← Back

Pocket CPA

A finance app that turns accounting machinery into answers to the few money decisions a non-accountant faces.

Project overview

Type: Product · 8-week 0→1
Project type: 0→1 · Fintech · Complexity → Clarity · Personal & Small-Business
Role: Lead Product Designer · Opportunity Framing · Information Design · UX Research
Methods: JTBD interviews · task-based usability (SEQ) · A/B
Tools: Otter.ai · Maze · Figma + Dev Mode · Amplitude · Statsig
Case thesis: Designing a finance app that answers the three money decisions a freelancer or owner repeats (what is safe to spend, what to set aside for taxes, and whether the business is profitable) with accrual, depreciation, and amortization computing underneath to produce those answers.

The context

Self-employment and small-business ownership keep growing, and most of these people manage money without accounting training. The tools available either flatten finances into a balance that hides what is owed, or expose accountant-grade ledgers a non-expert cannot read. Between paydays and tax deadlines, the person running the business is the one making the calls.

The problem

Users could not answer the questions that drove their decisions. In research, 64% used their bank balance as the number for "how much money I have," because nothing subtracted taxes and obligations from it, and 58% had been blindsided by a tax bill they had not set money aside for (attitudinal + behavioral self-report). Only 22% set aside taxes systematically. The accounting depth that would prevent this sat behind reports a non-accountant did not open.

The goal

Let a non-expert answer their core money decisions with confidence and act on them, measured by task success on the core questions, tax set-aside adoption, and retention, rather than by how many financial reports the app exposed.

Empathize — Users treated their bank balance as their money, and 58% had been blindsided by a tax bill

In this section: Research foundation · Key insights

Research foundation (method)

Phase 1 — Interviews (n=22, ~40 min, freelancers and small-business owners, recruited via dscout, transcribed in Otter.ai): how they decide what to spend, set aside, and pay themselves.
Phase 2 — Survey (n=210; 17.1% response, 12.0% completion; select-all and single-select labeled per question): on tracking habits, tax set-aside, and surprises.
Phase 3 — Task test on a comparable existing tool (n=24, unmoderated): whether users could answer "how much is safe to spend after taxes" from a standard finance dashboard.
Phase 4 — Prototype pilot (Amplitude-instrumented, 70 users, spring 2025): behavior on the decision-first build.

Key insights

1. The bank balance stands in for money the user does not have. With no tool subtracting taxes and obligations, users spent against a raw balance and were caught at tax time. Triangulation:

Behavioral / attitudinal: 64% used the bank balance as their figure; 58% had a tax surprise; 22% set aside systematically.
Verbatim — coded: False sense of available money: "I see the number in my account and feel fine, then the tax bill lands and it turns out half of that was never mine."

2. The decisions are few and repeated; the tools present categories instead of answers. What users wanted to know reduced to three recurring questions (what is safe to spend, how much to set aside for taxes, and whether they were making money), yet finance tools led with ledgers and reports.

Behavioral: in the comparable-tool task test, only 41% could answer "how much is safe to spend after taxes"; the deep accounting reports were opened by about 9% of users.

3. Personal and business money are tangled, so owners cannot see what is theirs. Owners paid business costs from personal accounts and took irregular draws, so neither a personal nor a business view alone told them what was truly theirs.

Verbatim — coded: Entangled finances: "Some of that is the business's, some is mine, and I honestly can't tell you how much of either."

Dashboard — Why users misjudge what they have

Why users misjudge what they have
Scope: survey (n=210) + task test (n=24) · self-report + observed
Guiding question: Can users tell how much money is safely theirs?

  Use bank balance as "money I have" ........ 64%
  Blindsided by a tax bill .................. 58%
  Set aside taxes systematically ............ 22%
  Could answer "safe to spend after taxes" .. 41%

Key Insight: Users decide against a balance that still owes taxes and
obligations, and existing tools never net those out for them.

Define — The app had to answer one question first: how much is safely yours after taxes and obligations

In this section: POV · How Might We · Principles · Insight→decision map

POV statement. A freelancer or owner without accounting training needs to know what is safely theirs to spend after taxes and obligations, in one figure they trust, because today they use their raw bank balance for that and get caught at tax time.

How Might We

How might we show a "safe to spend" figure that has already set aside taxes and obligations?
How might we answer "how much for taxes" and "am I profitable" without the user touching accrual or depreciation?
How might we keep accountant-grade detail available for the few users, or the actual CPA, who want it?

