01 / 20
Speaker notes
Desktop presentationMade for a laptop or larger screen. Here, swipe or use the arrows to move through the slides.
Case study · TD Bank · Payments

Money that quietly manages itself.

A Personal Finance Manager for TD's mobile app: budgeting, automated savings, and transaction clarity. Built by a small team, validated through multiple rounds of user research.

Lead Design Prototypes Research
Banking Native mobile Money movement Working prototypes, embedded ahead
What this could be worth · modeled, not booked · the math and every source are on slide 19 and the last page
$3.4–4.0Bmodeled value: the product as assessed by Visa Consulting & Analytics for TD, plus the patent this research produced
$2.5Brevenue lift for the PFM MVP, Visa Consulting & Analytics assessment for TD, 2022
$835M–1.5Bthree-year value of the patent this research produced, modeled for TD's IP committee, Nov 2025
$621Mof the $2.5B from savings automation alone, the round-up engine. Largest single line
42–150×value against the $10 to 20M build. Six patents filed; bank GUI verdicts have passed $100M
RoundUp +

What increment do you want to round up?

Dinner $28.90 rounds to $30.00 +$1.10 saved
$0saved last month, from change alone
$0a year, without deciding

Jay Frank · Principal Product Designer / Strategist

9:41
TD Bank
DepositZelle®Pay a BillTransfer
Financial Management
ACCOUNT SNAPSHOT
$0
$5,969 over the past month
Checking…1234$20,000.00
Savings…4534$20,000.00
Credit Card…8987- $3,795.50
Credit Cards…3732- $3,795.50
Linked through open banking
BABank of America…7741$4,120.88
CHChase Sapphire…2209- $1,284.30
Add Account +
Spending & TransactionsView
1 month6 month1 yearAll time
May 2023Total Spending$0All Accounts
Top 4 CategoriesView
Mortgage & Rent$1,990.00 | 35%
Groceries$600.00 | 20%
Utilities$200.00 | 15%
Electronics$150.00 | 10%
View more categories
My BudgetsView
You have
$3,409
Remaining across all categories
Shopping$345 / $500
Restaurants$345 / $500
Education$345 / $500
Electronics$345 / $500
My Savings GoalsAdd Goal ⊕
New Car
Saved towards goals$5,347 of $12,000
Spending AnalysisView
3 month6 month1 year
Up 7% from last month
8007006005004003002001000
AprMayJunJulAugSep
Things to know
798↑ 12 points
Your credit score went up 12 points on 1/14/23.Here’s why
×
Quick Links
RoundUp +

A new way to round up. Control the amounts you save with spare change on purchases

Roundup + is active

What increment do you want to round up?

$1$5$10$20
Linked toSavings…4534
You would have saved last month$50.00
Over the past year$250.00
Upcoming Activity

See what is coming before it happens.

Next 2 weeks

RentSep 1
$1,450.00
HydroSep 3
$86.40
PhoneSep 5
$72.00
Car paymentSep 8 · tight week
$310.00
Projected balance after$1,236.10
Accounts Pay & Transfer Financial
Management Menu2
How I am going to spend this

Fifteen minutes. The first eight are the problem.

There is a line people put in Einstein’s mouth: given an hour to solve a problem, spend fifty-five minutes on the problem and five on the solution. I cannot source it to him. I went looking and it is not there. I use the ratio anyway, and honestly the fact that I checked is the more useful half of the story.

The case study15 minutes
8 min · understanding the problem
7 min · the product and the evidence
The first eight

Who the customer was, what the bank believed, and what the research actually found. The person every budgeting tool on the market broke on. The two decisions where we tested a real alternative.

The last seven

The product itself, every decision inside one control, the measured results, the honest ending: it was defunded before launch, and how the work reads from the seat you are hiring for. The final four are an appendix, and I will tell you when it starts.

A problem well stated is a problem half solved.Charles Kettering, head of research at General Motors. That one is attributable.
Context

A bank betting on financial wellness

TD's Payments organization set out to build personal financial management into the bank itself: one place where budgeting, savings, and spending clarity live next to the money instead of in a third-party app. Our team owned the experience from discovery through validated design. Through open banking, the picture is not limited to TD: customers connect accounts held anywhere, so the snapshot, budgets, and spending insights cover all of their money, not just the part that sits with us.

9:41
Spend Insights
DepositZelle®Pay a BillTransfer
MAY 2023
TOTAL SPENDING
$0.00
ALL ACCOUNTS
This Month6 monthsThis Year
BankingFilter by Account
Credit AccountsFilter by Account
Spending GraphsView
0501005001K3K
Amount
Total of all Categories$3,200.00
80% of May income

Spend insights · all accounts, TD and connected

3product pillars: snapshot, budgeting, savings Click for the pillars
The three pillars
  • Central Snapshot · "show me everything in one place." Every account, TD and connected, in one view
  • Budgeting Tools · "help me avoid overdrafts." A starter budget built from real spending, editable from there
  • Savings · "save without thinking about it." Round-ups and rules that move the money automatically
15competitor apps audited in the discovery teardown Click to see them
Discovery teardown
  • Mint
  • YNAB
  • Monarch Money
  • Copilot
  • Rocket Money
  • PocketGuard
  • Simplifi
  • Goodbudget
  • EveryDollar
  • Acorns
  • Chime
  • Empower
  • Albert
  • Cleo
  • Qapital
…
9MVP milestones sequenced with engineering Click to see them
MVP sequence
  • Account aggregation
  • Central snapshot
  • Transaction categorization
  • Category correction
  • Starter budget
  • Budget management
  • Spend insights
  • Savings goals
  • Round-up automation
2+rounds of moderated user interviews, plus surveys
Timeline

About 14 months, discovery through validated design and patent filing.

Team

Two product designers, an HCD research lead with two researchers, product, engineering, content strategy.

Platform

Inside TD's existing mobile app, iOS and Android, on the bank's design system.

Constraints

Research was a shared resource. The design system had gaps. Every number on screen went through review. Then the program was defunded.

My role

One team, clear lanes

The HCD team I led co-created the product end to end with our business partners and other lines of business: shared flows, daily critique, one system with all of our fingerprints on every screen. Alongside that shared design work, four lanes sat with me.

DiscoverDefineDesignPrototypeTestrepeat
  • Product design, co-created with my HCD team full spectrum
  • Prototyping: every tested build came through my file all rounds
  • Research leadership, with our HCD research lead discovery & testing
  • Workshop facilitation across the feature pillars define phase
The core team
Product DesignersShared flows and daily critique, one system across every screen, through to delivery
HCD Research LeadMethod design, dedicated researchers, synthesis
Lead EngineerIn the workshops from day one, MVP feasibility
Content StrategistVoice, tone, and the plain-language transaction work
Chief Product Owner & POsPriorities, scope, and the funding narrative
Business AnalystRequirements and the competitive landscape
Research

We built the research muscle before the product

Research at a bank is a shared resource, so our HCD research lead and I made the case together and got dedicated researchers. We also did not start from zero: Payments already held prior study reports and running customer research diaries. We mined those first, so our own discovery began where the bank's existing knowledge ran out.

