← Back

Case Study: Killing Vantage at $10 a Month

Vantage was a consumer fintech product my co-founder and I built through the UVA Darden i.Lab accelerator in 2026. The idea: connect your bank account with Plaid, and an AI autopilot finds the loyalty programs and rewards you qualify for but are not using, then enrolls you.

The whole business rested on one number. How many dollars a month is a typical Gen Z user leaving on the table? We set a target of $25. Below that, nobody pays, and nobody stays.

So instead of building more product, we built a way to measure that number honestly.

The research stack

• 40 customer interviews, distilled into 14 structured personas with spending behavior, enrolled programs, and pain points
• A database of 122 loyalty programs and 26 credit cards, each with an earn rate and a value per $100 spent

• A Python savings calculator that runs every persona against every program they shop at but are not enrolled in

• A written attribution model defining exactly what counts as a missed dollar, so the answer could not drift toward what we wanted to hear

Confidence tiers mattered more than the model

Most of our spend data came from interviews ("I eat out once or twice a week"), not transactions. So every estimate got a confidence tier. Transaction screenshots were high confidence. Behavioral anchors from interviews were low. Only high and medium tiers counted toward the headline.

We also audited the database itself. The program rates had been aggregated from several AI research tools, so I verified about 20 high-impact programs against official terms and T&Cs. One major retailer's rewards rate was inflated 2.5x in our data. Rebrands, discontinued tiers, and duplicate programs all had to be resolved before the simulation meant anything.

The answer

Across all tiers, including estimates, the typical user was missing about $9 to $11 a month. Counting only hard data, the floor was closer to a few dollars. Either way, it was well under $25.

The interviews pointed the same direction. The dominant money emotions were avoidance and guilt, willingness to pay was low, and people rejected dashboard-style tools.

The decision

I recommended discontinuing the consumer product. Not because the build failed, but because the value equation failed. A product that saves someone $10 a month has to cost almost nothing and require almost no attention, and that is a very hard business to build.

What I took from it

• Find the one number the business depends on, and measure it before scaling the build
• Make the methodology strict enough that it can tell you no

• Killing a product on evidence is a product decision, not a failure

• The company, accelerator relationships, and research habits carried straight into Wellhound