Moneytor · Debt Recovery · 2021—2022
Hired to make it look good.
I rebuilt the system.
A solo designer, a broken fintech product, and twelve months of embedding myself into every role that would ever use it — before touching a design tool.
Role
Solo Designer → de-facto PM
Timeline
12+ months
Team
1 designer (me)
Scope
Product Design, UX Research
Background
Hired for a visual refresh. Ended up rewriting how the product ran.
Moneytor is a debt-recovery platform for an Indian fintech. Six roles use it daily — telecallers, field agents, telecaller managers, agency partners, portfolio analysts, compliance heads. Every role complained about a different surface; the symptoms pointed to one cause: nobody had ever designed the product from the operator's seat.
I joined as the only designer. The brief was cosmetic — fix the type, fix the dashboard, polish the spacing. Twelve months later I had embedded with every role, reframed the brief, and shipped ten modules end-to-end as solo designer and de-facto PM.
Impact
What twelve months of embedded research turned into.
Problem
Six roles. None of them could finish their job inside the product.
Telecaller
Couldn't see what to call next without the manager's morning hand-off.
Field Agent
Got assigned cases out-of-band; couldn't log a visit or settlement on the move.
Telecaller Manager
Allocated cases manually each shift. No way to simulate a rule before activating it.
Agency Partner
No live view of their team's performance. Numbers arrived after the fact.
Portfolio Analyst
Pulled data into a spreadsheet to find risk pockets. Insights arrived after the month closed.
Compliance Head
No reliable audit trail. Sensitive actions left no defensible record.
The product wasn't ugly. It was disconnected from the people using it. Every role had a workaround. The visual refresh I was hired for would have painted the symptoms.
Reframe
The brief was cosmetic. The product was broken at the workflow layer.
What I was hired to do
Refresh the visual layer.
Modernise the type, fix the dashboard, polish the spacing. A short cosmetic engagement.
What I had to do instead
Rebuild ten modules. Solo.
Embed with every role. Reframe the brief. Spec, design, QA, and ship the modules nobody had owned end-to-end.
The Work
Ten modules. One mental model. Built in the order the user runs their day.
01 · Dashboard
Four dashboards, not one — because no two roles open the day with the same question.
Manager opens to recovery health and the alerts that need a decision today. Agent opens to their queue. Analyst opens to portfolio risk. Role decides which view loads first.
Agency Performance for the partner channel. Portfolio Health for the analyst. Personal Growth Tracker for the operator. Same data, three lenses.
Plus a builder so a manager could compose a fifth dashboard without filing a ticket. Three steps, drag-to-arrange, shareable.
02 · Cases
Status, financials, contact, time pressure — surfaced at the row, not buried in a tab.
Borrower, loan ID, financials, contact channel, last-touched — all visible at the row. Filters by DPD, status, callbacks. Bulk actions selected without leaving the table.
Add. Upload. Export. The three actions a manager runs every day, designed once, consistent everywhere.
03 · Case Detail
An AI strategy panel that reads the case before the agent makes the call.
Recommended strategy with reasoning the agent can defend on the call: hard-pull, soft-pull, settlement, callback. Below it: every call, SMS, email, WhatsApp message — one chronological timeline, one source of truth.
The Quick Action Command Center is the shortcut layer for actions agents repeat all day — call, SMS, email, log PTP, schedule visit, callback, post note. Two keystrokes each.
04 · Allocations
Allocation rules. Self-serve. Engineering finally got to leave the room.
Four-step wizard — eligibility, allocation logic, simulation, activation. Operators design the rule, simulate impact on the live backlog, then publish. No more engineering tickets for routine rule changes.
05 · Offers
Settlement structuring became a workflow. Not a back-channel approval.
Discount slabs, auto-approve thresholds, multi-level approval hierarchy. Every settlement is auditable. Every override has a reason attached before it leaves the screen.
06 · People
Every person, every team, every reporting line — visible in one mental model.
Workload, performance, quality scores, payout eligibility per agent. Six tabs, one page. Reassign or suspend without leaving the profile.
Teams hold the recovery target and the backlog. The org chart makes reporting lines and capacity legible at a glance.
07 · Access Control
The trust model of an entire organisation, encoded in a visual matrix.
Per-role permissions across every module. View / Edit / No Access. Global data safeguards as defaults — not options. The footer keeps the receipt: who changed what, when.
08 · Capacity & Incentives
Operational guardrails on top. Motivation on the bottom.
