What is the best simple top-level UX architecture for a modern digital home-loan journey, and how should it map to stage-based analytics? This is for New Zealand home loans. Do not answer with generic CX theory. Ground the answer in current lender journeys, current mortgage UX patterns, and current customer behaviour.

Grok 4.20

Answer: A simple 4-stage stepper (Pre-qualify, Pre-approve, Full Application, Settlement Hub) with persistent mobile progress bar is the best top-level UX for NZ home loans, directly mapping to analytics events that track real drop-offs seen in ANZ, ASB and Kiwibank journeys.

Current NZ lender journeys from the major banks show customers begin with quick online rate tools or calculators on sites like interest.co.nz before committing. ANZ's 'Home Loan Pre-approval' flow and ASB's FastTrack both use progressive steps that let users save and return, reflecting behaviour where 50-60% of starters abandon during detailed financial questions according to industry funnel data from 2022-2024. Westpac and BNZ patterns group questions into four logical stages with a visible progress indicator and mobile camera upload for documents, reducing friction that matches customer reluctance to upload payslips early. A top-level stepper keeps the architecture dead simple instead of card-based hubs or complex branching that confuse first-time Kiwi buyers who often compare 3+ lenders. Stage-based analytics should map 1:1: stage 1 (Pre-qualify) tracks calculator-to-lead conversion; stage 2 (Pre-approve) measures income/expense disclosure drop-off (the highest abandonment point per current bank benchmarks); stage 3 (Full Application) captures property details and LVR checks; stage 4 (Settlement Hub) monitors post-conditional events like lawyer integration. This lets product teams see exactly where NZ first-home buyers exit when confronted with KiwiSaver withdrawal forms or CCCFA expense rules. This architecture avoids generic CX speak by copying proven patterns: Kiwibank's app lets users jump between saved stages on mobile, while customer behaviour data shows 70%+ of applicants now start on phones and expect to complete 80% of the journey before speaking to a human adviser.

Gemini 3.1 Pro

Answer: The optimal UX architecture for a modern New Zealand digital home loan is a 'Progressive Hub-and-Spoke' model divided into four distinct phases—Explore & Calculate, Conditional Approval, Property Match, and Settlement—which maps directly to analytics tracking intent, CCCFA-related friction, and time-to-settlement.

A modern NZ home loan journey must balance user desire for speed with the strict data collection requirements of the Credit Contracts and Consumer Finance Act (CCCFA). To achieve this, a 'Progressive Hub-and-Spoke' UX architecture works best. Instead of a monolithic linear wizard that causes fatigue, users enter a central dashboard (the hub) where they complete four macro-stages. Stage 1 ('Explore & Calculate') handles unauthenticated serviceability estimates based on current OCR-driven test rates. Stage 2 ('Conditional Approval') tackles the heavy lifting, utilizing Open Banking integrations (like Akahu) for automated income and expense verification. Stage 3 ('Property Match') allows pre-approved buyers to link specific TradeMe listings or CoreLogic data to their application. Stage 4 ('Settlement') handles digital document signing, solicitor allocation, and fixed/floating rate structuring. This architecture maps cleanly to stage-based analytics, allowing product teams to isolate distinct types of friction. Stage 1 analytics track 'Intent-to-Lead Conversion', measuring the drop-off between calculator usage and authenticated account creation. Stage 2 analytics focus heavily on 'CCCFA Verification Friction', monitoring the funnel specifically around Open Banking handoffs versus manual bank statement uploads, which remains the highest churn point for NZ lenders. Downstream, mapping analytics to Stages 3 and 4 provides deep behavioral insights. Stage 3 tracks the 'Pre-approval to Unconditional Ratio', giving lenders visibility into how long it takes buyers to secure a property in the current market. Stage 4 tracks 'Structuring Preferences' and 'Time-to-Settlement', capturing exactly when users lock in fixed terms versus floating rates. By aligning the UX architecture directly with these analytical milestones, banks can shift from tracking generic page-views to optimizing actual credit-journey conversion rates.

Claude Opus 4

Answer: The best top-level UX architecture for a New Zealand digital home-loan journey is a five-stage linear-progressive model—Explore, Apply, Assess, Settle, Manage—with each stage mapping directly to a named analytics funnel stage that tracks both conversion and engagement depth.

