Independent product case study · Proposed experiment
A refinance landing-page experiment: show a repayment scenario before asking for contact details, then follow the journey through to a useful broker conversation.
A researched proposal and static prototype. No campaign, live test or measured uplift.
One comparison walkthrough reached ten question screens before a first-name request, without a personalised result. Lendi already offers an ungated repayment calculator elsewhere.
An earlier estimate may help suitable borrowers judge whether a conversation is worthwhile. The test brings existing value into the moment of uncertainty.
Angus Sewell’s demand-testing demonstration connects two offers, landing pages, matching outreach, captured interest and a later review. The useful move is making an idea specific enough for people to respond to.
For Lendi: would a useful answer earlier in the journey lead to better broker conversations?
This adaptation keeps the refinance proposition constant and compares two entry experiences. It follows the response beyond the first click: what visitors understand, whether they choose to continue, and whether a qualified conversation actually takes place.
The creator’s results screen uses sample data. It is inspiration for a learning process, with no verified conversion result to import into this proposal. Original Angus Sewell reel ↗
The first two steps below reflect the inspected comparison route. The proposed path and later broker stages describe the experience to test, rather than an observed customer outcome.
| Moment | Current entry · A | Proposed entry · B |
|---|---|---|
| Arrive | Rate-led refinance page, then a comparison action.Observed public entry | A clear question: what could a lower rate change for this loan?Proposed same refinance audience |
| Work it out | Ten question screens in the inspected route, then a first-name request.Walkthrough stopped here; later results not inspected | Balance, rate and remaining term produce a scenario, including switching costs.Estimate before identity capture |
| Choose | Keep the existing comparison and contact journey as control.Later behaviour needs baseline measurement | Change assumptions, leave with the estimate, or choose a broker review.A useful stopping point is allowed |
| Talk | Use the same consultation qualification rules in both arms.Proposed measurement | Pass the scenario with consent; broker checks the inputs and available options.Proposed handoff, customer inputs remain unverified |
| Learn | Compare qualified attended consultations per originally exposed visitor. Read feedback, lead quality and handling effort alongside the result.Later applications and settlements need their own mature outcome windows | |
A working treatment would test the combined calculator-first experience. Headline, tool placement and timing change together, so this experiment would not identify a single winning component.
Keep the current refinance page and comparison route as the control. Establish how many eligible visitors reach a qualified attended consultation today.
Show repayment differences and switching costs before an optional broker review. Preserve the remaining loan term and make the assumptions visible.
Hold steady: comparable acquisition sources, refinance audience, lender/compliance basis and broker qualification. Assign visitors persistently 50/50 and keep non-completers in the analysis.
Before exposure: agree the baseline, worthwhile effect, sample size, exclusions and stopping rules. A public prototype is ready to inspect; experiment plumbing is still proposed.
The example shows the repayment effect alongside upfront costs. A smaller monthly repayment can still leave someone barely ahead in the first year.
In a working version, a no-benefit scenario must remain a no-benefit result. A broker would then check actual products, eligibility, fees and features.
Open the example result ↗Illustrative repayment reduction
About $225 ahead in year one
after $2,000 in upfront switching costs.
$600,000 balance, 25 years remaining, 6.50% versus a hypothetical 6.00%. Monthly principal-and-interest repayments at constant rates. Cash-flow estimate; not a lender offer or an interest-saving calculation.
The existing Archify workflow connects the walkthrough, proposed analytics, A/B allocation and the shared outcome. Both branches return to the same decision. Personalisation follows only if the first test supports it.
Select any box to open its explanation beside it. Scroll across the map to follow both paths.
Explore full map ↗Walk the route, then check the baseline. Use existing analytics or proposed PostHog instrumentation to see paths and drop-off. The public walkthrough alone cannot explain why people leave.
Give the treatment a working, checked estimate and an honest optional next step. The current static mock demonstrates the intended experience; it does not submit information.
Use the same acquisition mix and persistently assign eligible visitors. If email is tested later, review recipients and matched drafts, record delivery and separate forwards. No outreach is sent by this prototype.
Join the assigned variant to lead, booking and attendance events. Collect only consented handoff context; keep loan amounts and contact details out of analytics payloads. Separate QA records and duplicates.
Ask what made proceeding worthwhile or unclear. Review qualified consultations over the same outcome window, alongside broker corrections and reasons for not continuing.
Stop, refine or extend the test from the evidence. More clicks without better conversations would not justify rollout. Visitor-selected goal tailoring is a later, separate hypothesis.
Within 14 days of exposure, using one agreed qualification rubric. Include everyone assigned, including visitors who stop before completing the tool. Establish the required sample before launch.
Credible improvement in the primary outcome, worthwhile incremental economics and guardrails within agreed limits. Expand cautiously, then follow applications and settlements.
Visitors use the estimate but misunderstand it, abandon the handoff or reach brokers with missing context. Use the feedback to choose a specific revision.
Guardrails deteriorate or the test shows no worthwhile benefit. If the sample is too small, record an inconclusive result rather than declare a winner.
Eligible exposures by variant; tool starts and result views; consented handoffs; booked and attended consultations; agreed qualification; handling effort; errors and complaints. Deduplicate visitors and consultations, and label missing or immature data.
Enrol for at least two complete weekly cycles and the required sample, with a proposed six-week maximum. Wait 14 days after the final enrolment for the primary outcome. A short checkpoint can check collection quality; it cannot substitute for mature outcomes.
Keep financial inputs and contact details out of analytics events. Test-data separation, calculation QA, privacy review, CRM joins and a read-only review must work before a real experiment starts. None are installed by this case study.
After a consented research interaction, ask: “What were you trying to work out?”, “Did this result change your next step?” and “What would you still need before speaking with a broker?” Capture reasons and corrections, rather than treating every positive reply as qualified demand.
A possible outreach draft would say: “We’re exploring a refinance calculator that shows an estimate before asking you to speak with a broker. Would you try the concept and tell us what is still unclear?” This is a draft for a separately approved research round, with no recipients selected or messages sent.
Built: the public-site review, this case study, a static customer concept and the interactive workflow. Proposed: a functioning calculator, persistent assignment, consented CRM handoff and outcome measurement. No results are presented as achieved.
The homepage, refinance page, repayment calculator and one synthetic comparison route were inspected on 26/10/2026. The walkthrough stopped at the first-name request. No identity, contact details, appointment or application was submitted.
Lendi already has an ungated repayment calculator. The opportunity proposed here is its timing within the refinance entry. Abandonment, motives, conversion performance and revenue impact remain unknown without customer research and Lendi’s own data.
Angus demonstrates a loop from two offers through pages, outreach and a later comparison. His visible figures are labelled sample data. This proposal adapts that loop to an existing refinance journey and judges the downstream consultation, rather than copying the demo’s signup figures.
There are no fabricated customer replies, waitlist counts, testimonials or performance claims here. Before a real visitor test, the calculator would need to work, its limits would need to be understood, and collection would need to be verified end to end.
Search Console and landing-page cohorts would help select the entry point and segment. Public page structure alone cannot establish search traffic or rankings. Keep acquisition comparable during this test, then explore visitor-selected refinance, first-home or investment goals separately if evidence supports it.
Angus Sewell · original demonstration ↗Lendi · refinance entry ↗Lendi · existing calculator ↗