Skip to content
gposse
ESEN

← Back to workRedesign concept · Value proposition, identity and system · 2024 · 2026

Bond

PomJuice is a 2024 platform for startups to present themselves to investors. I redesigned it with a single value proposition, a new identity and a design system: that became Bond.

---

role:
Product designer: value proposition, identity and design system
starting point:
PomJuice, which I designed at MARL Accelerator
outcome:
Bond: brand, system and product, as a personal concept
year:
2024 · 2026

---

bond.mp4

PomJuice did a bit of everything. In the redesign, I looked for the one thing it had to do well.

Where it started

PomJuice was MARL Accelerator's platform for startups to present themselves to investors. I designed it in 2024, for desktop and mobile.

Looking at it again two years later, the problem was clear: it tried to do too much. A startup page you filled in by hand, a dashboard of visits and deck views, a feedback form with stars and five sliders. It had no system either: every screen had been solved on its own.

What founders said

PomJuice bet on DocSend-style metrics: how many visits, how many deck views. While designing it I talked to founders, and those metrics barely helped them.

What they wanted to know was what investors thought of their startup. Almost nobody finds out why they got a no, and without that reason it's hard to know what to improve.

One value proposition

The value proposition became: founders and investors get to concrete answers fast. Founders learn how much to raise and why they hear no. Investors find startups that fit their thesis, with confirmed data.

I cut everything that didn't serve that. The founder's journey ends in one place: understanding each answer and what to do next.

Before and after

The same tasks, before and after. On the left, the screens I designed for PomJuice; on the right, Bond.

From PomJuice to Bond

Once the proposition was set, the name stopped fitting: PomJuice said nothing about the product. Bond means a link, and that's what it does: connect a founder with the right investor.

The identity comes from that idea: two crossing rings, lime on charcoal and lowercase Space Grotesk. Lime only appears on actions: confirm, connect, reply.

A design system in code

So the product wouldn't grow screen by screen again, I built a design system in code: color tokens with one role each, a type scale and the components every screen is made of, in Spanish and English.

I built it with AI agents. I set the rules and the references; the agents helped me implement them and apply them across the product. That let me iterate the whole product in days, and a new feature now comes together from pieces that already exist.

How I designed the answers

  1. 01

    Answers with sources

    In a fundraise, an assistant that sounds sure but makes things up is worse than none.

    1. Option AA free chat that answers in paragraphs.
    2. Option BStructured answers: the figure, what weighs most, the source and what isn't certain.

    What I chose

    The second. Every answer shows where it comes from and states its limits (a range, not a promise). And it ends in a concrete action, like building the round plan or requesting an intro.

  2. 02

    Data the founder confirms

    An assistant is only as good as its data. If it drafts a profile with mistakes, it answers with mistakes.

    1. Option ALet the assistant fill in the profile alone.
    2. Option BLet it read the deck, draft the profile and have the founder confirm every fact.

    What I chose

    The second. Nothing reaches an investor unconfirmed and every fact shows where it came from. If something is missing, the assistant says so instead of filling it in.

  3. 03

    A no with a reason

    What came up most in the conversations: founders almost never learn why they got a no.

    1. Option ALet the assistant guess the reason.
    2. Option BAsk the investor for the reason in one tap, anonymously, and use it to answer the founder.

    What I chose

    One tap for the investor. With those reasons, the assistant groups what keeps coming up and suggests what to change in the deck. The original words stay one click away.

01 / 03 · Investors ask for what they need and get startups with confirmed figures, the source, and what the assistant doesn't know. Sample data.

Takeaways

  1. 01

    Cutting is design work too

    PomJuice had too many features and no clear proposition. Picking one job, learning why you hear no, and cutting around it sorted out everything else.

  2. 02

    Start from what users say

    I also thought visit metrics were the way. The founder conversations said otherwise: the feedback was what mattered. Bond starts there.

  3. 03

    A system stops you deciding the same thing twice

    Without a system, every screen is decided from scratch and inconsistencies pile up. With tokens and components, adding something new is fast and matches the rest.

  4. 04

    With AI, trust is designed too

    Showing the source, asking for confirmation and saying what the assistant doesn't know are interface decisions. Without them, users have no way to tell whether an answer holds up.