DiscoverDefineDesignPrototypeTestrepeat
  • Prior Payments studies & research diariesmined first, set our opening hypotheses
    before discovery
  • Moderated depth interviewswhy budgets die at setup, and what "am I okay" really means
    discovery
  • Surveysscale and rank what the interviews surfaced
    discovery
  • Competitor teardown, 15 appssolved ground versus open ground
    discovery
  • Concept evaluation rounds, moderated with our HCD research leada stack of hypotheses per round, one prototype
    testing
  • Workshops with product, engineering and contentevidence into a committed MVP list
    define
Round up all purchases and put into savings or investments
Pull account information from non TD accounts and Credit Cards
Budget recommendations that can tailor to goals
Spending patterns per account or overall, can personalize
Customize savings goals: Family, Rainy Day, Kids, School
Potential Risk / OverDraft
Show transactions coming within 2 weeks
Overall health score
Check debt to income ratio
Offer debt consolidation
401k contributions, education and tips
Alert when unusual spending is done
Future Balance monitoring: predict checking or savings balance
Savings calculator: how savings are affected over time
Add others to work with the same budgets, spouses etc.
Show ranking of accounts that yield the most for you
Set amounts to be dispersed to specific accounts
Tailored education material
9:41
Recommend Budget
DepositZelle®Pay a Bill
Remaining After Budget$1,467.00
Recommended Budget
We recommend this budget based off of your transaction data for the last 6 months.
Projected Income$6,200
Avg Spent / Recommended Limit
Housing$2400 / $2,400
Food$300 / $300
Lifestyle$400 / $600
Transportation$150 / $300
Savings and Goals$0 / $833
Miscellaneous$300 / $300
Roundup +
We’ll round up any spare change from your deposit and credit accounts and place it in the saving account of your choosing.
What you could have saved with RoundUp +
If you set this to $5 last month it would have looked like this…
Last month$50.00
Last month
w/ round up$250.00
RoundUp +
What increment do you want to round up?
$1$2$5$10$20
Select the saving account you wish to link.
AccountSaving 6546
Enroll
‹July 2023›
Monthly Budget Insights
This month you went over budget by $300. We have put together a recommendation that will help you adjust your budget based off of your last 3 months history.
Budget Breakdown
A Summary of your budget progress so far during this period
Income$6,000
!
$0$300 over
Spent / Limit
Over Limits
!Income$6,300 / $6,000
!Housing over $300$2,700 / $2,400
Under Limits
Lifestyle$500 / $800
Perfect
Food$400 / $400
Transportation$300 / $300
Savings goals$850 / $850
Miscellaneous$400 / $400
Updated Budget Recommendation
Savings
Vacation$0 Spent / $200 Limit
Emergency Fund$0 Spent / $500 Limit
Coming out · next 2 weeks
N
NetflixAug 29 · subscription
$16.99
Amazon PrimeSep 2 · subscription
$14.99
GoodLife FitnessSep 3 · membership
$34.99
Uber OneSep 5 · subscription
$9.99
Venmo · rent splitSep 8 · scheduled
$310.00
5 charges scheduled$386.96

Every idea stayed on the board · Tested: recommended budget, built from 6 months of transactions

The evidence · outside the building

The strain was never just ours.
The public numbers agree.

  • 89% of Canadian students report financial stress in the past three months, and 23% say they lack basic budgeting skills TD survey, 2026
  • Only 37% of Gen Z Canadians feel in control of their money; 64% feel stress multiple times a week TD survey, 2025
  • 36% of Americans are satisfied with their financial health, 68% feel pressure to appear more financially successful than they are, and 72% of dual-income couples report regular money stress TD Bank US, Fiscal Fit & Love & Money
  • More than a third of Americans could not cover an unexpected bill from savings TD Financial Preparedness Report, 2025
  • One survey run identically in both countries: 16% of Americans and 12% of Canadians feel financially fulfilled, and 32% and 41% are in financial stress Gallup with Edward Jones, 2026

Publicly published TD research on both sides of the border, stories.td.com and td.mediaroom.com, plus one cross-border study, Gallup with Edward Jones, Money and Meaning, 2026. Our own discovery, two years earlier, found the same shape.

Three lenses, held from here to the business case
01 · Desirability

Do people want it? Proven in discovery, then re-proven in every testing round.

02 · Feasibility

Can we build it? The lead engineer sat in every scoping session.

03 · Viability

Should the bank fund it? The activation metric and the business case.

How people actually feel about their money CANADA Students feeling money stress Gen Z stressed weekly or more Gen Z in control of their money Say they lack budgeting skills UNITED STATES Couples reporting money stress Pressured to look successful Satisfied with financial health Cannot cover a surprise bill 89% 64% 37% 23% 72% 68% 36% 35% ONE SURVEY, BOTH COUNTRIES TD publishes different studies in each market, so the groups above are not direct comparisons. These pairs are: one study, run identically in Canada and the US. Financially fulfilled · Canada Financially fulfilled · US In financial stress · Canada In financial stress · US 12% 16% 41% 32%
The problem
DesirabilityFeasibilityViability

People were not failing at money. Their tools were.

1

"Help me avoid overdrafts"

Overdrafts were the fear, and budgeting was the only tool on offer. Budgets died at setup: the effort wall killed them before any value showed up, so the anxiety stayed.

Became the Budgeting Tools pillar
2

"Save automatically, without thinking about it"

People wanted effortless saving, but not surrendered control. Automation alone was not the answer. RoundUp+ was: every purchase saves, at an increment they set.

Became the Savings pillar, as RoundUp+
3

"Show me everything in one place, in real time"

Money lived across too many accounts and apps. No single surface answered "am I okay?"

Became the Central Snapshot pillar
The three quotes above came out of our own interview rounds, nearly word for word. They became the product's job description.Discovery interviews and surveys
How we agreed to measure it

Week-one activation as the leading indicator, with budget setup completion, savings opt-in rate, and return visits per month underneath it. The business case behind them: deposit growth and retention. Every number later in this deck maps back to one of these.

Primary persona

The customer every fixed budget breaks on

Dana R.

29 · shift lead at a regional grocer, plus weekend delivery work · TD chequing, TD credit card, one outside savings account

Income
$3,100–$4,400 a month
Pay cycle
Biweekly, plus irregular gig
Overdrafts
About twice a quarter
Saves
When the month is kind
Mode
Set-and-forget, until it is wrong

"Every app wants me to say what I spend in a month. I don't know what I make in a month."

Click a card for what sits behind it

What she wants

  • To know she can cover rent before payday, without doing the math
  • To save something without deciding to save something
  • One place that includes the account she keeps elsewhere

Heard in the interviews

  • "I check my balance before I get out of bed. That is my budget."
  • "I move $20 over when it feels safe. Some months it never feels safe."
  • "I do not want another app. My money already lives somewhere."

