← All work Alekhya.
Product Design Β· Mobile + Web

Cravant-i, a maritime food management app

Cravant-i mobile and web app overview

A role-based mobile + web product that lets a Chief Cook plan, inventory, and order food for a 24/7 offshore vessel in a few taps, and gives a shore-side Purchasing Director real-time budget control over the whole fleet, replacing the Excel-and-email chain nobody has disrupted.

Role
Product Designer at Creatum GmbH, sole designer on the MVP
Responsibilities
Stakeholder research Information architecture Interaction design Mobile + Web UI Prototyping & testing Dev handoff
Project
Cravant-i MVP for Sea World Management S.A.M. (SWM) pilot
Timeline
R1.0, pilot July 2026, market launch MTB Istanbul Oct 2026
Tools
Figma, Miro, Notion, Confluence, Maze
Context

The user works in a galley on a moving ship, with intermittent connectivity, in a dozen languages, under a captain.

Every commercial vessel at sea runs its catering on Excel. The Chief Cook is responsible for inventory, meal planning, procurement, and crew satisfaction, for weeks or months of a voyage, without a reliable connection to shore and without a user interface designed for a galley.

Cravant-i is Creatum's product to replace that stack with a role-based mobile + web platform. I came on as the product designer for the Release 1.0 MVP, working with product leadership at Creatum and SWM's team, who were piloting the system on a 24/7 offshore cable-laying vessel crewed predominantly by Italians.

My job: translate a dense maritime domain (vessels, ROB, parent-level SKUs, multi-stage approval, port price index) into two interfaces, a mobile app a chef can use one-handed in a swaying kitchen, and a web dashboard a Purchasing Director can trust to make budget decisions in London.

Cravant-i stakeholders, roles and context
The Problem

Six roles, two platforms, one dataset, and not a single UI that treats the Chief Cook as the primary user.

Before any screen, I mapped the six people the product had to serve and what each one actually did with it:

  • Chief Cook (mobile, primary) β€” the daily driver: inventory, recipes, meal plans, requisitions.
  • Captain (mobile + web) β€” approver and escalation point, in the galley and in his office.
  • Ship Crew (mobile, optional) β€” menu viewer, feedback.
  • Purchasing Director (web, primary) β€” fleet-wide budget and governance.
  • Ship Owner (email) β€” rare, high-value approvals above threshold.
  • Supplier (email β†’ API later) β€” receives the Excel export.

Existing maritime tools were built for shore staff, not chefs. Interfaces were dense, desktop-style tables, English-only, assumed stable connectivity, and buried approval status three menus deep. On a workshop with SWM (22 May 2026) and an alignment session (01 June 2026), three design-level pain points kept surfacing:

  • The chef is doing data entry on a tablet while cooking. Big targets, high-contrast, one-handed, zero typing.
  • The vessel runs 24/7, so "breakfast / lunch / dinner" is wrong at the data model level. Every existing tool forced three meal types; our planner couldn't.
  • The Purchasing Director doesn't want detail, they want the one vessel over budget. The web app had to open on an answer, not a filter.

The brief, for me: design a product where the Chief Cook finishes the morning inventory in under 5 minutes, and the Purchasing Director sees the fleet's budget health at a glance.

User roles and primary surfaces per role

User-role map, six actors, two platforms, and the one surface each role needs to feel the product is "theirs."

Research

The insight that reframed the whole product was not technical. It was emotional.

