Conversion-led journeys
Speed-optimised, branching, FCA-promotion-compliant journeys with Consumer-Duty-aligned disclosures.
10 — Lead generation
Lead-generation engineering for fintech — performance journeys, lender APIs, attribution.
Most credit-broker funnels were built for last decade’s tracking. Apple ITP, Consent Mode v2 and server-side requirements have quietly broken attribution at most lead-gen shops — but the marketing team is still reporting CPA off the old data. The numbers look fine until you compare them to the lender’s actual decline rates.
FCA · UK CFPB · US Quebec Law 25 · Canada
01 — Problem
Where lead generation builds usually go wrong.
Lead-generation platforms have three problems most engagements eventually surface. The first is conversion: the funnel was optimised for a 2018 browser model and Apple ITP / Consent Mode v2 have shifted the constraints. The second is lender economics: pricing engines that don’t reflect current lender appetite produce expensive declined leads. The third is attribution: the marketing team is making spend decisions off increasingly unreliable data.
We rebuild for the current environment. Server-side tracking that survives Consent Mode v2. Real-time pricing engines that match against actual lender appetite. A/B testing infrastructure that’s regulator-aware so journey changes don’t accidentally cross financial-promotions rules.
Plus the regulatory layer — because lead-gen platforms operate in an FCA financial-promotions perimeter that the next CP26/15 changes will tighten further, and a CCPA / Quebec Law 25 perimeter for cross-border traffic.
02 — What we build
End-to-end, audit-ready, in production.
Real-time pricing engines
Multi-lender, instant offer presentation with binding-offer caching and consistent rate-comparison surfaces.
Lender API integrations
High-volume, with circuit breakers and queueing — built for thousands of applications per hour.
A/B testing infrastructure
Server-side, decision-aware testing framework — regulator-safe for regulated journeys.
Attribution & analytics
Last-touch + multi-touch attribution, ITP-resilient — with reconciliation to lender-side outcomes.
Server-side tracking
GTM Server-Side, custom pipelines, BigQuery / Snowflake — Consent Mode v2 native.
Consent & cookie compliance
GDPR + CCPA + Quebec Law 25 — built once, applied per-region with the right disclosures.
SEO-grade landing pages
Core Web Vitals optimised, structured data, schema markup — built to actually rank.
03 — Patterns
Build patterns we ship from.
High-volume application funnel
Multi-step, branching journey with state recovery, abandonment-recovery and Consumer-Duty-aligned disclosures.
Next.jsInternal funnel orchestrationRedis
Multi-lender pricing engine
Real-time lender-API fan-out with offer normalisation and consistent customer-facing comparison surface.
AWS Step FunctionsRedisInternal pricing DSL
Server-side attribution pipeline
GTM Server-Side + custom event collection + BigQuery / Snowflake warehouse with lender-outcome reconciliation.
GTM ServerBigQuerydbt
A/B test infrastructure for regulated journeys
Server-side experimentation framework with financial-promotion guardrails and compliance-review hooks.
Internal experimentation serviceSnowflakeLooker
04 — Regulations
What this work has to comply with.
GlobalGDPR consentUKFCA financial promotionsUKConsumer Duty (marketing)UKASA advertising standardsUSCCPACAQuebec Law 25UKCP26/15 (incoming)
05 — Stack
What we build with.
Tracking & tagging
SegmentGTM Server-SideSnowplow
Attribution & analytics
BigQueryGA4AdjustAppsFlyer
Front-end
Next.jsInternal funnel orchestrationRedisCDN edge logic
06 — Recent build
A broker funnel re-platformed for 12k apps/week — 31% conversion lift in Q1.
A UK credit broker’s funnel was pre-Consent-Mode-v2 and pre-Apple-ITP-2. Attribution had quietly degraded to the point where the marketing team was overspending on the wrong channels. Lender pricing was a quarterly export to a static lookup. A/B testing was “ship a version, look at the data next week.”
We re-platformed the funnel over 14 weeks: server-side tracking on a GTM Server-Side + BigQuery foundation; a multi-lender pricing engine with sub-300ms fan-out and binding-offer caching; an A/B framework with financial-promotion guardrails; and a Consent-Mode-v2-native consent layer.
Conversion lifted 31% in Q1 — driven mostly by the journey-orchestration upgrades. Attribution accuracy improved enough that the marketing team reallocated ~20% of spend within the first two months. Lender-side decline rate dropped from 47% to 31% as the new pricing engine matched leads to lender appetite more accurately.
14 weeks
Replatform to production
31%
Conversion lift in Q1
47% → 31%
Lender decline rate
07 — FAQs
Questions we get every time.
How do you handle attribution under Apple’s ITP and Consent Mode v2?
Server-side first. The browser-side tracking is treated as one of several signal sources, not the source of truth. The source of truth is a server-collected event stream keyed to the customer’s journey, reconciled with the lender-side outcome data (decision, funded, default). That gives you attribution that doesn’t degrade with browser changes. Then you layer GTM Server-Side and Consent Mode v2 on top for the channel-level reporting Google’s ecosystem expects.
What’s the right server-side tracking pattern for FCA-regulated journeys?
Consent-state-aware event collection. Every event is keyed to the customer’s current consent state, so events from non-consenting customers are either dropped or anonymised at collection time (not after the fact). Plus a financial-promotion audit layer that records which promotional content the customer saw at each journey stage — because the FCA’s review will ask “what disclosures did the customer see at the point of application” and the answer needs to be query-able, not reconstructable-from-screenshots.
Can you A/B-test a regulated application form?
Yes, with guardrails. The experimentation framework needs to know which fields and which copy are subject to financial-promotion rules and prevent variants that cross the regulatory line. Compliance review is integrated into the experiment lifecycle — variants need approval before they go live, with the approval recorded as part of the audit trail. Once that infrastructure is in place, A/B testing in a regulated journey is no riskier than in an unregulated one.
How do you push lead quality back to marketing for optimisation?
Reconcile the lender-side outcome data (decline reason, funded amount, default after N months) back into the marketing attribution warehouse, keyed to the original journey. That gives the marketing team a quality signal beyond CPA — CPA-per-funded-application, CPA-per-quality-funded-application. Most broker funnels don’t do this, so the marketing team optimises for top-of-funnel volume rather than bottom-of-funnel quality. Adding it usually shifts spend allocation materially within a quarter.
What’s the GDPR-compliant pattern for lead-gen consent capture?
Granular consent options (not a single “accept all”), defaults-off for non-essential tracking, a visible consent state throughout the journey (so the customer knows what they consented to), and an easy revocation path. Plus per-jurisdiction logic — Quebec Law 25 has consent requirements GDPR doesn’t, CCPA has opt-out requirements rather than opt-in, and the journey needs to apply the right framework based on where the customer is.
— Adjacent practices
Most builds combine two or three.
Scoping a lead generation build?
Send us a few sentences. We’ll come back inside three business days with a credible delivery plan and a date.