Product design · AI features · Design system

Bonaca — A Revenue Copilot for Charter Fleets

An AI copilot that tells yacht charter operators when to lower a price — and when to wait. On the phone for the daily call, on the desktop for the weekly plan.

Try the prototype

Client

Revenue copilot for yacht charter fleets (B2B SaaS, per-fleet subscription)

Role

Product designer, end to end

Year

2026

Services

Research · Product strategy · Interaction design · Visual identity · Design system · Motion

In short

Problem: charter operators guess when to cut prices, and a wrong guess costs money either way.
Solution: one recommendation per boat, with the reason, and a decision in one tap.
My role: market research, usability studies, UX design and UI design.
Outcome: design complete and in development. There are no results yet; success will be measured by how many recommendations operators act on and how long the Monday decision takes (see Outcome).

In short

Problem: charter operators guess when to cut prices, and a wrong guess costs money either way.
Solution: one recommendation per boat, with the reason, and a decision in one tap.
My role: market research, usability studies, UX design and UI design.
Outcome: design complete and in development. There are no results yet; success will be measured by how many recommendations operators act on and how long the Monday decision takes (see Outcome).

Walkthrough of Bonaca: a Monday push, one decision per card, undo until Friday, a stalled-bookings shortcut, and asking Bonaca why a price was recommended.

Context

Croatian charter fleets rent boats by the week, and a boat is booked about sixteen weeks out of fifty-two. An empty week isn’t a gap, it’s the margin.

Challenge

When a week isn't booking, someone has to make a call: lower the price now, or wait. Lower too early and you give money away; wait too long and the boat stays empty. Croatian operators make that call earliest — peak occupancy arrives by mid-June — and with the least information.

Today's tools show numbers, not answers. They compare competitor prices and leave the decision to you. Someone already worried about an empty boat shouldn’t have to read a dashboard to know what to do.

16 of 52

weeks a charter boat is booked

1,244

islands on Croatia's coast

83%

of property managers reprice at least weekly

The question I designed for: how do you help someone make an expensive decision when they're already worried?

Constraints

The people pricing aren't revenue managers. It's the owner or base manager, between other jobs. A Monday decision had to take about ten minutes.
Trust before features. An AI that recommends a price cut is asking for money. It has to show its reasons, admit mistakes and stay easy to overrule.
Two moments, two devices. The weekly plan happens at a desk; the urgent call happens on the dock. Both had to feel like the same product.
No confidence percentages. Research showed people trusted “78% likely” blindly or ignored it. Confidence had to be shown another way.

Approach

Research first. Summer 2026 · 13 participants · interviews, a first-click test, a pilot and A/B tests. Three moments came out of it: the Monday brief, the stall (when bookings stop) and the season review.

First version: a desktop brief. One white page, one accent colour, a table of boats with a recommendation per row. It tested well for clarity — but it was a tool you sit down for, and the most stressful calls happen away from the desk.

Second version: one decision at a time. I moved the daily work to the phone and kept the desk for planning. Every screen answers one question and offers one action. Glass surfaces and a large two-tone headline set the tone; numbers stay sharp and colour is spent only on meaning.

Iterating the decision card. I tried pill rows, photo cards, a big-price card and a table. I also tried a second button, “Change”; it led nowhere specific, so it went (a different price goes through Ask Bonaca). The final card answers three questions in order:

What? The boat and the signal: “Quiet for 9 days”.
How sure? Calm, ripple or choppy water.
What to do? Now → Recommended, a one-line reason and one button that carries the number: “Drop to €5,800”, “Hold at €6,200”, “Wait until Friday”.

The reason sits on the card, so the operator never has to open anything to understand the call.

The operator's year: what research showed at each moment, and how the design answers it

One Monday in about ten minutes, and three more ways in — each one a push, not a menu item

Key decisions

1. A phone for the call, a desk for the plan.

I explored a desktop-only tool and a phone-only app, and chose both — with one job each. The Monday brief and the season review need overview; a stalled week needs a decision on the dock. Both surfaces use the same components, so a card looks and behaves the same everywhere.

Desktop brief: four boats, one row, each with the move and the reason

The same brief on the phone: one boat in focus, the next one waiting at the edge

2. The headline counts decisions, not greetings.

I tried “Good morning, Ana” and “Four need you.” I chose the count, because the first thing an operator wants to know is how much work is waiting. The card title carries the signal (“Quiet for 9 days”, “Two enquiries”), not “needs a decision” four times over.

3. Confidence as a sea state, not a percentage.

I tested percentages against shapes. Percentages were trusted blindly; the water shapes — calm, ripple, chop — were understood without a legend and work in greyscale. They sit next to every recommendation.

4. The answer before the evidence.

Tapping “Why this price?” opens a card whose headline is the answer — “It sits €600 above 14 similar boats” — with the boat's price shown outside the group of similar boats. The table I started with was accurate but had to be read; the strip can be seen. Ask Bonaca is also where a different price is set: type or speak, from any screen.

Ask Bonaca: tap a question, type or speak, from any screen

5. The crisis screen is the calmest one.

When bookings stop, the first instinct is red alerts. A worried person plus alarms makes a worse decision. Panic has no red and no menu item — it arrives as a push, gives one recommendation per boat, with the same card as the Brief, and one “Do all four” as a shortcut. After a decision, the card says what happened, the next boat is one tap away, and undo stays until Friday.

Panic on the phone: the same decision cards, with one shortcut for all four

After the decision: what happened, undo until Friday, and the next boat

6. The season review answers the owner's questions first.

At the end of the season an operator asks: was this a good year, and where did the money go? The review answers in that order: revenue against last season, three numbers, then the money left behind, split by cause (empty weeks, late cuts, Bonaca’s own wrong calls). Each cause says what changes next season. I first put Bonaca’s mistakes at the centre; that was my story, not the operator’s question, so they became one line of the breakdown, with each call one tap away. There is no decision here: the season is over and decisions live in the Brief. A tool that never admits mistakes never gets trusted twice.

Season on the desk: the year against last year, and where the money went

Where I was wrong: each call, what it cost, what changed

7. Events the model can't see.

A regatta in Split moves demand for one week. Bonaca spots it, shows the week next to its neighbours and recommends a price for that week only.

The event: the regatta week next to its neighbours, with the recommended price applied in one tap

Gallery

The design system

One glass surface, one type family, one button per card. Colour is spent only on meaning.

One glass system: colour, type and the parts every screen is built from

Outcome

The design is complete for both surfaces and Bonaca is now in development. There are no results yet, so this is how I'd measure it once it ships. Each question maps to a decision in the design:

Do operators act on the brief? The share of recommendations accepted or adjusted each week.
Does the card explain enough? How often “Why this price?” is opened before a decision.
Is Monday really ten minutes? Time from the push to the last decision.
Does Panic calm the decisions? Price cuts made after a stall, against last season.
Does it pay? Revenue per boat-week against last season.

What ships first: the Monday brief for one fleet (the weekly habit that builds trust), then Panic once there's enough data, then Season and an owner report after a full year.