Where tools fail her

  • Fixed monthly budgets assume a fixed monthly paycheck
  • Setup asks for numbers she cannot answer yet
  • Overspend alerts arrive after the money is gone

The numbers she cannot answer

  • "Your monthly income?" It is $3,100 one month, $4,400 the next
  • "Your monthly grocery cap?" Depends if it is a two or three paycheck month
  • "What percent will you save?" A percent of a number she does not have
Setup dies on screen one. The app is gone by week two.

What we built for her

  • A starter budget drafted from her own history, not from her guesses
  • RoundUp+ at an increment she sets, so saving survives a thin week
  • A snapshot across every account, TD and connected

Why it holds for her

  • The starter budget reads six months of her real transactions; she edits it instead of inventing it
  • Round-up increments from $1 to $20, so a thin week still saves
  • Her outside account joins through open banking, so "am I okay?" is one screen

A composite built from the interview and survey base, not a single participant. Gig and shift workers were the group our concepts had to earn.

Define · the workshops

Every feature had to survive the board

The research and the persona came into the room with us. Same gauntlet for every candidate: open questions, ideas, concerns, competitor evidence, then MVP questions and a committed feature list. If it did not work for someone paid like Dana, it did not survive.

Click any column to see where it landed in the product

Open questions4
Do users know what is available to them, and where to go?
What entry points would they expect?
How much data do they want on the first screen?
Who owns the money-movement rails?
Where it landed
  • Quick links on home, so the tools are reachable without hunting
  • Snapshot as the default landing surface, not a sub-menu
  • Entry points named in plain language, not product names
4/5found spending analysis unaided after the nav change
Ideas5
Round up all purchases and put into savings
Pull account information from non TD accounts
Budget recommendations that tailor to goals
Spending patterns per account or overall
Customize savings goals: Family, Rainy Day, Kids
Where it landed
  • RoundUp+ shipped as a core savings feature
  • External accounts became MVP 1 of the snapshot
  • Named goals: Vacation, Rainy Day, Kids
4round-up increments tested before we settled on user-set
Concerns4
Aggregation depends on open banking rails
Recommendations need 3 months of history first
Too much data on the first screen kills the glance
Automation without control tested badly
Where it landed
  • Users set their own increment, rather than a fixed one
  • Pause and edit sit on the same screen as the switch
  • Recommendations wait for enough history before firing
0participants chose full automation without a control
Competitor evidence4
Round-ups exist, but at a fixed increment
Aggregation is common; budgeting on top of it is not
Strong budgets, but heavy manual setup
Nobody handles income that changes week to week
Where it landed
  • Custom increments became the differentiator, then the patent
  • Budgeting layered on aggregation, which competitors do not do
  • Variable income handled as a first-class case
15competitor products torn down; none offered a user-set increment
MVP questions4
Feasible in release one, per the lead engineer?
Does it serve one of the three jobs?
Can it ship without new rails?
What breaks if we cut it?
Where it landed
  • Feasibility agreed in the room, not after
The 9, in build order
  1. Account aggregation
  2. Transaction categorization
  3. Category correction
  4. Central snapshot
  5. Starter budget
  6. Budget management
  7. Spend insights
  8. Savings goals
  9. Round-up automation
Committed5
Central financial snapshot
One-tap starter budget
Spend insights
Round-up +
Savings goals
Where it landed
  • Every committed item is a screen in the tested prototype
  • One-tap starter budget came straight out of setup drop-off
  • Nothing shipped into design that skipped this board
7 min → 2 minbudget setup time, before and after the one-tap starter

The gauntlet, left to right. Click a column to focus it. Only what reached Committed went into design.

The road not taken
DesirabilityFeasibilityViability

Two decisions, and what settled them.

1

How much does a round-up round to?

The alternative on the table: the industry default, round every purchase to the nearest dollar. Simple to build, simple to explain, and already proven elsewhere in the market.

What I proposed instead: the customer sets the increment, anywhere from $1 to $20, with a custom value underneath that.

Bank-chosen increments were refused in every testing round. Control is what turned the feature on.
2

How does someone enroll?

The alternative on the table: the spec as written, a single enrollment form of fifty to sixty questions. Cheap, one screen, uses components that already existed.

What I proposed instead: ten conversational steps, one decision each, encouragement between them. New components, more build cost.

Twenty of twenty participants preferred the conversational flow in a moderated preference session. Not a production A/B, and I say so.
What we judged them against

Agreed before either test ran: opt-in rate under a control condition, completion of setup, and build cost against enrollment return. Naming the criteria first is what stops a preference test from becoming an opinion contest.

Who held the other side

One product partner on the enrollment flow, in those words: “nobody is ever going to use that.” I brought drop-off data and competitor patterns. When the argument stalled, I stopped arguing and called the test.

What I refused to trade

The increment picker. It cost new components and a longer build. Every round said the same thing, so I held it and framed the cost as investment against enrollment, which is the language that funded it.

Customization was my instinct. I never asked the team to take it on faith. We put it in front of users every round, and the evidence is what carried it, not my argument.The rule I work by when the idea is mine
Define

Three pillars, each mapped to a job, each sequenced to ship

How the scope was set · three lenses applied to every candidate below
01 · Desirability

Every candidate had to answer one of the three jobs from our interviews. Anything that did not was cut.

02 · Feasibility

The lead engineer sat in every scoping session, so build cost was named in the room, not after commitment.

03 · Viability

Nine MVP milestones sequenced against the Payments roadmap, with several patents filed from the work.

The job it answers
MVP 1
MVP 2
MVP 3
Central Snapshot
"Show me everything in one place"
AggregationInternal + external accounts, visualized
Custom snapshotsUser-defined views
Upcoming activityBills and events ahead
Budgeting Tools
"Help me avoid overdrafts"
Spend insightsForecasts and trends, so the data exists to build on
Creation & monitoringOne-tap starter budget, built from that history
CustomizationCategories, periods, alerts
Savings
"Save without thinking about it"
CustomizationUser-set round-up amounts
AutomationsRules and recurring transfers
OptimizationGoal pacing, aggressiveness

The five we committed to first: MVP 1 across all three pillars, plus the one-tap starter budget and savings goals pulled forward from MVP 2 because testing proved they mattered.

Central SnapshotView flow
New Car24 days left Saved towards goals$5,347 of $12,000 Camera Lens$0 of $1,000 Add another accountopen banking
My Savings Goals$5,347
Budgeting ToolsView flow
Housing · $2,400
Food · $600
Housing over $300
Monthly earnings $6,000 · remaining $1,467.00
SavingsView flow
Vacation$4,000 of $10,000 · 40% · Dec 2024
New Car$5,347 of $12,000 · 45% · 24 days left
$9,347saved across goals
RoundUp+ on · $77 in change this month · Vacation pacing 1 month early

Every screen in these flows is rebuilt from the concept evaluation file. Click a pillar to slide the real flow in.

