deal express is a global bank's crm for selling. sales teams and product owners use it to move a prospect from first conversation to signed deal, across offices from london to chennai.
using it was hard work. a simple deal update meant hopping between screens, and each product's form asked 246 questions, much of it information the bank already held. people made mistakes, support tickets piled up, and every new product took two weeks of front-end work.
lead ux architect, 2018 to 2024. i owned the ux end to end: research, redesigning the workflow, the design-system foundations and the handover to code. i led and mentored three designers on the account and partnered with four engineering squads, working daily with sales users, approvers and product owners.
questions per product: 246 → 168 (32% fewer). in testing after the redesign: 56% fewer usability issues, 68% fewer data-entry errors, 47% fewer support tickets. 40% of fields filled in automatically. a new product's front end: two weeks → four days. built to wcag 2.1 aa, checked by internal audit at component level.
measured in q4 2022 against the pre-redesign baseline, from the service-desk report and 8 usability sessions.
a global crm is an ecosystem, not a screen. getting five time zones to agree took as much design as the interface did. next time i'd put users in the room sooner; product owners had been speaking for them.
a 2026 reimagining of the deal dashboard, keeping only the original structure and layout.
drag, or use the arrow keys,
to compare the dark and light themes.
the ai insights panel is a 2026 addition, not part of the original delivery.
the core finding: the system kept asking people for information it already had. the repetition and ambiguity users described were not screen-level usability issues; they were systemic communication problems between products, regions and teams.
baseline, by cognitive walkthrough across scenarios: creating a deal and submitting it took 10 minutes and 56 mouse clicks, counting adding a product to the deal, not configuring it.
screen 1 / 6
hover a heuristic to read its score for the screen on show
each product had added its own flow over years, so long, repetitive forms made people re-enter known client data. errors followed.
sales users spent more time feeding the system than selling, and confidence in the data dropped.
without clear status, people chased approvals by email and phone, which drove support calls.
product owners in separate time zones made decisions for users without them, and developers couldn't see the intent behind screens.
drag, or use the arrow keys, to compare. the reimagined form groups products into cards, moves client context into a side panel and prefills what the bank already holds, so people confirm rather than re-enter. reimagined in 2026, keeping only the original structure and layout.
research pointed to three decisions. each card shows what i chose, what it cost and what it changed.
"unified design language across crm modules."
i established the core foundations of the design system (colour, typography, spacing, elevation and interaction tokens) mapped directly to production code. these foundations created a single source of truth across all crm modules, enabling visual consistency, accessibility, and rapid ui assembly.
front-end teams cut a new product's build from two weeks to four days using reusable components and templates. a governance model and bi-weekly design guilds kept regions aligned and stopped design drift.
research cadence
constraint
product owners in five time zones decided for users, not with them
what i changed
i ran repeated sessions with the business to argue for meeting users directly, then set up periodic surveys and remote walkthroughs so feedback kept arriving.
outcome
users were in the room
and the re-entry problem surfaced within weeks.
shared decisions
constraint
prefill was not mine to decide alone
what i changed
i ran question consolidation workshops with product owners in every region, and worked with data analysts on historical data-flow trends to establish what could safely be pre-populated.
outcome
246 questions became 168
with 40% of fields filled from what the bank already held.
design system
constraint
engineering was split by region and by vendor, each with its own code practice
what i changed
i defined shared colour, type, spacing and interaction tokens with accessible colour standards, and globalised the layout and interaction patterns.
outcome
one visual language across the crm modules
with accessibility built into the components.
platform delivery
constraint
the front end ran on a vendor platform none of us controlled
what i changed
i learned the vendor's standalone tool and spent three months integrating the design system into its core css library rather than styling screen by screen.
outcome
a new product's front end dropped
from two weeks to four days.
two questions guided the testing:
walkthrough sessions and feedback loops with sales users and approvers
question consolidation workshops with product owners
audits of component and template usage across products
rewriting system messages in plain, human language
measured in q4 2022 against the pre-redesign baseline.
a global crm is an ecosystem, not a screen.
