B2B SaaS, UI — 04 / 000%EN日本語繁中
Case 04 — OneDegree Global — 2021

How might we improve the experience for SME insurers by simplifying and digitising core workflows?

Company
OneDegree Global
Role
Senior Product Designer
Platform
Web, SaaS
Period
2021

Overview

IXT is an enterprise SaaS platform that helps SME insurers digitise insurance operations — underwriting, policy administration, claims. After the 2.0 release I drove ongoing UI, interaction and motion work for the parts where a wrong number costs real money.

I translated agreed UX definitions into production-ready screens, states and specifications, working alongside a dedicated UX designer rather than owning the flows myself — and applied a UX mindset to the execution: hierarchy, task flow, and where a user could plausibly get it wrong.

Five modules — Product, Policy, Application, Claim, Member. The work concentrated on the first three.

Goal

Agreed with product & business

Let SME insurers roll out new insurance products faster, in Asian markets where financial regulation is unusually strict.

The platform’s promise was speed at scale. That put the design problem in a specific place: not making insurance beautiful, but making a regulated, calculation-heavy process survivable for the people who run it every day.

The problem

SME insurers face high operational friction and real adoption barriers when they digitise core workflows. The tooling is not the hard part — the cost of switching is.

  • Underwriting workflows are inefficient. Manual, paper-based steps and long approval chains, carried over into legacy systems with poor usability.
  • Digital transformation is expensive. It demands resources smaller insurers do not have, which is exactly why they are still on paper.

Scope

Product · Policy · Application

Five modules, three prioritised — chosen where business requirements and the daily operator’s needs overlapped most.

I worked with product managers to clarify the proposition and direction, ran competitor comparisons, and sat in on expert interviews with actuaries and insurance product managers. Clarifying the business requirement was the first priority and the one most likely to be wrong: insurance rules are not guessable from the outside.

The challenge

Insurance products carry complex premium and policy-period logic, and several products together produce interlocking rules. The constraint was to deliver clarity inside the existing framework, without growing the design system to fit.

Every new pattern is a cost paid twice — once in design, once in engineering. So the question on each screen was whether an existing component could carry it.

Process

Definition · UI · States · Prototype · Reuse
  1. Align on workflow and rule definitions
    Confirmed decision points and exceptions with PMs and UX early, where rework is cheap.
  2. Translate flows into production-ready UI
    Structured dense forms and tables with clearer hierarchy, grouping and progressive disclosure.
  3. Define UI states, validations and edge cases
    With engineering: validations, warnings, confirmations, loading, empty, error. Calculation steps had to be explainable, not just correct.
  4. Use prototypes to reduce ambiguity
    Lightweight prototypes and motion studies for the things a static screen cannot settle — table interactions and step-based flows.
  5. Scale with reusable patterns
    Documented typography, spacing and component usage so delivery stayed consistent without the system expanding.

The frame

What every module inherits
A consistent page frame

Every module had to open on the same frame, so a reader arrives already knowing where to look.

Workbench pages are long and information-dense. Staying oriented is not a courtesy here — a reader who loses their place has to rebuild it before they can do anything else.

Canvas
Left navigation and a central content canvas with generous spacing.
Context header
Entity identity and status stay visible throughout — policy or product ID, current status, key timestamps.
Sections
Card-based blocks break dense content into scannable units, with tables only where a list actually needs comparing.
Product workbench
Product workbench
Workbench frame02
A separate view mode

A review state has to say read-only, not merely be read-only.

After editing, the work gets checked, shared and confirmed. A screen that is safe but does not look safe still gets handled as though a keystroke could change something.

Reading order
Information presented as structured blocks to be read top to bottom, rather than as a form standing still.
Telling
Disabled fields and read-only styling carry the state, so nothing depends on the user remembering which mode they opened.
Grouping
Content grouped by hierarchy, so key facts can be verified without scrolling through raw form fields.
Version history
Version history
Review surfaces03
Navigation and version state

Navigation had to answer two questions at once — where am I, and am I editing the live version?

Decision-heavy work means jumping across areas constantly, and a product is not one thing but a run of versions. Neither question should cost a detour to answer.

Rail and list
A fixed icon rail for global areas and an inner section list for the current workspace, collapsible when dense content needs the room.
Predictability
Consistent labelling and hierarchy, including expandable groups, so a user can predict where a setting lives.
Version state
Going to launch, launched, shelved — the selector puts the state inside the label rather than beside it, so the question is answered by the control the user was already reaching for.
Product version — launch states
Product version — launch states
Navigation and version state05

Reading

Review first — the state every screen opens in
Policy servicing

Servicing opens in review, and every action is something you reach for deliberately.

Servicing means high-impact operations on a live policy. Whatever state the screen opens in is the one most work gets done in, so the default decides how easy it is to change something by accident.

Context
Users need immediate answers to what policy is this, what is its status, what period am I looking at. A summary header carries the policy number, primary and secondary status, key metadata and a Change Status action.
Actions
The default is review — readable card sections — with actions pulled out: global at top right, row-level in tables, contextual where they belong. Accidental edits get harder.
Rhythm
Every tab follows the same pattern: policy context, section navigation, card blocks, then tables for dense lists. A predictable shape across Policy Parties, Policy Details, Billing and Excluded Item.
State
Effective against Pending, Overdue against Paid, renewed relationships. State carries clear labels and badges, exceptions surface their own signal, and status changes live in consistent places.
Billing
Billing is designed around what needs attention now — next payment and method up top, then a full schedule with status tags for reconciliation. Excluded Item is an exception workspace with row-level edit and delete.
Policy parties
Policy parties
Servicing surfaces04
Plan viewer

Contract-like content had to become readable without simplifying away detail that is legally load-bearing.

