Improving Cash Advance Decision-Making Through Personalized Insights

Users can quickly select a cash advance amount, but the current experience does not provide enough context to help them evaluate affordability. Without visibility into their post-repayment balance or upcoming expenses, users may commit to an amount they cannot comfortably repay, creating avoidable repayment failures and trust issues.

Right now the flow opens by asking how much you want to draw, and it leads with the biggest number you qualify for. That quietly turns approval into a target to hit instead of a decision to think about. If you're choosing between $150 and $250, there's no easy way to see how each one lands against the rent that's due, the paycheck that's coming, or what's actually left in your account after you pay it back, instant-delivery fee included.

And when you're stressed about money, "what you can get" and "what you can safely repay" are two very different things.

Tilt already has everything it needs to help increase transparency, it just isn't showing it at the moment of decision. To approve an advance, Tilt already analyzes income patterns, recurring expenses, and pay cycles. Surfacing that same data back to the user in plain language turns an opaque number into an informed, confident choice.

So the work isn't about talking anyone out of borrowing, or adding friction. It's about taking data Tilt already runs and handing it back in a way that helps someone make a choice they feel good about. My bet is that when people understand what a choice does to their next few weeks, they repay more reliably and they come back. That's good for them and good for the business.

Everything in this project comes back to one belief I can actually put to the test:

Transparency and guidance are what I'm adding; better decisions, repayment, and trust are what I'm hoping falls out of it. Everything after this, the flow, the experiment, the metrics, exists to get that belief in front of real people and let what they do tell me whether I'm right.

I didn't want to redesign Tilt. Cash Confidence is five screens that live inside the flow people already use and just make each decision a little clearer. It opens where Tilt opens today, on picking an amount, and from there it adds the recommendation, the real instant-delivery fee, and a plain repayment preview. The point wasn't to add screens. It was to take some of the guesswork out.

One thing I didn't want to hide: the fee. Tilt charges an instant-delivery fee ($1–$8 for advances under $300, or 3% once you're at $300+, with standard delivery free). It's a real cost that shows up in "total to repay," so the prototype assumes someone picks instant delivery and shows the fee in every breakdown. Affordability isn't honest if you save the fee for the last step.

A few principles kept me honest while building it, all of them coming back to the fact that people usually reach for a cash advance when money is already tight:

There's also an optional AI layer sitting on top: one short, human sentence, something like "Based on your recent income and expenses, this amount covers what you need and still leaves room for your bills." It helps with the decision. It never makes it for you.

A hypothesis is just a guess until the data backs it up. So rather than ship this to everyone, I'd run Cash Confidence as an A/B test against the current flow and watch repayment, the number I care about most. The variant only changes how the decision is framed, not who qualifies or how much they can get. That way, if repayment moves, I can trace it to the experience and not to something else.

Eligibility, underwriting, and repayment policy stay the same for both groups. The only thing that changes is how much visibility and guidance someone gets while they're deciding.

One number decides whether it worked. The rest are there to explain why it moved, and to catch a version that looks like a win on the surface but quietly costs the business.

Then the guardrails, the things that keep a "trust" feature from quietly costing the business. If clearer guidance just means people borrow less than they actually need, or revenue slips, that isn't a win:

Every one of these choices had a real tension behind it. Here's what I decided, and why.

A few things I deliberately left out of scope: changes to eligibility, underwriting, repayment policy, credit scoring, or any new banking products. The whole point was to show that a real improvement to trust and repayment could come from how you present data Tilt already has, not from rebuilding the risk stack underneath it.