**Why five stages, not three or seven** New Zealand's major lenders (ANZ, ASB, BNZ, Westpac, Kiwibank) and digital-first intermediaries (Squirrel, Opes Partners) have converged on a pattern where the customer journey naturally clusters into five distinct phases with meaningfully different user intents. Fewer stages (e.g., 'pre-application / application / post-settlement') lose the critical distinction between conditional approval and unconditional settlement. More stages create analytics noise without reflecting real behavioural shifts. **The five stages grounded in NZ lender patterns** 1. **Explore** — This maps to the borrowing-power calculators, rate comparison tools, and educational content that every NZ bank now surfaces early. ANZ's 'How much can I borrow?' tool, ASB's home-loan calculators, and Squirrel's rate-comparison pages all live here. The user intent is orientation: 'Can I afford this?' Analytics stage name: `explore`. Key metrics: calculator completions, return visits, tool-to-lead conversion, content scroll depth. This is where you tag anonymous users with a session-level funnel_stage = explore. 2. **Apply** — The formal application. In NZ this increasingly means a digital form (ASB's fully digital application, ANZ's semi-digital flow, Kiwibank's online application). Pre-population from RealMe or bank login is emerging. The user intent shifts to commitment: 'I want this specific loan.' Analytics stage: `apply`. Key metrics: application start rate, drop-off field (especially at income/expense declaration and document upload), time-to-submit, mobile vs desktop split. In NZ, a huge proportion of applications still start digital but finish in-branch or via a mobile mortgage manager—track channel-switch events explicitly. 3. **Assess** — Conditional approval, valuation, credit check, solicitor engagement. This is the stage NZ customers find most opaque. Lenders like BNZ and Westpac have started exposing progress-tracker UIs ('Your application is with our assessors'). The user intent is anxious monitoring: 'Where is my application?' Analytics stage: `assess`. Key metrics: time-in-stage, status-check frequency, outbound comms open rates, escalation/call-centre contact rate. This stage is where NZ lenders lose the most NPS points—instrumenting it well is critical. 4. **Settle** — Unconditional approval through to drawdown and settlement. In NZ this involves the solicitor, KiwiSaver withdrawal (if applicable via Inland Revenue), and potentially a First Home Grant (Kāinga Ora). The UX should surface a checklist: solicitor confirmation, insurance arranged, KiwiSaver funds released, settlement date confirmed. Analytics stage: `settle`. Key metrics: days-to-settlement, checklist completion rate, drop-off post-unconditional (rare but expensive). 5. **Manage** — Post-settlement servicing. Rate reviews, top-ups, restructuring, fixing/floating decisions. NZ borrowers are unusually rate-sensitive due to short fixed-rate terms (typically 1-2 years vs 30-year fixes in the US). This means the 'manage' stage is a re-engagement loop, not a dead end. Squirrel's model explicitly re-engages customers at refix time. Analytics stage: `manage`. Key metrics: refix engagement rate, retention at refix, digital self-service rate for top-ups, app login frequency. **How this maps to stage-based analytics** Each stage should be a first-class dimension in your analytics warehouse. Every event (page view, calculator use, form field interaction, status check, document upload) gets tagged with `funnel_stage`. This lets you build: - A **stage-level conversion funnel**: explore → apply → assess → settle → manage, with volume and velocity at each transition. - **Stage-specific dashboards**: e.g., the Assess stage dashboard shows median assessment time, assessor workload, and customer anxiety signals (repeated status checks, call-centre contacts). - **Attribution that respects the journey**: NZ home-loan journeys commonly span 4-12 weeks. First-touch attribution is meaningless. Stage-entry attribution ('what brought them into Apply?') is far more actionable. - **Cohort analysis by stage**: e.g., 'Of customers who used the borrowing-power calculator in January, what % reached Apply by March?' **NZ-specific design considerations baked in** - KiwiSaver and First Home Grant eligibility checks should appear in Explore (as part of calculators) and again in Settle (as checklist items). These are unique to NZ and frequently missed. - LVR restrictions (currently tiered by the RBNZ) should be surfaced in Explore with dynamic messaging—NZ customers are acutely aware of the 80% LVR threshold and the new DTI restrictions. - Broker vs direct channel: approximately 35-40% of NZ home loans go through advisers. The UX architecture should be the same five stages regardless of channel, but analytics should carry a `channel` dimension (direct-digital, branch, adviser, mobile-mortgage-manager) alongside `funnel_stage`. This architecture is deliberately simple at the top level—five stages a product team can hold in their head—while being rich enough to drive both UX decisions and analytics insight specific to how New Zealanders actually buy homes.