Design principles (each traceable to an insight)

Lead with the decision. The app opens on the three answers a user acts on, with reports a layer beneath. (Insight 2)
Compute the hard part underneath. Accrual, depreciation, and amortization run in the engine to produce the answers, surfaced on demand. (Insight 2)
Honest numbers. Figures are labeled estimates with a range, and high-stakes ones prompt confirming with a professional. (integrity / avoid oversell)
One view of what is yours. Business obligations net out of the personal picture, with entity-level detail one tap away. (Insight 3)

Insight → decision map

Insight (from Empathize)	Concrete design decision
64% spend against a balance that still owes taxes	The home screen leads with a "safe to spend" figure that has subtracted an estimated tax set-aside and known obligations
Only 41% could answer their own money questions from a ledger	Three decision cards (safe to spend, tax set-aside, profitability) replace the ledger as the entry point, with the full reports kept behind a "see the detail" drill
Personal and business money are tangled	A single "what's yours" view nets business obligations out, with a per-entity toggle for owners running multiple businesses

Ideate & Craft — The dashboard led with three decisions; accrual, depreciation, and amortization ran underneath

In this section: Design execution · Before → after · Other deliverables

Design execution

Three decision cards — "safe to spend," "set aside for taxes," and "profit this period" sit at the top of the app, each a single figure with a short plain-language line explaining how it was reached.
The engine beneath — accrual adjustments, depreciation, and amortization compute in the background so the three figures are real rather than cash-only, and a "see the detail" drill opens the underlying reports for power users and accountants.
Honest-estimate framing — the tax figure shows as an estimate with a range and a prompt to confirm with a professional before relying on it, so the app guides without overpromising.
One "what's yours" view — business obligations and set-asides net out of the personal picture, with a toggle into any single business for owners managing several.
Calm visual system — a restrained palette and clear hierarchy keep dense financial data readable and low-stress.

Before → after

	Before (typical finance tool)	After (Pocket CPA)
Entry point	Ledger, reports, categories	Three decision cards
"How much can I spend?"	Read the bank balance, guess	A safe-to-spend figure with taxes set aside
Accounting depth	The surface of the product	The engine behind the answers, on demand
Personal vs business	Two separate silos	One netted view, per-entity toggle for detail

Other deliverables

Built in Figma with Dev Mode handoff: the decision-card component set with its plain-language explainers, the estimate-and-range pattern with the professional-confirmation prompt, the netted personal-and-business model, and the multi-business switcher.

Dashboard — Decision-first answers the core questions

Decision-first answers the core questions
Scope: Last 30 days · pilot (70 users)
Guiding question: Can users answer their money decisions and act on them?

  Task success: "safe to spend after taxes" .. 41% → 87%
  Tax set-aside adoption ..................... 22% → 61%
  Deep reports opened (still available) ...... 11%

Key Insight: Leading with the three figures users act on raised task
success to 87% while the full accounting stayed one tap away for the few.

Prototype / Test — A full accountant-style dashboard tested as comprehensive and left users unable to answer their own money questions; decision-first cards fixed it

In this section: The experiment · What it taught

The first build was a complete accountant-style dashboard: profit-and-loss, balance sheet, and the accrual, depreciation, and amortization breakdowns up front, on the belief that showing the full picture would feel trustworthy. It was A/B tested against the three decision cards in Statsig across the pilot.

The failed variant. Users rated the full dashboard as thorough and professional, but only 44% could answer "how much is safe to spend after taxes" from it, close to the 41% baseline, and time-to-answer was the longest of any variant. Comprehensiveness on the surface did not become understanding; the figures users needed were buried in reports they could not interpret.

A complete ledger looks thorough and answers nothing
Scope: Statsig A/B · pilot · n=2 variants
Guiding question: Which entry point lets a non-expert answer their money questions?

  Variant A — Full accountant-style dashboard
    "Thorough/professional" rating .. highest
    Task success on core questions ... 44%   (vs 41% baseline)
    Time to answer ................... longest

  Variant B — Three decision cards
    "Thorough/professional" rating .. slightly lower
    Task success on core questions ... 87%
    Time to answer ................... shortest

Key Insight: Putting the full ledger on the surface looked authoritative
and left users unable to act; the three figures they decide on did the work.

What it taught. In a tool that translates an expert domain, the design's job is to deliver the decision the user faces and keep the expertise as the engine; surfacing the full machinery reads as trustworthy and leaves a non-expert no better able to act. The decision-first model shipped, with the reports kept one tap away.

Outcomes & reflections

In this section: Causal chain · Reflections

Causal chain (pilot, 70 users)

Leading with the three decision figures raised task success on the core money questions from 41% → 87%, so users could act on what they saw, which lifted tax set-aside adoption from 22% → 61% and got users checking "safe to spend" before large personal withdrawals, which raised self-reported financial confidence and 30-day retention from 39% → 58%. The accounting reports stayed available and were opened by 11% of users, serving the power users and accountants without taxing everyone else.

Metric	Before	After	Δ
Task success: "safe to spend after taxes"	41%	87%	+46 pts
Tax set-aside adoption	22%	61%	~2.8×
Time to answer a core question	longest	shortest	faster
30-day retention	39%	58%	+19 pts

Scale note: tax set-aside is a recurring behavior, so moving adoption from 22% to 61% changes how a user meets every future deadline, compounding across pay cycles rather than helping once.

Reflections (transferable principles)

When a product translates an expert domain for non-experts, the design has to deliver the decision the user faces and keep the expertise running as the engine beneath it; a full expert interface reads as authoritative and leaves the user unable to act.
In a financial tool, honesty is a design feature: labeling figures as estimates with a range and prompting a professional check on high-stakes numbers builds more trust than a confident figure the user cannot verify.
For freelancers and owners, the useful unit is not the entity but the person's actual position, so netting business obligations out of the personal picture answers the question they really have.