BestTrip

02 — Brand

The one thing I own outright.

The brandbook was the first deliverable of the mandate and shipped six weeks in. It was not a restyle: I defined the audience, wrote the tone of voice from research into how the company already positioned itself, expanded the palette and type, set the illustration style and chose the icon set. The CEO approved it directly, with no design or marketing layer between. Version two followed in March 2026 with everything made since, plus a trimmed PDF cleared for clients.

Decisions

• Audience first, visuals second. With no defined audience, every aesthetic argument was unfalsifiable. Defining it — travel managers, secretaries, finance, directors — let the rest of the book be argued on evidence, not taste.
• The illustration style was tested, not preferred. Flat, 3D, gradient and photography were all options. I put them to real users at eight client companies, in the roles that use the product; flat with gradients and effects tested best, and that is why it won.
• Photography was dropped over the law, not the look. Images used to be pulled off the open web — real legal exposure for an accredited IT company, and nobody had raised it. I took it to leadership as a question of liability and set up an approved stock list the company can buy from. It is rarely used, since illustration is faster, but a compliant route now exists.

03 — Design systems

A system with a history, not a component library.

The first landing page had a one-week deadline. The cheap move was one bespoke page; instead I built a base design system under it, derived from my brandbook so brand and product would connect by design rather than by luck, and kept the full process under the compression: UX map, flows, style directions, structure, then the page. I have maintained and extended it ever since.

Four months later a landing page for another brand in the group shipped in two days, built entirely on that system. The week it cost was repaid by one project.

In February 2026 a second generation brought the company’s graphic design into the same system, so brand and product are one artefact rather than two kept in sync by hand, with developer pages that carry the code and the system files for engineers to take directly, instead of translating from Figma.

The third generation is written to be executed rather than read: tokens the build consumes and rules a generator builds pages against. The tender sites already run on it; the multi-page site replacing the first landing is next.

Four systems share these roots: product, marketing site, presentations and print, and one for a second company in the group. None was assigned; each was built inside delivery work, so the next piece would be cheaper.

04 — Product

Rules first, then the screens: 419 of them.

The interface broke on foldables, at resolutions the layouts had never been drawn for. There was no settled practice to copy, and fixing screens as they broke would have produced as many answers as there were screens.

So the rules came first — how elements reflow, what scales, what happens at the fold — from Samsung’s and Google’s guidelines, fitted to the interface the apps already had: foldables were too small a market to pay for a rebuild. Then I built 419 screens across the B2B and B2C apps against those rules instead of deciding each by hand. The rules from October 2025 still govern foldable work a year on.

Alongside it

Ten product sections needed empty states. Instead of ten one-off illustrations, I set the principle — draw the section’s empty container, never an abstract symbol — then the series rules: one main object and at most one supporting, a single accent, a detail budget set by the 144–160 px render size. Components were named by screen, not by what they show, so a redraw keeps the name. The rules did the design; generation only executed them. A later maintenance notice was drawn to the same rules, outside the set.

05 — Production pipeline

Templates that assemble themselves.

Much of what reaches graphic design has the same shape every time, so I turned it into automations: templates taken one step further, with rules, components, spacing and structure written into files a model assembles from. Bring the content, and the template builds itself.

Decisions

• One automation per task, not one for the company. A single company-wide setup came first: the model mixed up files, the context outgrew it, similar components collided. Precision matters more than reach, so each kind of work has its own rules.
• The depth of the input follows the job. Our own materials get thorough rules; a tender gets a few shared ones, with the design built from the client’s brandbook, site and logos.
• Everything is built in HTML, the format the model handles most reliably, then goes to PDF, PPTX or Figma, so any result can still be edited by hand.
• Where the tools stop is part of the method. Print is prepared for press by hand in Illustrator, PPTX exports are checked, every deck is reviewed by a designer, and production artwork is vector: generated images can’t hold small text or exact positions.

About twenty-five presentations have gone through it. The tenders are for organising events for large companies, where each pitch must look like the client, not like us; their sites come from the rules in 03.

The habit is older than the tooling: I build working HTML from my own designs, so rendering is checked at once. The company digest has run on it since December 2025, BestTrip’s itinerary receipts replaced a supplier format that could change at any time, and at a conference won by tender the venue’s schedule boards were HTML pages that followed the clock.

Some of it already runs without me. Other departments proofread their documents with the automation, against my rules; business units use a PDF tool I built for them; and the designer I hired has worked on the automations from the first weeks.

06 — Where it stands

Still building — and some of it has outlived me already.

The product is early and has no metric culture yet, so there are no conversion or retention numbers here, and I would rather say so than invent one. What can be stated is what held: the foldable rules from October 2025, the digest running since December 2025, and the clearest — I launched the company’s social presence from zero on five networks, style, tone and every asset, then handed the rules and templates to an outsourced team that has stayed on brand without me since.

That is the outcome I would point an interviewer to, because it can be checked: the system kept working after I stopped touching it. A second test came when the designer I hired started on the rules, templates and automations already in place, not on habits that lived in my head.

Next is depth: flows taken past first-build quality, the multi-page site on the third-generation rules, and more of the automations in the hands of the people who use them.