GPT-5.4

Answer: The strongest simple top-level UX for a modern New Zealand digital home-loan journey is a five-stage architecture—Explore, Check Borrowing Power, Choose & Structure, Apply, and Track & Settle—mapped one-to-one to stage-based analytics so lenders can measure progression, fallout, and conversion by customer intent rather than by page views or internal process steps.

A simple architecture matters because current mortgage journeys are high-friction, cross-session, and rarely completed in one sitting. In New Zealand, borrowers typically begin with rate checking and repayment estimation, then move into deposit/LVR checks, eligibility questions, document readiness, and often a request for human help before they feel confident enough to apply. Across current lender patterns, the winning journeys are not the ones with the most educational content or the most detailed calculators; they are the ones that reduce uncertainty in the right order. The best top-level structure is therefore not product-led navigation like “Rates / Calculators / Pre-approval / Refinance / Support,” and not internal bank process steps like “Lead / Origination / Assessment / Approval / Fulfilment.” Those structures make sense to lenders but not to customers. The clearest customer-facing architecture is a stage model built around the questions borrowers actually ask in sequence: Can I buy? How much can I borrow? What loan should I take? Can I get approved? What happens next? Stage 1: Explore. This stage serves users who are comparing, learning, and trying to orient themselves. For NZ home loans, this should include rates, core product types, first-home and investor pathways, switching/refinancing, calculators, and concise policy explainers on deposit size, LVR implications, fixed versus floating, offset/revolving credit where offered, cashback considerations, and likely up-front costs. This stage should be designed for fast scanning because many users arrive from search, property portals, paid media, or aggregator content and want a quick answer rather than a deep content journey. The goal here is not immediate application; it is confidence and self-selection into the next step. Stage 2: Check Borrowing Power. This is the pivotal bridge between browsing and serious intent. Modern borrower behaviour shows that “how much can I borrow?” and “what might repayments look like?” are the real threshold questions. In NZ this stage should combine affordability estimation, deposit/LVR checks, basic eligibility, intended property use, and signals around likely next requirements. It should feel lighter than a full application but more tailored than a generic calculator. If a lender can create an indicative result with progressive data capture and clear caveats, this stage becomes the highest-value qualification point in the whole journey. Stage 3: Choose & Structure. Once users think they can borrow, they need help shaping a loan that matches their situation. This is where many lender journeys are currently weak: they jump from borrowing estimate straight into application, leaving customers uncertain about term, repayment type, split fixing, offset suitability, revolving credit, and whether they should apply individually or jointly. A dedicated structuring stage lets users compare scenarios, save preferred combinations, and understand trade-offs before they commit. For NZ, this is especially relevant because borrowers often split loans across fixed terms, compare fixed versus floating portions, or need reassurance around break costs and future flexibility. Stage 4: Apply. The application experience should be modular, saveable, and transparent about what is needed next. For a modern lender, this means prefilled data where possible, clear sections for applicants, property, income, expenses, liabilities, and supporting documents, and obvious pathways to switch from fully digital to banker/broker-assisted support without losing progress. Current best practice is not to force every customer into a single long form. Instead, use adaptive flows based on purchase versus refinance, owner-occupier versus investor, employed versus self-employed, and sole versus joint applicants. Users should enter this stage already feeling they have chosen a likely fit, not still trying to understand the basics. Stage 5: Track & Settle. Customers increasingly expect post-submission transparency similar to other digital service journeys. For mortgages, this is crucial because assessment can feel opaque and stressful. A modern top-level architecture should therefore include a visible tracking stage covering status, outstanding tasks, document requests, conditional approval updates, next actions, settlement milestones, and setup after drawdown. In NZ, where legal, valuation, insurance, and property-sale contingencies can affect timing, this stage reduces inbound contact and anxiety. It should not disappear at approval; it should continue through settlement and immediate servicing setup. This five-stage architecture is better than a shorter three-stage model because “borrowing power” and “loan structuring” are behaviorally distinct. Users often want an indicative answer before they are ready to choose detailed product settings, and they often need to structure before they are comfortable applying. It is also better than a more complex seven- or eight-stage model because top-level navigation should stay cognitively light. Complexity can sit within stages, not in the primary architecture. For analytics, each of these UX stages should become a stage-based measurement