All projects
B2B Wealth Analytics Enterprise SaaS Information Architecture A/B Testing

The wealth-analytics dashboard, in one screen.

Led product design for the software financial advisors use to check client investments, work out tax, and plan where money should sit. Around 800 advisors used it every day. I cut the time to review one portfolio nearly in half.

Case Study Tata Consultancy Services B2B Wealth Analytics · Sep 2022 — Aug 2023
Role Senior Product Designer
Timeline Sep 2022 — Aug 2023
Platform Web SaaS · B2B Advisor Dashboards
Domain Wealth Mgmt · Portfolio Analytics
At a glance Full read · 16 min  |  This panel · 40 sec

The problem

Reviewing one client portfolio took 25 minutes against a 12-minute target. About 9 of those minutes went on jumping between four sections to make numbers agree that should have agreed already.

What I did

Put all four sections on one screen, had the software flag disagreements itself, and let advisors open any headline number down to the calculation behind it.

The result

−44%Review time +31%Dashboard use 9,167Hours saved / mo 8→3Eng questions

The Stakes

Advisors were taking 25 minutes to review a portfolio when the target was 12. With about 800 advisors and 50,000 portfolios a month, every extra minute cost the bank roughly ₹1.25 lakh in advisor time.

Outcomes

25 min → 14 min per review · ~9,167 advisor hours saved a month · ~₹1.37 Cr recovered

-44% portfolio review time, from 25 min to 14 min per portfolio
9,167 advisor hours per month reclaimed at scale across ~800 RMs
+31% core dashboard engagement post-redesign, measured via A/B testing
8 → 3 engineering clarification cycles per sprint after component patterns shipped
01
Overview

Advisors kept getting lost inside their own software

TCS built this platform for a large Indian banking group. It is the screen around 800 advisors sit in front of all day, checking investments for ordinary and wealthy clients. It had everything an advisor needed: returns, tax forecasts, plans for where money should sit, and risk checks. The numbers in it were correct. Getting to them was the problem.

I joined as Senior Product Designer to redesign the dashboards and the way advisors move through them. The brief on day one was a number. Advisors were averaging 25 minutes on a portfolio review against a 12-minute target. At that scale the slowness sits directly in what the bank pays for advice.

Client

Large Indian banking group (B2B SaaS)

Industry

Wealth Management · Financial Services

Advisor Base

~800 RMs across 3 regional clusters

Portfolios/Month

~50,000 reviewed

Modules

4 (Returns, Tax, Allocation, Risk)

SLA Target

12 min per portfolio review

02
Problem

Advisors spent a third of every review just finding things

The pattern showed up on the first day of interviews. Advisors were not spending 25 minutes analysing a portfolio. About 9 minutes of every review went on jumping between four separate sections of the software, trying to make numbers agree that should have agreed already. The analysis was quick. The software was making people check things by hand that it should have checked itself.

Problem statement Mid-tier wealth advisors (relationship managers, 2–8 yrs tenure) struggled to deliver portfolio reviews within SLA because returns analysis, tax forecasting, and asset allocation data were fragmented across 4 disconnected modules requiring manual cross-referencing, resulting in average review times of 25 minutes against a 12-minute SLA — a 108% overshoot documented across 18 workflow interviews and 24 timed sessions.

"I open returns, copy the number into a notepad, switch to tax, copy the next number, switch to allocation, copy the third. Then I open a fourth tab to do the math myself because the platform won't tell me if they reconcile."

— RM, 4 years tenure, workflow interview
03
Research

Four sources of evidence, one clear cause

I used four separate sources of evidence. None of them proved much alone, but together they were clear. The analysis side of the product was strong. Moving around it was costing the bank an entire target review time on every single portfolio.

01
Workflow Interviews (n=18)
I interviewed 18 advisors across three regions. Every one of them raised the same thing: time lost switching between sections. Five had built their own Excel templates to pull the numbers together before they even opened the platform. When people build a second tool to make your tool usable, that is the finding.
02
Timed Task Observation (n=24)
I recorded 24 full portfolio reviews with screen capture and a stopwatch. The average review took 24 minutes 47 seconds. Of that, 9 minutes 12 seconds went on switching sections, copying and pasting, or checking one number against another by hand. Over a third of the time was spent on the software rather than on the client.
03
Support Ticket Mining
Of 487 support tickets in a quarter, 312 of them (64%) were people asking where a piece of data lived or why two reports disagreed. Neither is a question about analysis. The support team had quietly become the thing that made the numbers line up, covering for a product that should have done it.