Plan content is dense and nested — main clauses, conditions, riders. None of it can be thinned out, so the only thing left to change is how it is set.

Hierarchy
A document-style viewer with real typographic hierarchy and spacing, rather than a form holding legal text.
Grouping
Grouping that reflects the clause relationships instead of flattening them into a single run.
Core plan
Core plan
Plan viewer02
Advanced filter

Filtering had to stay reversible — no losing your place, no accidental commits.

Workbench lists need flexible filtering, and a filter is something a user tries several times before it is right. The cost of a wrong one has to stay near zero.

Panel, not page
A drop-down from the header, so the list context it is filtering survives the act of filtering.
Structure
A Field × Operator × Value structure that scales to multiple conditions, with autocomplete and type labels to prevent mismatches.
Safe iteration
Apply and Cancel, so trying a filter is not the same act as committing to one.
Advanced filter
Advanced filter
Filter panel
Date picker placement

Acting

Where a wrong number costs money

Everything above is what the platform does while nobody is changing anything. What follows is what it does when somebody is — which on this platform is another way of saying where the mistakes get expensive.

Editing screens

Complex input had to be hard to misconfigure — without anyone relearning the pattern in every module.

Many fields feed calculations and downstream behaviour, so an error here is not a typo — it is a wrong number that reaches a customer.

Hierarchy
Long forms structured into consistent sections and blocks; headings, spacing and alignment say what belongs together and what depends on what.
Reuse
Dropdowns, toggles, add-another rows, inline help and table actions standardised, so the same interaction behaves the same everywhere.
States
Editable against read-only, required against optional, validation and warnings — the difference between a mistake caught and a mistake shipped.
Underwriting — edit
Underwriting — edit
Editing surfaces04
Rule editor

Rule authoring needed its own workspace that still belonged to the platform.

Advanced rules are written, not configured. The workspace had to balance focused writing against fast attribute discovery, and keep errors readable while both are happening.

Layout
Left tools and right content, matching the workbench so context switching stays cheap.
Discovery
A collapsible attribute selector — expand to search, collapse for focus. Two patterns behind it: hierarchical layer filters for browsing, search with type cues (Num, Enum) for speed, and insert actions to cut copy-paste.
Feedback
Compile result and syntax guidance in their own area, so debugging never leaves the page.
Surface
A dark editor against lighter UI, which says this is code without a label.
Rule editor
Rule editor
Rule editor05
More of the product module
Rule lists, and the pricing table behind them

What these are
The underwriting rule list in its three states — nothing yet, a draft, one that has launched — the same list in a denser view, and the Excel-shaped premium table that pricing is authored in.
A note
The original case put these five screens under the heading “More product module screens” and said nothing else about them. They are here for the reason they were there: they are part of the module and they were built. Nothing is claimed for them that the source did not claim.
Rule list — launched
Rule list — launched
Rule lists and pricing06
Cancellation wizard

The highest-stakes operation on the platform had to make every decision explicit and reviewable.

A small mistake here cascades — wrong refunds, wrong billing, mismatched documents — and it is usually made under time pressure, halfway through a partial cancellation.

Steps
Three steps — target selection, cancellation info, refund calculation — with a persistent progress indicator.
Disclosure
Progressive disclosure keeps the default minimal: rider selection appears only for a partial cancellation.
Attention
A centred single column with Optional labelling, guiding attention to the fields that matter.
Verification
A refund preview with a plan-by-plan breakdown before totals, so the number is checked before it is committed.
Target selection
Target selection
Cancellation flow04

Design system

Area → Fundamentals → Components & Patterns

As the platform spread across modules, we needed a shared UI language that could scale without adding rework. I built the visual foundation from the corporate identity, then turned it into rules teams could apply.

  • It mirrors the real product structure. Area, then fundamentals, then components and patterns — so the guideline is navigable the way the product is.
  • Area-level rules gave cross-module consistency. Navigation, headers and content regions behave the same across workbenches, which is what prevents drift.
  • Components shipped with their rules. Layout, states and usage examples together, so a complex form is assembled rather than redesigned.
  • Documented interaction reduced handoff friction. Selection behaviour, action placement, validation states — written down, so implementation stopped being interpretation.
Area
How a workbench is composed — the rules that stop drift between modules.
Area
Area
Pages05
Fundamentals
Colour, type and icons, built from the corporate identity.
Colour
Colour
Pages04
Components & patterns
Each shipped with its layout, states and usage rules — which is what made implementation stop being interpretation.
Fields
Fields
Pages21
All 21 sheets

Marketing site

Motion, and a page that moves as you read it

The platform has a public face as well as a product one. Two pieces of it: the motion that carries the brand, and a landing page whose sections arrive as you scroll rather than sitting still waiting to be read.

Both are the same argument the product makes, aimed at a different reader. A workbench earns trust by being predictable; a marketing page earns attention first and has a few seconds to do it. The motion is what buys those seconds, and the scroll design is what spends them in an order somebody chose.

Brand motion
Scroll design

Outcome

The platform shipped and kept shipping. What the design work bought was a floor under the complexity — new modules could be built without renegotiating how a form behaves.

  • Clarity in decision-heavy workflows. A consistent information hierarchy and interaction structure made complex operations easier to scan and complete.
  • Consistency across the platform. Standardised components and usage rules cut confusion and shortened the learning curve.
  • Fewer input errors. Clear UI states and feedback patterns — editable and read-only, required and optional, validation, warnings — strengthened confidence in calculation-sensitive work.
  • Delivery that scaled. Close alignment with PMs and engineers on feasibility, edge cases and implementation kept the platform moving as it grew.
These are the outcomes the original case claimed. They are qualitative — no measured before-and-after was published for this project, and none is invented here.