The design · budgeting & savings
DesirabilityFeasibilityViability

Two features, designed as one loop

  • Smart categories pre-build the budget from real transaction history value first
  • Full review, edit, and custom categories on top fine-tuners
  • Threshold alerts shaped by customer feedback effortless tracking
  • Saved versus planned, month over month informed adjusting
  • Goal dates tuned by savings aggressiveness user in control
Patents filed from this work TD design system + new patterns
Manage Your Budget overview, expenses against income with category cards
Savings and Goals expanded inside the budget, Vacation goal with monthly contribution
Goal editing sheet opened, goal total and monthly contribution fields
Vacation goal changes, current versus new end date after lowering the contribution
Editing the goal total, recalculation preview in red
Editing the monthly contribution, save action
Saving the updated goal total
Updated goal saved, 8 percent completed by July 2024
Budget after the change, Savings and Goals recalculated to 60 percent used
Accessibility

Nothing carried by color alone: every over-limit state pairs the red with a label and a value. Type held to the bank's minimum sizes, targets at 44px, and the whole loop tested against WCAG 2.1 AA contrast before it went into a round.

Edge cases

What happens when income is irregular, a goal is missed, a connected account drops, a round-up would overdraw the chequing account. Every one of those got a screen, because the participants who broke the flow found them first.

Patterns

TD's design system where it served us, new patterns where it had gaps: the budget-to-savings recalculation, the increment picker, and the goal pace control were built and contributed back.

Interaction depth · one control, three levels
DesirabilityFeasibilityViability

The control on slide one, all the way down.

You already touched this one. It is the real component, not a picture of it. Three decisions sit inside a control most people would have shipped as a toggle.

RoundUp +

What increment do you want to round up?

Dinner $28.90 rounds to $30.00 +$1.10 saved
$0saved last month, from change alone
$0a year, without deciding

Live · the real control, same as the cover · try an increment or a custom amount

The interface never offers a move the system cannot honor. That rule is why the overdraw case has a screen instead of an error.The constraint we designed against, in a domain where the money is real
1

The choice: four chips, not a slider

A slider at money granularity is a precision problem solved with a thumb on a moving train. Four discrete increments answer almost everyone in one tap.

$1 · $5 · $10 · $20, and $5 as the default
2

The escape hatch stays hidden

Custom is the fifth option, and the stepper does not exist until it is chosen. The few who want $14 get $14; the rest never read past it.

Progressive disclosure, sized to the actual distribution
3

You see the consequence before you commit

The worked example and the yearly projection recalculate the instant the increment changes. Nobody commits to an automated money rule on faith; the outcome stays on screen while the decision is still reversible.

Dinner, $28.90, rounds to $30.00, +$1.10 saved
The design · snapshot & insights
DesirabilityFeasibilityViability

The first glance answers "am I okay?"
Every tap after answers "why?"

One snapshot aggregates every account. Spend insights layer category breakdowns, month and year comparisons, and custom ranges beneath it. Same screens, two depths: glanceable defaults for set-and-forget users, drill-downs for fine-tuners.

9:41
Manage Your BudgetExpense Overview
DepositZelle®Pay a Bill

Housing

Rent & Mortgage$0 Spent / $2,050 Limit

Food

Groceries$200 Spent / $400 Limit
Restaurants$200 Spent / $400 Limit

Transportation

Car Payment$0 Spent / $300 Limit
Gas$0 Spent / $100 Limit

SavingsAdd Item

Emergency Fund$0 Spent / $500 Limit
9:41
Manage Your BudgetExpenses
DepositZelle®Pay a Bill
MAR 18, 2022
VenmoDeposit · One Time · Checking 1234
+$400.00
MAR 17, 2022
NikeShopping · One Time · Checking 1471
$120.00
UberTransportation · One Time · Checking 1471
$24.00
Why this matters
Merchant name, category, accounton every row
9:41
Manage Your BudgetCategory detail · Rent
DepositZelle®Pay a Bill
$2,050.00 Rent · 1 Transaction this month 30% of this month's planned expenses
Planned$2,050.00
Remaining in Housing$0.00
Skyline Property MgmtMar 1 · Recurring · Checking 1234
$2,050.00
Skyline Property MgmtFeb 1 · Recurring · Checking 1234
$2,050.00
Skyline Property MgmtJan 1 · Recurring · Checking 1234
$2,050.00
Category insight
Same amount, six straight monthsposts on the 1st
9:41
Manage Your BudgetDaily Budget
DepositZelle®Pay a Bill
$91 Left to spend per day. There are 7 days remaining in this period
Estimated Projection
$425.00An estimated amount that will not be used for the budget period

The last three days

Friday$38 of $91
Saturday$104 of $91
Sunday$61 of $91
How it is figured
$637 unspent this periodspread across 7 days
9:41
Monthly Budget InsightsSpent / Limit
DepositZelle®Pay a Bill
Over Limits
Income$6,300 / $6,000
Housing over $300$2,700 / $2,400
Under Limits
Lifestyle$500 / $800
Perfecton four of six categories

On track

Food$300 / $400
Transportation$190 / $300
Savings$400 / $400
New Limits
9:41
Monthly Budget InsightsUpdated budget recommendation
DepositZelle®Pay a Bill
Recommend budget updates

To avoid going over and remain on your budget we have noticed that in the last three months, you average $300 less than what you budgeted for lifestyle. However, you also go over $300 in your housing. We have offset the differences to help you maintain your budget.

New Limits

Housing (updated)$2,700
Food$400
Lifestyle (updated)$500
Transportation$300
Savings$400
Miscellaneous$400
Accept
Customize
9:41
Monthly Budget InsightsCustomize new limits
DepositZelle®Pay a Bill

Tune each limit before it takes effect

Housing−$2,700+
Food−$400+
Lifestyle−$500+
Transportation−$300+
Savings−$400+
Miscellaneous−$400+
Live total
$4,700 plannedof $6,000 monthly income
Save Limits
Back
  • 01
    Expense OverviewThe glance layer. Every category carries spent against limit on one screen, so "am I okay?" is answered before any tap.
  • 02
    ExpensesOne tap deeper. Merchant name, category, frequency and account on every row, the fix for transaction confusion.
  • 03
    Category detailAnother tap: what a single line item is, how many transactions made it, and its share of the month.
  • 04
    Daily BudgetThe same data reframed as a decision. Not "you spent $2,050" but "you have $91 a day for seven days."
  • 05
    Monthly Budget InsightsOver and under limits, side by side, ending in one action: set new limits for next month.
  • 06
    Updated budget recommendationThree months of history, read as one move: consistently $300 under on lifestyle, consistently $300 over on housing, so the app offsets the two and hands back limits that fit.
  • 07
    Customize, on their termsAccept is one tap, but every recommended limit can be tuned first. The machine suggests, the customer decides.
The design · onboarding
DesirabilityFeasibilityViability

Sixty questions became a conversation.

