The experiment that changed the brief
Common Cents Lab and Irrational Labs randomized 9,035 people. Some saw a single spendable number, some saw full category budgets, and the rest saw neither. Spending did not fall in either treatment group.
I read that as a design constraint. Ralphy has to do more than display a number: it has to watch the days between paychecks, show its math, and speak up when the week changes.
Randomized controlled trial · Common Cents Lab and Irrational Labs, with Clarity Money · 2020 · n = 9,035
Why I am the person working on it
For six years I have built machine learning models from bank transaction data for lenders and insurers. The recurring work is recognizing what a charge really is and which deposit is a paycheck, the same raw problem underneath Ralphy.
That experience does not make a forecast automatically right. Ralphy shows the receipt behind its number, keeps the bank connection read-only, and tells you when the data is stale so you can check what it knows.
See the idea work through a week.
Use sample money in the browser. No download and no bank connection.
