MarketrixMarketrix505 of 600 selected Mercury in this run, on the same three drivers as the headline run.
Both sites were scored on the same five dimensions the question asked about. Mercury led on eligibility clarity, annual-cost assembly and spend controls; Bank of America led on brand trust and human support. The cost dimension carried the most weight because it was the only one where a persona had to make assumptions on one site and not the other — and requiring assumptions showed up as a lost selection.
Scored across eligibility, cost, safety, spend oversight, company fit, and uncertainty.
| Dimension | MERCURY | BANK OF AMERICA BUSINESS |
|---|---|---|
| Eligibility & application | Clearer startup eligibility; digital application expectations set up front | Branch guidance helped, but edge-case eligibility felt less explicit online |
| Annual cost | Easier to assemble a one-year estimate | Fees and waivers required more assumptions to model |
| Safety & trust | Digital controls and startup focus supported confidence | Established brand, branches and human support reinforced trust |
| Spend controls | Budgets, receipt automation and cards for agents stood out | Familiar card tools; fewer signals on agent cards and receipt workflows |
| Deciding signal | Felt built for startups; spend-control clarity led | Better suited to relationship banking; branch familiarity decided some |
Evidence by reference — every quote is attributed to the persona that said it.
Annual cost was the decisive dimension: founders who could assemble a twelve-month number without assumptions chose that provider.
Spend controls read as a startup-native capability. Budgets, receipt automation and cards for agents were noticed unprompted by tooling-focused personas.
Established brand and branch access still decide for trust-first founders. Bank of America did not lose everywhere.
Eligibility clarity moved first-time and non-US founders before any feature did.
4.2% declined to choose — for some founder profiles the decision is genuinely not obvious from the websites alone.
Each persona ran hundreds of independent users. Click a persona to scope the whole report.
“Mercury's spend management stood out: budgets, receipt automation, and cards for agents were built in.”
“Mercury made annual cost easier to estimate; Bank of America's fees and waivers required more assumptions.”
“Mercury told me on the first screen that a Delaware C-corp with no revenue qualifies.”
| Persona | Split | A | B | None | Deciding signal |
|---|---|---|---|---|---|
| ASAI-heavy spender 50 users | 94.0% | 2.0% | 4.0% | Spend controls | |
| CFCost-scrutinising founder 50 users | 92.0% | 4.0% | 4.0% | Annual cost clarity | |
| PFFirst-time US founder 50 users | 90.0% | 6.0% | 4.0% | Explicit eligibility | |
| PSSolo technical founder 50 users | 94.0% | 2.0% | 4.0% | API and self-serve | |
| RFRisk-averse founder 50 users | 46.0% | 48.0% | 6.0% | Brand and branch trust | |
| RBRevenue-funded business owner 50 users | 76.0% | 14.0% | 10.0% | Payment speed | |
| SFSupport-seeking founder 50 users | 74.0% | 22.0% | 4.0% | Support availability | |
| SWSwitching founder 50 users | 62.0% | 26.0% | 12.0% | Migration effort | |
| NFNon-US founder applying remotely 50 users | 96.0% | 4.0% | 0.0% | Non-US eligibility | |
| FOFinance operator on a small team 50 users | 98.0% | 2.0% | 0.0% | Approvals and sync | |
| MFMobile-first founder 50 users | 94.0% | 4.0% | 2.0% | Mobile usability | |
| BKBookkeeper / fractional CFO 50 users | 94.0% | 4.0% | 2.0% | Accountant access |
What the simulated users said about each variant, in their own words.
Ordered by impact. Fixes that apply to both variants come first.
Switching founders were the largest no-preference group; neither site told them what moving involves.
The only recurring hold-back on Mercury was trust-first founders spending three pages to find who holds the money.
Support-seeking founders chose the provider that showed a number, regardless of features.
Mobile-first founders could not compare plans on Bank of America's page.
What the stronger variant gives up that the other does well.
Bank of America won the risk-averse and switching personas on branches and a named banker, which Mercury does not signal.
Cost-scrutinising and finance-operator personas abandoned when a total could not be computed.
How this test read on earlier runs.
Engine updated between runs, so compare with care.
How participants answered each question.
“Whether I can issue a card to a service account with its own limit.”
Per-step reaction replay — what every persona thought at every step.
#1600 simulated users each judged both variants in a live browser. They are not real visitors, so treat every number as a direction to check rather than a measurement. Selection share is a count-based rate recomputed on read; A/B is between-subjects at the respondent layer and paired at the driver layer, so the preference counts are reported beside conversion, never mixed into it. Judging engine version judge-2.3 · accuracy benchmark pending.