Goal setup was specced as one long enrollment form, fifty to sixty fields of paperwork. I rebuilt it as ten supportive steps: one decision per modal, encouragement between steps, and a finish line you can see. People organizing their money are already stressed. The interface should cheer, not interrogate.

Step 1, name the goal Step 2, start date Step 3, optional end date Step 4, goal amount Step 5, funding account Step 6, recurring transfers choice Step 7, manual savings reminder Step 8, completion alerts Step 9, review with edit links Step 10, congratulations
  • 01
    Name the goalA goal with a name is a promise. One field, one decision, and the toggle folds the goal into the budget automatically.
  • 02
    Start dateToday by default, changeable in one tap. The road illustration keeps the journey framing going.
  • 03
    End date, optionalA yes or no before the date picker. Deadlines motivate, but only when the user chooses them.
  • 04
    The amountOne number. No income interrogation, no worksheet.
  • 05
    Funding accountOne account or several, and manual contributions stay allowed either way.
  • 06
    The recurring forkThe only nudge in the flow, stated as a fact: automatic savers put away about twice as much. No is a first-class answer.
  • 07
    The honest fallbackChose No? The flow says plainly that monthly savings are now manual. Clarity without guilt.
  • 08
    Alerts you chooseProgress alerts at a percentage the user picks. Encouragement, scheduled.
  • 09
    Review, all of it editableEvery answer on one card with an edit link. Nothing commits until Submit.
  • 10
    Celebration, not confirmationA dedicated goal account is created on the spot. The last screen claps.
"Nobody is ever going to use that." One product partner held the line for the form, so I called the test: twenty participants, A/B, and twenty of twenty chose this flow. It won its place in the validated build. PFM never launched, but the pattern is tested and specified, so when this rolls out again, onboarding starts finished.The real screens from the Figma file, shown here
Test & iterate
DesirabilityFeasibilityViability

We saw it, we changed it, we measured it

Four findings kept proving true across every round, and each one became a design change we could measure. One user model sat underneath all of them: set-and-forget people and fine-tuners, and nothing could punish either. Click a card for the evidence.

We saw people quit during budget setup

One-tap starter budget

A draft budget generated from the user's own spending history, editable from there.

Setup: about 7 minutes down to about 2
7 min

median time to finish budget setup in the first tested build

2 min

after the one-tap starter budget, built from real transaction history

Every round

setup was named as the reason to quit, in every testing round we ran

We saw automation without agency stall

Customizable round-ups

My concept: fixed nearest-dollar round-ups tested against increments the customer sets, from $1 to $20.

Meaningfully higher opt-in with user control
Control > automation

participants who could set the increment turned it on; those given a fixed rule refused it

RoundUp +

my design, and the basis of one of the patents filed from this work

Tested, not assumed

Customization was my instinct, but I never asked the team to take it on faith. We put it in front of users every round, and the evidence is what carried it, not my argument.

We heard "what is this charge?" constantly

Human-readable transactions

Merchant logos plus plain labels, like turning "WF Mortgage" into "Home Loan."

Transaction confusion cut by about one fifth
Merchant names

the single most common source of confusion in transaction review

Plain language

our content strategist rewrote transaction copy off the back of this finding

Trust, not UI

people who could not read a charge stopped trusting the whole balance

We saw warnings get ignored

Nudges over alarms

Notifications rewritten as gentle nudges, data requests explained in plain language.

More check-ins, more willingness to link accounts
Alarms read as blame

red overspend warnings made people close the app rather than open the budget

Plain permissions

saying why we needed a data connection raised willingness to link outside accounts

More check-ins

the softer tone brought people back more often, which is the whole activation game

From changes to proof

Every change earned its number.

What comes next is what we could prove, not what we hoped. Measured outcomes are what make a project like this hold its value: the product did not ship, but the evidence did, and it is already paying for whatever comes next.

Measured outcomes
DesirabilityFeasibilityViability

What we could already prove in testing

Click a number for how we measured it. The gold line is what each result is worth, modeled on slide 19.

Budget setup time 7→2 min How long it took a participant to get a working budget, before and after the one-tap starter budget. Worth $416M of modeled revenue: budget creation and monitoring in Visa's assessment for TD. On the cohort model, about $1.8M a year per 10 points of completion. Click for the method
Usability testing, before & after
  • Same moderated task every round: "set up a budget you would actually keep"
  • Manual flow benchmarked first: median about 7 minutes, with drop-offs at category setup
  • Retested after the one-tap starter budget: median about 2 minutes, completion up
  • Medians across testing rounds, not a controlled lab study
Transaction recognition ~20% fewer "What is this charge?" moments, once rows carried merchant logos and plain-language labels instead of raw descriptors. Worth a share of $670M: the holistic financial summary line in Visa's assessment, where recognised transactions live. Plus about $100k a year in avoided calls. Click for the method
Concept evaluation, round 2
  • Two versions of the same feed: raw descriptors vs logos + plain language
  • Participants walked their last two weeks of spending in each
  • We counted every "what is this charge?" moment per session
  • About 20% fewer on the redesigned rows; confusion clustered on subscriptions
Week-one activation ~3× the intent Nothing had shipped, so there was no app to keep using. Weeks after their sessions, the participants who had connected an outside account and set one savings rule were about three times as likely to still want it. Worth $150 to 300M of retention and lifetime value over three years in the committee model. On the cohort model, $8 to 15M a year of acquisition spend not needed. Click for the method
Follow-up interviews, not live usage
  • Nothing was live to keep using, so we re-contacted participants weeks after their moderated sessions
  • Split by what they did in-session: connected an outside account and set one savings rule, or not
  • Still wanted it weeks later: 68% vs 23%, about three to one
  • A leading signal on a small sample, which is why proving it at launch became the KPI
Round-up enrollment Control > automation Opt-in rose when the customer picked the increment themselves. A fixed, bank-chosen amount was refused. Worth $621M of modeled revenue: savings automation, the largest line in Visa's assessment. The fixed increment measured zero opt-in, so the customer-set increment is the switch on all of it. Click for the method
Prototype 4 testing
  • Two variants, order rotated: bank-fixed increment vs user-picked $1–$20
  • Measured stated opt-in and trust language in the debrief
  • The fixed amount was refused outright in every round
  • Choosing the increment flipped the same feature from "creepy" to "mine"
Activation metric we defined: connect one outside account and set one savings rule within the first week. In follow-ups the ones who did showed nearly three times the intent to keep going, the signal we would have proven at launch.The would-be north star
The business case
DesirabilityFeasibilityViability

Defunded before launch. So here is the case, modeled honestly.

How many customers would turn PFM on

Out of 1 million customers who get offered it, counted every two months for two years. Illustrative.
Expected caseConservative caseHover any point for that month

Read it as: uptake is slow for the first six months, steepest between months 6 and 14, and flattens once the interested customers have signed up. The shaded area is the gap between the two cases, which is the range we would actually be arguing for.