Across the workshop + alignment meeting, I sat in on 14 hours of stakeholder conversation, kept a running log of verbatim quotes against each user role, and clustered them in Miro into an affinity map. The quotes mattered more than any slide:

"It will be on 24 hours per day, so the system need to be open in this sense."SWM Captain, Workshop Transcript 22 May 2026
"For SWM cable-laying vessel full of Italians, if they want to have Barilla number 5, they go on strike."SWM Purchasing Director, alignment meeting 01 June 2026
"Just for ice cream and the Red Bull, if you do not supply, the crew will be really upset."SWM Captain, Workshop Transcript
"We want to spend X per month. If required more, require approval of management or from the owner."SWM Purchasing Director, Workshop Transcript
"There needs to be a quality staging for injecting new line items into the overall catalog. Human in the loop."Workshop Transcript, catalogue governance discussion

Underneath all five quotes: SWM wasn't buying features, they were buying the absence of anxiety. No anxiety that the budget was being blown. No anxiety that the crew would mutiny over the wrong pasta. No anxiety that the chef forgot the Red Bull. Every screen, every state, every empty state had to be designed to subtract worry, not add it.

That became the north star for the whole visual language: calm, confident, confirming. No ambiguous states. No "pending" without a next step. No red without a fix.

Affinity mapping and journey synthesis in Miro

Affinity map clustered into two primary personas (Chief Cook on the vessel, Purchasing Director onshore) and the three recurring jobs-to-be-done each one came to the product for.

Design Decisions

Six design decisions that shaped the Release 1.0 product.

1.

A single pre-filled requisition cart, not a search-and-build form.

The existing industry pattern is a blank order form you populate line-by-line. In workshop testing with a chef, that took roughly 40 minutes for a weekly order. I redesigned the requisition as a review, not a build: the system takes Meal Plan + ROB deficit, generates the cart automatically, shows each item at parent level with the shortage math visible ("45kg needed, 10kg on board, add 35kg"), and the chef approves, edits, or adds. The IA treats the cart as a draft that already thinks it's done. Completion is a swipe, not a form submission.

2.

A calendar planner without time slots, built around meal cards.

Every maritime catering tool I studied laid out the day as three fixed rows: breakfast, lunch, dinner. On a 24/7 vessel that layout is a lie, watchkeepers eat at 02:00 and 14:00. I replaced the three-row grid with a free-form meal-card stack per day, each card carrying recipe, headcount, and an optional meal-type tag. The ROB math drives off ingredients and portions, not time slots. The visual payoff is calmer too, days with 3 meals look like 3 cards, days with 7 look like 7, no empty rows, no forced structure.

3.

Budget status as a living band at the top of the fleet dashboard, not a report you run.

The Purchasing Director's first job every morning is "is anything over budget?" An old-school BI dashboard answers it in three clicks (filter vessel, select month, open budget tab). I made the answer the first thing the page shows: a horizontal band of vessel chips, each coloured by budget utilisation (green under 80%, amber 80–100%, red over 100%), sortable and drillable. The question becomes the opening screen. Detail is one tap away for only the vessels that need it.

4.

Brand specification as a line-item chip on the requisition, not a recipe field.

"Barilla no. 5 or the Italians go on strike" is a requisition-time constraint, not a recipe-time one. Putting brand at the recipe level would have forced parallel recipe libraries per vessel. I designed brand as an optional chip that sits on the requisition line, tap to pick from history, free-text comment for substitution notes. One recipe, many vessels, the brand travels with the order, where the supplier can act on it.

5.

Snacks & drinks as a parallel inventory flow, not a meal-plan workaround.

Red Bull and ice cream are not recipes, but they're $15k+ per order. Forcing them into the meal planner was the fast engineering answer; the design cost was a UI that hid a $15k category behind the wrong mental model. I split the IA: a dedicated "Snacks & Drinks" tab on the inventory, its own reminder cadence, same requisition cart at the end. One product, two inventory models, surfaced in the navigation so the chef never has to pretend a Red Bull is a lunch.

6.

Approval as a visible status chain on the requisition itself, not a notifications tab.

Multi-stage approvals (Chief Cook β†’ Captain β†’ Purchasing Director β†’ Owner) are the single biggest source of "where is my order" anxiety on a ship. Existing tools put approval status behind a notifications bell. I designed it as a four-node status chain that lives on every requisition card, each node showing current state, who owns it, and (if stalled) how long it's been there. Chef taps a stalled node, sees who to ping. The audit trail and the status UI are the same data; the design just surfaces what was always in the DB.

