Skip to content

B2B Fintech · Payments

Invisible Revenue

66% of UK revenue flowed through 300+ external sellers the company couldn't see. I designed the 0-to-1 platform that made that pipeline visible, and redefined the data model three teams depended on before a single screen existed.

300+
Field agents served
10s
Deal capture time
17
Iterations to earned simplicity
Role
Senior Product Designer
Focus
0-to-1 platform
Duration
2025–2026

Owned the problem definition and end-to-end design of a 0-to-1 sales platform: lead capture, deal management, lifecycle governance. Operated as a peer to Product, Engineering and Sales leadership, including decisions that reshaped the backend data model.

Sales platform overview

The business problem: an invisible revenue pipeline

External sellers generated 66% of UK revenue, but the company had zero visibility into their activity until a merchant was already registered. 15-day activation had dropped from 62% to 48% in four months; 50% of sellers churned within their first three months. A previous CRM had failed because it demanded too much data upfront. The brief said 'build lead capture'. I treated the brief as a hypothesis: the real problem sat a level deeper, because nobody had agreed what a lead actually was.

The lead and deal taxonomy that restructured the backend
A lead is a contact identity. A deal is a sales attempt. That clarification restructured the backend before any interface work began.

Capture in 10 seconds, comply over time

300+ field agents needed to log opportunities on-site; compliance needed comprehensive data. Those goals look incompatible until you separate creation from enrichment, so that's what I designed: create a deal with just a name in 10 seconds, enrich with contact details, offers and compliance data later. Deduplication only triggers once a unique identifier is added. Across 17 iterations I stripped the flow to what agents needed in the moment, while the system absorbed complexity (multiple offers per deal, stage transitions, conflict resolution) without exposing it upfront.

Progressive disclosure capture flow
Radical simplicity at the point of capture, depth on demand.

Key decisions

01

Cut AI override functionality from V1

Scope included automated capture edge cases. I challenged the actual frequency and risk before anyone built it: weeks of engineering work came out of V1, the pilot launched faster, and no core value was lost. Knowing what not to build was as consequential as anything I designed.

02

Accepted incomplete records to win capture

The trade-off was explicit and accepted by the business: temporarily incomplete records in exchange for opportunities actually being captured. The previous CRM died on this exact hill by demanding everything upfront. Completion improved because the initial friction was gone.

03

Treated the brief as a hypothesis

The brief said 'build lead capture'. The real problem sat a level deeper: nobody had agreed what a lead actually was. I kept asking until the answer changed, and the previous CRM's failure made the cost of skipping that question concrete.

04

Redefined lead vs deal before any screens

A lead is a contact identity. A deal is a sales attempt. That clarification restructured the backend data model three teams depended on before any interface work began.

05

Separated capture from enrichment

Field agents and compliance had incompatible needs until creation and enrichment became different moments: create a deal with just a name in 10 seconds, enrich with contacts, offers and compliance data later. Deduplication only triggers once a unique identifier exists.

06

Spent 17 iterations earning simplicity

Each pass stripped the capture flow to what agents needed in the moment while the system absorbed the complexity: multiple offers per deal, stage transitions, conflict resolution. Depth on demand, never upfront.

Outcome

Beyond the platform: a shared lead/deal taxonomy now encoded in the backend, a capture-then-enrich pattern the team can extend to future field tooling, and a precedent that design interrogates the data model rather than decorating it.

300+

Field agents served

10s

Deal capture time

17

Iterations to earned simplicity