  1. $2.5B. Visa Consulting & Analytics, business impact assessment for TD's PFM MVP, 2022: savings $979.5M, budgeting $868.8M, holistic financial summary $669.8M. Savings automation alone $620.8M. Visa's own caveats: no costs or time to implement; features may double count.
  2. $835M to 1.5B. The patent this research produced, modeled for TD's IP and Ideation Committee, Nov 2025: deposit and investment growth $420 to 850M, cross-sell $180 to 360M, service cost $35 to 70M, retention and lifetime value $150 to 300M. Three years, against a $10 to 20M build.
  3. 3 in 5 enroll, 99% stay. Financial Health Network, Keep the Change, 2019. Curve shape.
  4. 14M+ users, $30B+ saved. Acorns press release, May 2026. Scale check.
  5. 95% retention for engaged digital users; $800+ to acquire a customer. Lumin Digital, Jan 2026. 16% higher net profit; 43% vs 5%. Fiserv and Northwest Bank, 2023.
  6. $35 to 46M. 318,000 enrolled × $110 to $145 a year of round-ups ($9 to $12 a month); Keep the Change ran $2B across 12M+ enrollees.
  7. Patents. Software royalties 8 to 12% (PatentPC 2024, Stanzione 2026). USAA mobile-deposit verdicts: $200M and $102.7M v Wells Fargo, settled; $218M v PNC, overturned 2025.

Why the first week decides everything

Customers still using PFM six months later, split by what they did in week one
Activated in week one
68%
Did not activate
23%

Activated means they linked one outside account and set one savings rule. Same product, same customers: doing that in the first seven days makes a customer three times as likely to still be here six months later. That is why week-one activation, not sign-ups, became the metric we defined.

$2.5Bmodeled revenue lift for the MVP, Visa Consulting & Analytics for TD, 2022. Savings $980M, budgeting $869M, summary $670M
$835M–1.5Bthree-year value of the patent this research produced, against a $10 to 20M build. Modeled for TD's IP committee, Nov 2025. Pending, not granted
$35–46Ma year in customer savings from round-ups on our 318,000 cohort. The floor, built only from public benchmarks

Two scales, both modeled and labeled that way in the room. Bank scale: Visa's assessment and the committee model, built for TD. Cohort scale: our research signals on public benchmarks, 1 million offered, 318,000 enrolled. The committee model builds on Visa's savings automation line, so the $3.4 to 4.0B cover total is an upper bound. Every formula and source: the last page of this deck.

Reflection

It did not ship. The work still compounds.

✓

What survived

Patents filed from this work

User-set round-up increments, and the budget-to-savings reallocation loop with the goal recalculation that follows it. Several of the six patents I have filed to date came from this work. The reallocation loop became the filing modeled for TD's IP committee at $835M to $1.5B over three years against a $10 to 20M build.

  • A backlog that is designed, not imagined. Nothing here is a sketch. The flows are built, tested and specified, so when funding returns the work restarts at implementation rather than at discovery. That is a year of runway already paid for, against the $2.5B of revenue lift Visa assessed for this MVP.
  • Features that travel. These pieces do not have to wait for a PFM launch to earn their keep. Round-up increments, the goal-recalculation loop and the plain-language transaction rows all drop into existing experiences on their own, and each one is a win wherever it lands. Savings automation alone carried $621M of Visa's number.
  • What the organization kept regardless. Validated patterns the wider design org absorbed, a research corpus other teams drew on, and the week-one activation insight, which is portable to any engagement product in the bank.
↻

What I would do differently

Three things, and they are all the same mistake: I proved the design worked long before I proved it paid.

  • Speak in the bank's numbers, from day one. I reported setup dropping from seven minutes to two. Leadership funds retention and deposits, not minutes saved. The same finding, framed as "activated customers stay three times longer, savings automation is worth $621M of a $2.5B lift, and round-ups route tens of millions into savings accounts," is a different conversation.
  • Ship a thin slice before the budget cycle, not a complete design after it. Everything we had was tested, not live. Even one pillar in production would have replaced my modeled curve with real enrollment data, and real data is much harder to defund.
  • Put finance in the room for the research readouts. They heard our conclusions second-hand, months later, from a deck. Had they watched a participant abandon setup, the business case would have been theirs to make, not mine to sell.
The business case slide you just saw is the argument I should have made a year earlier. I now treat the business narrative as a design deliverable.Honest hindsight
Where this travels

Three problems took most of this runway: hard rules the interface can never break, trust in automation that has to be earned one control at a time, and a first week designed on purpose. Every product I have worked on since has had all three, inside banking and well outside it. The next slide reads the same work from the seat you are hiring for.

9:41
Manage Your Budget
DepositZelle®Pay a BillTransfer
Expenses$3,000.00
Income$5,000.00
Expenses Spent of Income60%
Savings and Goals
Total Monthly $2,200 | 73% of Expenses
Vacation Goal Total $10,000 | 8% Completed
Contributed $800 of $2,200 Monthly36%
3 Months leftDecember 2024
Food2 categoriesTotal $300 | 40% used
Lifestyle2 categoriesTotal $250 | 8.3% used
Transportation2 categoriesTotal $250 | 8.3% used
Savings and Goals
Total Monthly $1,200 | 80% of Expenses
Vacation Goal Total $10,000 | 8% Completed
Contributed $800 of $1,200 Monthly66%
⚠ 7 Months leftApril 2024
Vacation Goal Changes
Current36% | 3 Months Left | Dec 2024
New66% | 7 Months Left | April 2024

Your budget change will affect your end date. If you want to cut cost, you can view our recommendations so you can keep your goal on track.

View New Budget Recommendations
$10,000Save
Goal total
$1,200Save
Total Monthly Contribution

Patent-pending · edit any input and the goal recalculates live, then the product recommends the budget moves to recover the date

How this translates

The same work, read from six seats.

The seat changes what you look at. The work does not. Here is what to pull from this case study depending on the role you are hiring for, and where in the deck it lives.

1

Product design, senior to principal

One control taken three levels down, edge cases that each got a screen, WCAG 2.1 AA before any round went out, and every tested build through my file.

The round-up control · Onboarding · Test and iterate
2

Design leadership

Clear lanes inside one team, a research muscle built before the product, criteria agreed before either test ran, and a stakeholder fight settled by evidence rather than rank.

My role · Research · The road not taken
3

Design systems

The bank's system where it served us, new patterns contributed back where it had gaps. In the appendix, a system I own end to end: tokens aliased in Figma and consumed by the code.

Budgeting & savings · The design system, published
4

AI and agentic product

Automation people actually turn on: control before automation, the consequence shown before the commit, and a recommend-then-recalculate loop that keeps the person deciding. Then a product built by 19 agents I direct.

