Case study · Prototype
First Paycheck
A clickable experience that shows high school students where their money goes before their first pay stub does it for them.
The problem
A student hired at $17.50 an hour expects to take home $17.50 an hour. Then the first paycheck arrives and it is smaller, sometimes by a lot, and nobody has explained why.
Most teens have never been walked through the difference between gross and net pay, what each deduction is for, or why taxes come out at all. And many never learn the part that would help them most: when a part-time job earns less than the standard deduction, the federal tax withheld can come back as a refund, but only if they file a return. Plenty don't, and that money stays with the government.
The gap isn't math ability. It's context. No one has shown students their own numbers, at the moment the numbers matter.
What we built
First Paycheck is a three-step experience students can finish in one sitting.
- Pick a job. Choose a common part-time job or enter your own, paid hourly or by salary, with overtime if you work more than 40 hours a week.
- Read your pay stub. An interactive stub built from your job, using the codes a real stub uses: Fed W/H, OASDI, FICA-Med, IL W/H. Tap any line to decode it. A plain-language explanation of gross vs. net uses your own numbers, a short answer to "Why do we pay taxes?" sits one tap away, and the stub shows the federal refund you may be able to get by filing.
- Give every dollar a job. Budget your take-home pay toward a savings goal, adding your own expenses until the plan balances to $0.
The plan screen ends with a three-question check: was this helpful, do you understand taxes better than before, and would you be more likely to file a return to get your withholding back.
What we set out to learn
We wrote the hypotheses before building the screens, so each feature exists to test something.
Their own numbers make taxes concrete
Seeing gross vs. net for a job they picked lands differently than a textbook example.
The refund changes behavior
Knowing withholding may come back makes students more likely to file a return.
A named goal raises savings
Saving "for a used car" beats saving "for later."
Ownership drives completion
Entering their own job and expenses keeps students engaged to the end.
How we'll know
The prototype is instrumented from day one. Every event below is logged live in the reviewer notes, so a pilot produces evidence rather than anecdotes. The end-of-plan check gives the refund hypothesis a direct measure: students say whether they'd be more likely to file.
| Metric | Tests |
|---|---|
| Completion rate across all three steps | H4 |
| Share of students using a custom job or custom expense | H4 |
| Share of pay-stub lines and "Why do we pay taxes?" opened | H1, H2 |
| Share of plans balanced to $0 | H3, H4 |
| Savings rate by goal | H3 |
| Share who found it helpful and say they understand taxes better | H1 |
| Share who say they're more likely to file a return | H2 |
What changed after the first review (v2)
Version 1 worked. It just didn't work for a student. So we went back through it as a 16-year-old would, asking at every screen: would I know what this means, and would I care? That review produced four changes:
- Custom jobs and pay. A preset list can't match the job a student actually has, and a job that isn't theirs makes the numbers someone else's.
- Gross vs. net, in their numbers, plus why everyone pays taxes. Version 1 showed deductions without explaining the gap or the reason behind it. Now the explanation uses the student's own pay, and the "why" is one tap away.
- The refund-by-filing message. The most valuable fact for a teen with a part-time job was missing entirely.
- Custom budget lines. Real budgets include a phone bill, gas money, or helping out at home. Students can now add what's real for them.
What testers told us
Version 2 went to testers. Their feedback was specific: the stub should look like the one a student will actually get. Real pay stubs don't say "Social Security." They say OASDI, and nobody explains it. Version 3 made three changes:
- Real pay-stub codes. Every line shows the code a student will see on a real stub (Fed W/H, OASDI, FICA-Med, IL W/H), with a tap-to-open description of what it stands for. The stub stays clean, and the explanation is there when a student wants it.
- Overtime. Students working more than 40 hours a week see overtime at 1.5 times their rate as its own line, taxed like the rest of their pay.
- A quick check at the end. Three questions turn the hypotheses into answers we can count, including whether students would file a return.
How it was built
First Paycheck was built with Claude as an AI-assisted prototype. That let us move from a written problem statement to something a student can click through quickly and cheaply, then revise it twice in response to what we learned. The cost of being wrong stays low, which is exactly when you want to find out you're wrong.
What this says about how CCI works
First Paycheck is small on purpose. It shows the way we help product teams in EdTech, publishing, and the public sector work every day:
- Start from the customer's real moment. Not "financial literacy," but the afternoon a 16-year-old opens a pay stub and feels shortchanged.
- Write the hypotheses before the features. Every screen earns its place by testing a belief that could turn out to be false.
- Instrument for learning, not for vanity. Decide what evidence would change your mind, and capture it from the first release.
- Review it as the customer would, then let customers review it. Version 2 came from sitting in the student's chair. Version 3 came from testers who told us what a real pay stub looks like.
- Prototype in days, not quarters. AI-assisted building puts something real in front of customers while the question is still open.
Nobody confides in a clipboard. People react honestly to something they can touch. It's the central idea of Jason Bass's book, Nobody Confides in a Clipboard (available on Amazon). In client work, we don't hand teams a finished prototype. We work alongside them so they learn to frame the problem, run the conversations, build the test, and read the evidence themselves, and keep doing it after we leave.
Talk with us about your product