Checking the competition: the gap was closable

I went through three comparable tools used elsewhere in the industry. Published reviews put a typical portfolio review at 12 to 15 minutes. This platform was 10 to 13 minutes behind, and almost none of that gap was about missing features. It was about how the product was arranged. That gave me a realistic target and a number the business could hold me to.

04
Options Considered

Three ways to fix it. I ruled out two.

Option A — Rejected
Global search across all modules — "find anything fast"
A search bar that would index every report and figure across the four sections. It sounds powerful. I ruled it out because searching finds data but does not make two numbers agree, and agreeing was the actual bottleneck. Advisors were not hunting for things. They were cross-checking them. Search would have built the wrong product faster.
Option B — Rejected
Aggregated "summary tab" sitting on top of existing modules
The cheap fix: leave the four sections alone and add a summary screen that pulls the headline numbers together. Less work and less risk. I ruled it out because advisors told me plainly that they do not trust a summary they cannot open up. Without joining the data underneath, a summary is just a fifth disconnected screen.
Option C — Chosen
Unified portfolio canvas — one screen, cross-module data, drill-down hierarchy
Put returns, tax, allocation, and risk on one screen, link them to each other, flag it automatically when numbers disagree, and let an advisor open any headline figure down to the calculation behind it. This fixes the actual cause. Advisors stop switching, the software does the cross-checking, and opening something up replaces hopping about. It was the biggest build of the three and the only one that closed the gap.
05
Decision & Tradeoff

We designed for the slower majority and accepted a fight with the fastest advisors

The single screen was the right call and it had a real cost. The top 18% of advisors, who bring in 40% of revenue, had memorised the old layout and moved through it fast. The redesign broke habits they had built over years. For two weeks after launch they were the loudest voice in the building.

+ Gained
Review time fell from 25 to 14 minutes for the other 82% of advisors, who between them do 78% of all reviews. Use of the main dashboard features rose 31%. Tickets asking where data lived stopped almost entirely.
− Lost
The fastest 18%, who bring in 40% of revenue, lost their shortcuts for the first two weeks. We got 47 escalation emails. Three senior advisors asked us to put the old layout back.
Why accepted
The 25-minute problem was hurting the advisors doing most of the work, and it happened on every single review. The senior advisors' problem was learning a new layout once. Fixing the recurring cost was right, even though it made for an uncomfortable fortnight.

How we softened the change for senior advisors

I built a toggle back to the old layout and left it in place for 60 days, so senior advisors could switch back while they learned the new screen at their own pace. After 60 days, 71% had moved across on their own. We removed the old layout in the third release and heard nothing further. When you change how an expert works, give them a way across rather than a drop.

06
Design Decisions

Every decision came back to one number: minutes per review

One screen: the order that replaced four sections

Returns at the top, allocation in the middle, risk at the bottom.

This came from watching people work. Advisors read a portfolio from the top down. Returns tell them what happened, allocation tells them why, risk tells them what it means. The screen follows that order. Across 24 timed sessions, not one advisor read them in a different sequence.

The tax forecast sits on the same screen, not behind a tab.

Tax was the number people most often had to check against something else. It needed to sit next to returns rather than a click away. It is a panel on the same screen, closed by default for advisors who want less on screen and open for those who want it.

The disagreement flag: the quiet favourite

The feature people liked most after launch was a small warning that appears when a portfolio's stated allocation does not match what it actually holds, usually because a company action had not fully come through in the data. Before the redesign advisors caught these by hand maybe 60% of the time. With the flag they catch all of them, and nobody has to remember to look. A whole category of quiet mistakes disappeared.

Testing the new screen against the old

I ran the testing myself, comparing the new single screen against the old sections across the three regions. Use of the main features, opening up returns, rebalancing allocation, and modelling tax, rose 31% in the group on the new design. The summary screen some stakeholders had asked for was settled by the data: where we shipped it, people used it 84% less than the main screen.

Reusable parts, and what they did for the build

I built a set of reusable parts for this kind of work: portfolio cards, the pattern for opening a number up, the disagreement states, tax input fields. They live in a documented Figma library mapped to the code. The effect showed up in the build. Questions from engineers each sprint dropped from about 8 to 3, and later work on the risk section and client reports moved faster because nobody was reinventing the same pieces.