The round-up control · Reflection · What the room holds
5

Brand and interactive design

PFM lived inside a brand set before I arrived. Hello Court is a brand authored from nothing, and this deck is itself an interactive build with the real components embedded in it.

The appendix, start to finish
6

Strategy and business partnership

Three lenses held from discovery to the business case, a week-one activation metric the research produced, and a modeled case I now treat as a design deliverable.

Outside research · Measured outcomes · The business case
Underneath every seat

The same three problems: hard rules the interface can never break, trust in automation that has to be earned, and a first week designed on purpose. Regulated or not, consumer or enterprise, that is the work.

Four minutes, and then your questions

Two things that case study cannot show you.

PFM is the case study. What follows is not a second one. It is evidence for two things PFM structurally cannot show, however recent it is. One, a brand I authored from nothing: at TD I designed a great deal from scratch, but always inside a palette, type and principles set before I arrived. Two, a whole product I built and run end to end: the systems, the security, the agents and a working app, not just the screens.

The honest part first, before you ask. Neither of these has launched. Hello Court is fully built and functionally tested and it is the one I am carrying forward. Quorum I paused deliberately to make that possible. I am showing you craft and systems, not adoption numbers.

hellocourt.app
Hello Court home page on desktop
Hello Court home page on mobile
hellocourt.app · coded responsive, not two compsNext.js 16 · React 19 · TypeScript · Tailwind 4 · Supabase · Vercel
Appendix · brand authorship · hellocourt.app/planning

The Blueprint. The whole build, published while it happens.

Hello Court’s planning room is not a deck about the work. It is a live site I designed and built, generated from the work itself: 65 pages across 13 sections, behind an access code I share with employers, collaborators and investors.

hellocourt.app/planning · behind an access code
The Blueprint home page: the planning room, open while the work is happening
The planning room, open while the work is happening · created 2026
Generated, not hand built

A Python pipeline I wrote renders every page from one style source. The data pages read sprint status, the time log and git history at build time, so the site cannot drift from the work it describes.

The sections

Planning, case study, team, design system, data, SEO, security, prototype, journeys and a change log. Each one is the real working artifact, not a summary written after the fact.

Honest by construction

Where something is unfinished, the page says so. Showing the room mid flight is the point: decisions get revisited and the counters refresh on every rebuild.

Appendix · what the room holds

Security, search and an agent workforce, with receipts.

Security, set as the floor

28 caught, 17 high severity

8 independent reviews, 6 tools run against the code. Every finding is published with its impact and status, never the exploit path.

Zero secrets. Zero user data used to train models
The toolkit Cipher runs
gitleakstrufflehogsemgrepnpm auditpip-auditNmapNuclei

Secret scanning, SAST, dependency and supply-chain, recon and vuln scanning, OWASP Top 10 and access-control review, plus an edit-time tripwire on risky patterns. The full kit stays private by design.

2

A search engine from a standing start

One day in June took the site from one indexable article to a 49 page library. Three weeks later: 50 pages indexed, 272 queries, and the first clicks, including a searcher who looked up a Georgia statute and landed on the Georgia guide from page one.

Search Console numbers published, including the zeros
3

19 build agents, one legal bench

The product is built by 19 agents I direct across desirability, viability and feasibility, over a research bench of 411 AI personas across all 50 states. The page says it in bold: AI personas, not lawyers, and none of it is legal advice.

Every agent documented: what they run, what they produce
The security page stat tiles: 28 issues caught, 17 high severity, 8 reviews, 6 tools, 0 secrets exposed, 0 data used to train models
hellocourt.app/planning · security · the work, by the numbers
Appendix · the design system, published

A design system you can open, connected to Figma.

Published as a living section of the site and mirrored into Figma variables: foundations, 15 components and 3 interaction patterns, documented with the same tokens the code consumes.

Tokens, aliased not hardcoded

The Figma file carries 67 Color variables plus Radius and Spacing collections. A component consumes a semantic name, accent, surface, danger, and that name aliases to a primitive, so one edit moves the whole system. Design file and running product cannot drift.

Accessibility is a section, not a footnote

An ADA AA audit sits in the foundations beside type and color. Every pairing is verified for WCAG AA in both modes, and sage text on light is pinned darker to clear 4.5 to 1.

No detached copies, every state previewed

Every app screen is built from component instances, zero detached copies. Anything with a hover, a load or a transition has a live preview: if you cannot watch it move, it is not specified.

The changelog cuts against me

The first entry in the color changelog reverses a decision I had defended: danger moved from amber to red, so danger reads as danger. A system that can overrule my taste is the proof that it is a system.

Figma variable panel for the Hello Court design system: 67 Color variables plus Radius and Spacing, semantic names aliasing to primitives
Figma · Color, Radius and Spacing variables · semantic names alias to primitives
The published Hello Court design system page: foundations, components and patterns
hellocourt.app/planning · design system · foundations, components, patterns
Appendix · the shipped surface · hellocourt.app

Not mockups. The app, running.

The working build, recorded · a finished case, then a new one from intake · loops
The Hello Court case home: packet completeness, pick up here, and a plain language to do list
The case home · packet completeness, pick up here, and a to do list in plain words
What the screens argue

One decision per screen. Save and exit always there. English and Spanish from the first tap. A ten step intake that meets a frightened person in plain language, then a case home that always answers: what do I do next.

Appendix · range, and a decision

The second one I paused on purpose.

Quorum is an AI deliberation tool for boards and executive teams. Next.js 16, React 19, Prisma, real auth, Playwright end-to-end tests, and a design-system route that lives inside the application rather than in a separate file that drifts away from it.

1

Why it is here

The second production-grade build, and it shows the pattern I work in: the system is not a document about the product, it is a route in the product. Design and code are the same artifact.

Design system as a running surface, not a PDF
2

Why it is paused

Two products, one person, and both getting a fraction of what they needed. I picked the one with the larger problem and the clearer user, and I stopped the other. It was the right call and it was not a comfortable one.

The pause is the point, not the footnote
quorum.plus · session
Quorum session workspace
quorum.plus
Quorum landing page
Paused, not abandoned · Next.js 16 · React 19 · Prisma · Playwright
Working together

Strategy needs a maker.

I have spent years at senior and director level, in the rooms where roadmaps get funded: executives, partners, vendors, and the tradeoffs between them. That is unusual for a designer, and it travels to any seat. It does not make me a better designer than the people already on your team. It makes me an easier partner for them, because I can carry the stakeholder conversation without anyone having to translate the work first.

The seat can be an individual contributor or a lead, and the loop is the same. Give the team a direction on Monday and there is a testable prototype by Thursday: UI, interaction, edge cases, accessibility, and evidence in the room. Leading, I run that loop through the team, set the criteria before the test, and clear the stakeholder path. When the business case is what is missing, I will write it, and the credit stays with the people who built the work.