Telecaller max load, agent radius, holding limits — capacity rules that prevent burnout and bad allocation. Incentive schemes — algorithmic, payout-tracked, audit-ready.
09 · Import Data
Self-serve data ingestion. The engineering bottleneck — eliminated.
Excel/CSV upload, field mapping with preview, sanity-check rows, then commit. Operators ingest fresh data without an engineering ticket.
10 · Audit & Compliance
Every action, every export, every permission edit — logged, searchable, downloadable.
Real-time activity stream. Security flags, bulk export tracking, role permission changes, settlement approvals. One downloadable compliance pack the head can hand to an auditor.
Even the boring page closes with a compliance tag — security as a primitive, not a checkbox.
Design Decisions
Seven decisions. What each one reveals about how I think.
Decision
What it reveals
Embedded across six roles before opening a design tool.
Research isn't a phase — it's the foundation. Without it, the work is rooted in assumption.
Built role-aware dashboards instead of one master view.
User context determines hierarchy. A manager and an agent don't need the same first viewport.
Made the AI panel a recommendation, not a decision.
AI augments judgment, doesn't replace accountability. The agent still owns the call.
Designed allocations as a self-serve wizard with simulation before activation.
Trust comes from transparency. Operators trust rules they can see, dry-run, then ship.
Encoded global data safeguards as defaults, not toggles.
Compliance is the floor everyone stands on. It can't be a checkbox someone can untick.
Treated the audit log as a product surface, not an admin afterthought.
If it's not auditable from day one, it isn't defensible at the day-one-thousand audit.
Wrote the spec, the flows, the QA criteria — and ran the standups.
When you understand the user better than anyone, you become the PM whether you signed up for it or not.
Context
What the rebuild changed for the people who run it daily.
Sales
Demos became something the team was proud to run.
Before the redesign, sales demos required engineering to pre-configure the environment. After launch, the team could demo self-sufficiently — and prospects could see themselves using it.
Support
Recurring complaint patterns dropped.
Allocation confusion, navigation loss, settlement approval gaps — the three patterns I'd identified during research all had direct design solutions shipped. Tickets on those patterns reduced post-launch.
Training
Onboarding shortened.
The product had previously required extensive daily training. The AI strategy panel, role-aware views, and guided wizards made the product learnable without a manual.
Scope boundary
What I designed and shipped at Moneytor in 2021–2022 was the production product, working with the in-house engineering team. The live demo linked above is a frontend prototype I rebuilt in 2026 to make the original design accessible end-to-end — navigation, screens, and interaction patterns are intact; backend and data layer are not connected.
Closing
Six roles, ten modules, twelve months — solo. The product wasn't broken because of design. It was broken because nobody understood the user well enough to make the decisions. So I did.
Try the prototype rebuild — moneytor-seven.vercel.appDeep Dive
Four chapters on the moves that shaped the rebuild.
Chapter 1 · The AI strategy panel
The agent's reasoning, made visible.
Senior agents had a feel for which case needed which tactic — soft pull, hard pull, settlement, callback. Junior ones didn't. The AI strategy panel encodes that heuristic as a recommendation with reasoning the junior agent can defend on the call. It never decides. It makes the decision defensible.
Chapter 2 · Quick Action Command Center
Designed for the agent who runs the floor.
The original product made an agent click through nested menus to record a callback or log a PTP. The command center is the shortcut layer for everything an agent does often — two keystrokes, then back to the call.
Chapter 3 · Self-serve everything
Eliminating the engineering bottleneck wasn't a feature. It was the principle.
Every workflow that used to require an engineering ticket got a wizard. Data import. Allocation rules. Settlement structuring. Custom dashboards. Each wizard followed the same skeleton — choose, configure, simulate, publish. Operators learned the pattern once and applied it to every new module.
Chapter 4 · Three lessons that travel
What I'd hand to the next solo designer walking into a broken platform.
01
Embed in the work before designing for it
Time on the floor before opening a design tool feels slow at the time. It saves redesign loops later. Embedded research isn't a luxury — it's the only way solo design at platform scale becomes possible.
02
Compliance is a primitive, not a layer
Audit log, RBAC matrix, global safeguards — these had to be in v1, not v3. Retrofitting compliance into shipped modules is twice the cost. Build the floor before the rooms.
03
When you become the PM, accept it
A solo designer who understands the user better than anyone in the room is the PM. Pretending otherwise wastes everyone's time. Write the spec, run the standup, own the criteria — that's the role the work needs.