.
The Inbound Shift System — running an air-hub operation as a designed system
CS-2026-02 · 2026 · Air cargo operations · Published anonymized: method and numbers only
How an operations supervisor at the air hub of a global logistics carrier turned an inherited overnight shift into a designed system — cutting injury frequency by roughly 6–8× and taking container data integrity from 42.71% to 98.81%.
The context
Every night, cargo aircraft land at the air hubs of the world's large logistics carriers. Inside each aircraft: unit load devices (ULDs) — containers holding thousands of time-sensitive packages. The clock starts the moment the aircraft blocks in. ULDs must be offloaded, moved across the ramp, located in the tracking system, linked to the correct unload lanes, emptied onto inbound belts, and their packages sorted — all inside a window measured in minutes, because timely arrivals and departures are the product.
The supervisor running an inbound shift here manages more than package flow: coordination with airline personnel, ground support equipment, inventory across transportation modes, safety and security protocols under aviation and transportation regulation, cost and performance reporting, and an hourly workforce whose training, evaluation, and development are the supervisor's responsibility. It is one of the least forgiving operating environments there is: heavy equipment, moving belts, hard deadlines — and a missed container is a service failure multiplied by everything inside it.
The author of this case study supervises exactly this operation. What follows is the system built there — anonymized, method and numbers only.
The problem
The shift, as inherited, ran the way most operations run: on experience, vigilance, and effort. Three structural gaps kept surfacing:
Data blindness — containers were physically moved but inconsistently registered against their unload lanes in the tracking system. Quick-location — whether a container is instantly locatable and linkable in the system — averaged 42.71%. For more than half the containers on the ramp, the data layer didn't reliably know where they were.
Effort-dependent safety — injuries were being registered at a rate of roughly one per week. Safety outcomes depended on individual attentiveness rather than designed structure — and around tugs, dollies, and powered belts, "be careful" is not a system.
Role opacity — employees executed tasks without always understanding why their step mattered to the whole: the gap that turns a workforce into disconnected hands instead of an operating system.
The principle applied
You will not out-discipline a poorly designed operation. The fix was not more supervision pressure, more reminders, or more heroic recoveries. It was architecture: a shift system founded on three pillars — safety, training, and employee involvement — where the data discipline everything else depends on is built into how the work is done, not inspected in afterward.
The system
Safety as structure, not slogan — safety practices embedded into the container-movement process itself, backed by recurrent job-methods training and immediate correction of unsafe equipment usage. The same motion that moves a ULD correctly is the motion that moves it safely — so safety stopped competing with productivity and started producing it.
Training as the multiplier — every employee group (union, technical administrative, management) trained not just on what to do but on where their step sits in the flow. When the carrier deployed a new enterprise operations platform for the ramp, this training system became the adoption engine: the supervisor ran the feedback loop with the platform's developers and coordinators, then trained all three employee groups on it.
Employee involvement as the feedback loop — frontline observations fed process adjustments continuously. People maintain what they help build; involvement converted the workforce from executors of a process into co-owners of a system.
The keystone habit: locate → link → unlink. Every ULD, every time: quickly located in the system, linked to its unload lane, unlinked when complete. One small, deliberate, repeatable action — the operational equivalent of a compounding deposit. It looks like housekeeping. It is the foundation of the entire data layer.
The results
Injury frequency: ~1 per week → ~1 per 6–8 weeks. A roughly 6–8× reduction. Safety-as-structure turned injury prevention from a poster on the wall into a property of the process.
Container quick-location: 42.71% → 98.81%. From a data layer that was wrong more often than right, to near-perfect integrity — functionally the difference between analytics as fiction and analytics as fact. Accurate container data now feeds capacity analysis, process improvement, and the enterprise platform's optimization loops.
Service and flow. Improved container movement across the department contributed to the prevention of service failures and to optimal package flow on inbound belts.
Range as a by-product. The same system thinking transferred sideways: the supervisor absorbed the irregular/restricted-freight operation by applying the identical container-movement and capacity-management discipline, and built cross-department relationships that widened the system's inputs.
Why it compounds
Data integrity is the interest-bearing account of operations. Every correctly linked container makes the analytics sharper, which makes the process better, which makes the next shift easier to run well. Every trained employee becomes a trainer. Every frontline suggestion adopted increases the rate of suggestions. Every injury that doesn't happen keeps an experienced operator on the floor, compounding the system's knowledge instead of resetting it. None of these are heroic acts — they are small, deliberate actions, repeated nightly, compounding into an operation that runs on structure instead of adrenaline.
Takeaway
The instinct in modern operations is to reach for the software first. This case ran the other way: the human system came first — safety, training, involvement — and the data discipline it produced is what made the technology layer work. A platform fed by 42.71% data integrity automates confusion. The same platform at 98.81% becomes an engine. Before you automate, standardize. Before you standardize, involve the people who do the work.
Build the systems that build you.
More case studies are in the build. One essay a week on how they're made → subscribe on the homepage.