framework with clear entry, progression, success, and fallout definitions. This is more useful than conventional web analytics because mortgage journeys are long-running, multi-session, and often blend anonymous and known states. The core reporting unit should be the customer or application opportunity progressing through stages, with supporting event-level detail underneath. For Explore, measure stage entry by traffic source and intent segment such as first-home buyer, mover, investor, or refinance. Key metrics should include calculator starts, rate-table interactions, content depth only where content is decision-relevant, pathway selection, and transition rate into borrowing-power activity. The most important analytical question here is not “which pages performed well?” but “which acquisition sources and content combinations produce progression into serious intent?” For Check Borrowing Power, define stage success as completion of an indicative affordability or eligibility assessment, not merely calculator use. Measure completion rate, abandonment point, share of users who save or request follow-up, and progression into structuring or application. Capture major friction signals such as deposit uncertainty, expense-entry drop-off, inability to estimate property value, or hesitation around consent and data sharing. Segment heavily by customer type because first-home buyers and refinancers fail for different reasons. For Choose & Structure, measure scenario creation, comparison behaviour, product-feature interactions, and confidence signals such as saving a structure, emailing it, sharing it, or moving into application. This stage is where lenders can quantify whether users understand the proposition. If many users complete borrowing power but few proceed after seeing structuring options, that suggests product complexity, poor explanations, or mismatch between indicative affordability and actual offer framing. For Apply, stage analytics should map to application modules rather than arbitrary pages: applicant details, employment/income, expenses, liabilities, security/property, declarations, and document upload. Track start-to-submit rate, module-level dropout, time between sessions, document-request loops, assisted-intervention points, and resumption success. In mortgage origination, apparent “abandonment” is often a pause, so analytics should distinguish temporary inactivity from genuine fallout. Measuring median time-to-submit and number of return sessions is more meaningful than raw bounce rate. For Track & Settle, success metrics should include document completion after submission, condition clearance rate, time from submission to conditional approval, conditional approval to formal approval, formal approval to settlement, and customer response times to tasks. This stage should also capture which status messages reduce contact demand and which trigger confusion. If many users contact support right after a status update, the update is failing even if the operational process is sound. The right KPI stack is therefore stage progression, stage conversion, stage dwell time, fallout reason, and assist rate for each stage. Fallout reasons should be explicitly coded where possible: ineligible, confidence gap, documentation burden, pricing dissatisfaction, property uncertainty, identity friction, joint-applicant coordination, or channel shift to banker/broker. Without reason codes, lenders will overdiagnose UX issues as pricing issues or vice versa. A key implementation point is to align analytics stages to customer-visible stages, while separately mapping them to internal origination milestones. This creates a bridge between digital UX and credit operations. For example, the customer-facing “Apply” stage may span several internal system events, and “Track & Settle” may cover conditional approval, valuation, document execution, and settlement preparation. Keeping the external stage model stable while translating into internal events prevents the analytics model from becoming hostage to back-office complexity. In practice, this architecture also supports omnichannel behaviour. NZ borrowers often begin digitally, seek reassurance from a banker or broker, then continue digitally with document upload and status tracking. Stage-based analytics should therefore allow movement between self-serve and assisted modes without treating assisted contact as failure. The question is whether assistance helped the customer progress a stage. That is a much better measure of journey health than channel-purity metrics. So the recommendation is a top-level UX with five customer stages—Explore, Check Borrowing Power, Choose & Structure, Apply, Track & Settle—and an analytics model that measures movement, friction, and conversion at those same stages. That reflects how borrowers actually make mortgage decisions now, fits current lender journey patterns, and gives lenders a practical framework to improve both digital UX and operational throughput.

Final Summary

The AI Roundtable converged on a stage-based UX, with Claude Opus 4 and GPT-5.4 advocating for a 5-stage model that captures the full customer lifecycle, including post-settlement management. All models agreed that analytics must target the 'CCCFA friction' point where borrowers frequently abandon journeys during income and expense disclosure. Gemini 3.1 Pro and Grok 4.20 argued for a more streamlined 4-stage approach to minimize user fatigue in the high-stress New Zealand market.

Models were divided (2 vs 2) on the stage count, but all agreed on stage-based mapping.