07
Outcome

25 min → 14 min, across 50,000 portfolios a month

What changed

Review time fell from 25 to 14 minutes, a drop of 44%. Measured the same way as before the redesign, across 30 reviews in the same three regions. The result held in all three, which suggests the gain came from the design rather than from one office.

Use of the main dashboard rose 31%, measured by testing across the three regions. The summary screen we tested alongside it saw 84% less use than the main screen, which ended a feature request before it reached everyone.

Questions from engineers dropped from 8 to 3 a sprint once the documented library shipped, and the share of work sent back at design review fell from about 22% to about 7%.

What it is worth at scale

Eleven minutes a review does not sound like much on its own. Across the whole user base it looks different:

(25 − 14) min × 50,000 portfolios a month = 550,000 minutes, or about 9,167 advisor hours a month

At roughly ₹1,500 an hour for advisor time, an industry figure published in 2023, that is about ₹1.37 Cr a month in advisor time recovered. The point of the number is not precision. It turns a design fix into something the executive team can defend against their own budget.

"The redesign didn't make the analytics better. It made the platform finally match the way advisors actually work. That's a far harder problem — and a far bigger payoff."

— Product Manager, internal launch retro
08
What's Next

Next: AI-drafted client notes for the slowest quarter

14 minutes is good, and the spread underneath it is interesting. The fastest quarter of advisors are at 10 to 11 minutes. The slowest quarter are at 16 to 18. Almost all of that gap is now the writing of the note the client actually reads. Experienced advisors write it quickly. Newer ones labour over it.

My theory: an AI-drafted first version of that note, which the advisor then edits and approves, would take 4 to 5 minutes off the slowest quarter and pull everyone closer to the target. The tracking is already there, so we know what advisors keep, change, and throw away from any draft. The risk is that AI-written text in regulated advice needs careful limits, something I had already worked through in detail at Goldman Sachs.

09
Constraints

A client, a live user base, and a fixed release schedule

TCS work runs on visible limits. The client sets the release schedule, the software is in use every day and cannot go down, and designs either land inside the sprint or they do not ship. The redesign had to hold up to the advisors using it, the bank's product team, and TCS engineering leadership at the same time.

I protected quality and scope. Time was what I let move.

Cutting scope was not an option, because half a unified screen is a worse layout than four honest sections. Cutting quality would have cost the bank's confidence in TCS. So the schedule moved. The screen shipped across two sprint cycles, with the old layout kept alongside it to absorb the disruption for senior advisors.

01
Finding Assumptions Inside Client Constraints
The client's data rules meant I could not recruit bank customers, so the research had to come from the advisor side. Eighteen interviews, 24 timed observation sessions, and a read of the support tickets gave me three independent sources without ever touching customer data. That was enough, because the delay was on the advisor side.
02
What Moved to V3
AI-drafted client notes, with limits for regulated advice. Nudges to rebalance based on the new disagreement flags. Reviewing several portfolios at once. All written up with tracking already in place, so the team taking over inherits a brief rather than a blank page.
03
How Documentation Compressed Sprint Cycles
The reusable parts did more than keep screens consistent. Writing down what each one does, how it behaves, and what happens in the awkward cases stopped engineers having to ask every sprint. That alone gave back roughly half a day per engineer per sprint.

"In a B2B SaaS engagement with a live user base, the constraint isn't creative freedom. It's the discipline to ship the change advisors will actually adopt without breaking the workflow that pays the bills."

10
Reflection

What I took from this project

01
Architecture Is The Feature
The strongest analytics product I have worked on did not lose people on the maths. It lost them on getting around. When someone lives in your software for eight hours a day, how it is arranged is the product. Features come after that.
02
Optimise For The Volume, Bridge The Power Users
The loudest users are rarely the biggest group. The right call was to design for the advisors doing most of the volume, and to give the fastest ones a way to cross over at their own pace. Within 60 days, 71% had crossed on their own.
03
Documentation Is A Force Multiplier
The reusable parts lasted longest. The dashboards have been changed several times since I left. The written record of what each component does still shapes how the team builds.

Conclusion

When software is arranged the way people actually work, the saving is not a few percent. Here it was 44%, on every advisor, every portfolio, every month.

The unified portfolio canvas, reusable component patterns, and A/B-validated decision framework remain in production use across the bank's wealth analytics platform.

Glad we could cross paths.
Out of anywhere you could be, you're here.

Next project

Maxtra Technologies
All projects