Cravant-i mobile app, Chief Cook view

Mobile app (Chief Cook), 5-screen daily journey, home, incoming stock, cookbook, meal planner, pre-filled requisition cart.

Meal plan calendar view ROB inventory view

Left, meal-planner with unlimited meal cards per day. Right, ROB inventory with current + Planned ROB projection and shortfall chips.

Pre-filled requisition cart with brand chip

Pre-filled requisition cart, parent-level items with optional brand chip per line, visible approval status chain below each requisition.

Web App fleet dashboard for Purchasing Director

Web App fleet dashboard, budget band on the top opens on the answer: which vessels are green / amber / red right now.

Prototyping & Testing

We prototyped the mobile flows in Figma and ran them past SWM in two rounds before any pixels were locked.

Two rounds of interactive Figma prototypes went to SWM's Chief Cook, Captain, and Purchasing Director between the workshop and the June alignment. I framed each round around a question, not a screen set:

  • Round 1: "Can the chef complete a weekly meal plan + requisition in under 10 minutes?" Target hit at 7 minutes on the second attempt, down from ~40 in Excel. Surfaced that the auto-generated cart needed a visible explanation of why each quantity was there, otherwise the chef felt the app was doing magic they couldn't audit.
  • Round 2: "Can the Purchasing Director spot an over-budget vessel in under 15 seconds on the dashboard?" Target hit at 8 seconds once the budget band was coloured by utilisation instead of showing raw numbers.

Outputs fed directly into a Figma component library with 42 components, documented states (default / loading / empty / error / success), and a design-token layer for cross-platform parity between Mobile and Web.

What Didn't Work

Three designs I built and had to kill.

  • A rich onboarding tour for the Chief Cook. I built a 6-step guided tour on first launch to explain ROB, parent-level items, Planned ROB, and the approval chain. In testing, both chefs tapped Skip. They wanted to start cooking, not learn the vocabulary. Replaced with inline tooltips triggered on first hover of each unfamiliar term, used 3x more often in testing.
  • A clickable fleet map on the web dashboard. Early concept had vessels pinned on a world map (seemed visual, geographical, maritime). Testing showed the Purchasing Director never asked a geographical question, they asked a budget question. The map was a decorative distraction. Replaced with the vessel band sorted by budget status.
  • A separate "alerts" tab for budget breaches. I designed a notifications inbox for over-budget and over-threshold alerts. During testing the PD opened it twice in two weeks; everything they needed was already on the dashboard home. I collapsed alerts into a single counter chip on the home nav and killed the tab.
Reflection

The Release 1.0 design handed off to engineering in June 2026, pilot shipping July.

Design outputs at handover:

  • Two platforms (Mobile App for vessels, Web App for shore) designed from a shared Figma library of 42 components, 8 feature groups, 3 user roles, with all 5 states documented on every component (default, loading, empty, error, success).
  • Chef task time on weekly planning + requisition: prototype testing 7 minutes vs ~40 minutes in Excel (82% reduction).
  • Purchasing Director time to spot an over-budget vessel: 8 seconds on the dashboard vs multiple clicks on existing BI tools.
  • Design-to-dev handoff: single design-token file powering both Mobile and Web, zero spec-sheet translation needed.
  • Design system scoped as the foundation for Release 2.0 (AI meal suggestions, Bonded Store, fleet analytics) without re-work.

What I carry forward: in a domain this dense, the designer's job is to subtract vocabulary, not add visuals. Every design decision that worked came from a word the user was already using (cart, chip, card, status). Every one that got cut (onboarding, map, alerts tab) came from a word I had imposed.

Limitation: the prototypes were tested with 3 SWM stakeholders, not a broader crew sample. Real-world shipboard conditions, low light, glove-wearing hands, interrupted connectivity, could only be approximated in the Figma environment. Post-pilot usability observations in July will be the honest validation, and they sit after the handover window.