Figma + live prototypes Design systems Research fluency Regulated-domain depth Stakeholder & partner management 6 patents filed
2025
Nominated for Visionary of the Year Put forward for this work, which also produced several of my six filed patents
Let's dig into whatever interests you most

Jay Frank · Principal Product Designer / Strategist · Thank you

Appendix · sources and math · not for presentation

Where every number comes from.

The cover total: $3.4 to 4.0B

Two models built for TD, added together. Both are modeled, not booked; PFM was defunded before launch.

ComponentLowHigh
A. PFM MVP revenue lift (Visa, 2022)$2,518M$2,518M
B. Patent filing 1, three years (committee, 2025)$835M$1,500M
Total$3,353M$4,018M

Caveat: the committee model cites Visa's savings automation line as evidence for its deposit growth, so A and B overlap in part. Treat the total as an upper bound. Visa's horizon is not stated on its slide; the committee figure is three-year cumulative.

A. Visa Consulting & Analytics, business impact assessment: increased revenue

Built for TD to rank the MVP features across credit, debit, checking and savings. Slide 12, 2022, Visa Confidential. Reviewed with the team Jan 31, 2023 (source file timeline).

CapabilityCheckingSavingsTotal
1.0 Holistic financial summary$349.8M$219.2M$669.8M
1.1 Account aggregation and visualization$306.5M$195.3M$542.6M
1.2 Customizable visualization$23.4M$12.6M$68.1M
1.3 Upcoming transactions$19.9M$11.3M$59.2M
2.0 Budgeting$464.3M$289.9M$868.8M
2.1 Budget creation and monitoring$219.1M$148.4M$415.9M
2.2 Budget customization$25.6M$13.0M$72.1M
2.3 Forecasted spend insights$219.6M$128.5M$380.7M
3.0 Savings$534.6M$333.7M$979.5M
3.1 Savings customization$167.5M$96.6M$290.5M
3.2 Savings automation (round-ups)$346.1M$225.5M$620.8M
3.3 Savings optimization$21.0M$11.6M$68.2M
All three capabilities$1,348.7M$842.8M$2,518.1M

Totals include credit and debit columns not shown here (credit $117.2M, debit $209.4M). Visa's caveats, quoted: checking and savings portfolio size factors heavily; revenue impact does not factor in costs or time to implement; a feature's impact may double count another's.

B. The patent this research produced

Automated Adjustment of Visual Elements in a Personal Financial Management System. Filed through TD Invent; 2025 Visionary of the Year finalist, one of three bank-wide. Modeled for the Intellectual Property and Ideation Committee, Nov 2025. Not audited results.

Component, three yearsLowHigh
Deposit and investment growth$420M$850M
Cross-sell revenue$180M$360M
Service-cost savings$35M$70M
Retention and lifetime value$150M$300M
Total$835M$1,500M
Build cost, once$10M$20M
Operating cost a year$5M$7M
Value to build42×150×

Inputs as submitted: managed deposit uplift +5 to 10%; digital cross-sell conversion +15 to 25% (JD Power, BCG, BAI benchmarks: satisfaction up 20% lifts cross-sell 15 to 25%); digital interaction about $0.10 against $4.00 in a branch or call centre; digital transformations lifting bank revenue about 31%; Visa's savings automation line as the deposit evidence, discounted. Adoption ramp 20 to 30% to 50 to 60% over three years.

Patents, three lenses kept separate

  • Income: filing 1 as modeled above. Filing cost for six patents $180 to 360K (AIPLA, $30 to 60K each through grant), so 2,300 to 8,300× if filing 1 ships as modeled.
  • Market: software patent assets listed at a $163K median asking, $95K median sold (Richardson Oliver, 2024): $0.6 to 1.0M for six. Closing prices run 0 to 75% below asking.
  • Licensing: software royalties 8 to 12% of attributable revenue (PatentPC 2024; Stanzione 2026). On a licensee with our cohort's $9.7 to 17.8M engagement lift: $0.8 to 2.1M a year per licensee.
  • Precedent: USAA v Wells Fargo $200M (2019) and $102.7M (2020), settled; USAA v PNC $218M (2022), overturned by the Federal Circuit 2025 as abstract. Eligibility is the risk for software claims.

C. Cohort model, public benchmarks only

1,000,000 customers offered; 318,000 enrolled at month 24 expected, 174,000 conservative (slide 19 curve).

  • Round-ups $35 to 46M a year: 318,000 × $110 to $145 ($9 to $12 a month, $1 to $5 increments on 20 to 40 purchases). Conservative $19 to 25M.
  • Interest on balances: $35 to 46M × 3.23% FDIC net interest margin (Q3 2024) = $1.1 to 1.5M year one; $2.3 to 3.0M year two as balances stack at 99% retention.
  • Engagement lift $10 to 18M a year: 16% higher net profit (Fiserv and Northwest Bank, 360,000 customers) × $350 revenue per checking account (StrategyCorps) = $56 × 174,000 to 318,000.
  • Setup completion, about $1.8M per 10 points: 31,800 more live budgets × $56. Time returned: 5 minutes × 318,000 = 26,500 hours.
  • Activation $8 to 15M a year: 40% of enrolled activate in week one (assumption); retention 80% to 95% (Backbase, CoinLaw, Lumin) keeps 10,400 to 19,000 customers × $800 acquisition cost (Lumin).
  • Calls avoided about $100K a year: 1 in 4 enrolled call once, 20% fewer confusion moments (our round 2), $6.47 a call (MaestroQA).

Sources

  • Visa Consulting & Analytics, TD business impact assessment, slide 12, 2022. Held in the TD PFM working files; the MVP ordering from it sits on the "TD Bank Business Impact Exec Readout" board in the source Figma file.
  • TD IP and Ideation Committee submission, Nov 2025: Visionary OPI deck, script and Q&A (questions 2, 6, 7, 8, 9, 15). Not audited.
  • Financial Health Network, Keep the Change, 2019. finhealthnetwork.org/keep-the-change-how-bank-of-america-pioneered-roundup-savings
  • Acorns press release, May 2026, 14M+ people, $30B+ saved. prnewswire.com (Acorns, Money Therapy and 2026 Financial Wellness Report)
  • Lumin Digital, Driving double-digit engagement growth, Jan 2026. lumindigital.com/insights/driving-double-digit-engagement-growth
  • Fiserv and Northwest Bank, Are your banking customers digitally engaged? fiserv.com/en/insights/articles-and-blogs/are-your-banking-customers-digitally-engaged.html
  • FDIC Quarterly Banking Profile, Q3 2024, NIM 3.23%. StrategyCorps, Economic Realities of Retail Checking. Backbase and CoinLaw retention statistics, 2026. MaestroQA cost per call.
  • PatentPC 2024 and Stanzione 2026 royalty rates; Richardson Oliver 2024 patent market; AIPLA cost survey.
  • USAA verdicts: TPR 2019, Insurance Journal 2020, Law360 settlement; Bloomberg Law 2022 and Banking Dive 2025 (PNC reversal); Banking Dive 2024 (Truist settlement).