FinTech Product design 04 April 2026

Zenxo

An F&O strategy builder and backtesting platform. Retail options traders assemble a multi-leg strategy and test it against years of historical market data before they put real money behind it. The strategy builder was nine form sections in one scroll, and it couldn’t express the relationships traders actually needed.

  • RoleUI/UX Designer
  • TimelineMay – Aug 2026
  • Team2 PM · 3 designers · 10 engineers
  • PlatformWeb
Zenxo strategy builder
Context

A strategy is a structure, not a form

A strategy runs four or five legs at once. Each leg carries its own entry, stop loss, target and re-entry rules, and strategy-wide rules sit on top of all of them. Some rules belong to a single leg. Others are shared across several — one stop loss governing two short legs, entry conditions shared across a group.

Traders kept asking for the same thing, and it wasn’t a shorter form. In a form, a shared rule is a checkbox list inside a panel, and it disappears the moment you close the panel. With three or four shared rules you can’t audit the strategy at all.

Research notes from trader conversations
Research · trader conversations
The user is rarely a professional quant, so the product has to be powerful enough to be real and clear enough to trust.
Problem

Nine sections. One scroll.

You hit twenty-five controls under “Overall Settings” before you described a single leg. Nothing told you where you were or how much was left. Legs sat below as stacked, independent form blocks.

The deeper limit wasn’t length, it was capability. A form can list settings; it can’t show relationships.

Traders kept using one word for it: overwhelming.

Before · the nine-section scroll
Approach

Draw the shape of the strategy

Entries sit above, legs run across, exits sit below. A shared rule becomes a single node wired to its members and labels itself with how many legs it covers. A leg with nothing beneath it is visibly a leg with no exit.

Changing one number on a canvas is slow, and traders optimise for speed — so the form stayed the default and the canvas became its live mirror. It can’t drift from the form, because it reads the same data.

Wiring a shared rule · the node labels itself
Say it, and the canvas builds itself

Zenbot turns a sentence into structure. Ask for a 30% stop loss on both legs and the canvas acknowledges the moment you hit send, then the blocks arrive one after another, already wired onto the legs they belong to. No form to fill in, no mode to switch into.

Decisions

Two calls that set the shape of the product.

Decision 01

More freedom was the version nobody would learn

The empty + Exit slot doubles as the affordance and the warning.

  • Option A · Drag-to-reveal connectors

    Expressive, and closer to how node editors work. But it puts a learning curve in front of the thing meant to simplify the product.

  • Option B · A button on the block itself

    Click + Exit and an exit appears. Less elegant, zero discovery cost.

  • Chosen · B

    Every leg carries its own + Entry and + Exit, so the next move sits on the thing it applies to. The interface teaches itself as you use it.

Decision 02

A form can list settings. It can’t show relationships.

The split screen is the product’s defining layout: you see the structure while you build it.

  • Option A · Form plus grouping checkboxes

    Cheapest to build, and what every competitor ships. Close the panel and the relationship vanishes.

  • Option B · A canvas of nodes and wires

    A shared rule becomes a permanent shape. Costs a new interaction model, and a much harder build.

  • Chosen · B, on a split screen

    A narrow form was the harder sell — a wide form makes the eye travel, and every trip costs time. That reclaimed space is what lets the canvas sit beside it.

Results page

Twenty-one panels deep, still guessing

Every card had the same border, the same padding, the same importance. Net profit sat in a tile the size of the worst losing streak. Scroll far enough and a pink box said the strategy would lose money over time, in the same treatment as everything above it.

A trader opens a backtest to answer one question: should I run this.

A descending funnel

The headline verdict comes first, then shape over time, then patterns, then the raw evidence. The reader picks how deep to go. Exit reasons are broken out by type, and insights are labelled by their coloured tab before you read a word.

The data shown is demo data. The backtesting engine is the backend team’s work.

Before · twenty-one equal panels
After · the page walked top to bottom
System

Nothing decorates

There’s one accent colour, and it’s rationed: orange means “act here” or “you are here”, and nothing else. Motion reports state — the orbiting dots stall when the strategy does, and a pulse runs along the wires when a rule is grouped. The assistant has intent instead of jargon. The form and the canvas use the same components because they read the same data. 26 strategy templates, 24 supported brokers.

What’s next

Overlapping shared rules currently drop to a second row. It’s readable, but it’s the layout at its least elegant. Smaller screens come next: the form column is fixed, so the canvas is the part that has to adapt to different viewport sizes.