Generative Research

Which features would people actually pay for

Product had twelve candidate features for a proposed paid tier. Rather than one large study, three sequential rounds at increasing scale turned a features wishlist into a shipped revenue stream.

ClientA US consumer finance marketplace
RoleLead Researcher
MethodsSentiment survey, open response analysis, weighted feature ranking
Scale1,527 participants across three study iterations
1,527Total participants across three rounds
3Sequential study iterations
12Candidate features ranked
4Features that shipped

The question

Leadership needed to know which of twelve candidate features people would actually pay for, which three to five mattered most, and who those willing payers were.

Approach

I ran three sequential iterations rather than one large study, adding a variable each round.

Round one established a baseline feature ranking with 66 participants. Round two added a payment prompt, reframing the question from what is useful to what would you pay for, with 904 participants. Round three added open response questions about financial stressors and attempted solutions, with 557 participants.

Rankings were assigned point values and summed to produce a preference-weighted order. Recruitment ran through a participant panel, surveys through Qualtrics.

The finding that mattered most

The feature list assumed the wrong problem

Advanced budgeting led by a wide margin, roughly 50% above the next feature. Users wanted category-based recommendations and guidance on which bills to pay first. Financial coaching, automated transfers to savings, and connected brokerage accounts followed in a tight cluster.

The open responses reframed the whole problem. The stressors people named were not the ones the feature list assumed. Debt dominated: student loans, credit cards, medical bills. Cash flow going entirely to bills with nothing left for saving. Unexpected expenses with no buffer.

People were already solving these problems on their own, with spreadsheets, budgeting books, and competitor apps. One competitor app was named 22 times unprompted. That told us we were not entering an empty space.

What the ranking showed

Points summed from three rounds of weighted ranking, top five of twelve candidate features.

Advanced budgeting3,500 pts
Cashflow coach2,350 pts
Automated money transfer2,350 pts
Investment coach2,150 pts
Connected accounts1,950 pts

Advanced budgeting was not close. It led the field by roughly 50% over the next closest feature, and the margin held across all three rounds regardless of how the question was framed.

The demographic split held signal too. Younger participants centered on loan and credit application confusion and lack of visibility into their finances. Older participants centered on retirement planning and not knowing which information sources to trust.

What happened next

Findings fed directly into evaluative research

Wireframes tested the top-ranked concepts against the same population, which produced a series of mockups and three prototypes.

Top features shipped as a new paid tier

The highest ranked features were structured into an information architecture and shipped as a new tier of the company's flagship app.

The work opened a new revenue stream

The tier remains a component of the product today and became a new revenue stream for the company.

Iteration beat a single large study

Adding one variable per round, rather than testing everything at once, is what surfaced the payment prompt and stressor findings that changed the ranking's meaning.