Reference Architecture
Architecture pattern — built and tested. Not currently in production with paying customers. See wintura.ai for Soatech's flagship production case study.
PropTech Platform: The Iteration Sprint Scaling Pattern
Architecture pattern — built as a reference implementation, not in production with paying users. The featured production case study is wintura.ai (see /case-studies/wintura-ai). This pattern documents the PropTech Iteration Sprint playbook a Soatech engagement would follow for that domain.
Architect-led, AI-accelerated. 2× faster than hand-coded shops.
Hand-coded teams, hourly billing, scope creep. Multi-month ramps before the first production-grade PR.
A Veteran Architect leads the Pod. AI tooling captured as reviewed throughput, not someone else's margin. Fixed sprint price.
Key Results
The Challenge
This reference implementation documents the Iteration Sprint pattern for a property management SaaS that has outgrown its prototype. The scenario: a London-based platform with 500+ landlord customers and 8,000+ rental units, where the existing codebase ships features too slowly to keep up with demand.
- Tenant portal is the #1 requested feature — tenants need to pay rent, submit maintenance requests, and view documents online
- Maintenance workflows are manual — property managers coordinate repairs via email and phone
- Infrastructure costs are ballooning — the AWS bill has doubled in 6 months without corresponding traffic growth
- Technical debt from early-stage shortcuts slows every new feature
The constraint is delivery throughput, not headcount. The Iteration Sprint model answers it differently than staff augmentation: instead of adding bodies, a senior architect owns the roadmap end-to-end and ships scoped feature modules every two weeks at a fixed price per sprint.
The Approach
Sprint cadence: fixed-price feature modules
Under the Iteration Sprint model, your architect delivers one to three feature modules per two-week sprint. Each module is scope-capped (≤5 screens, ≤3 entities, ≤2 integrations) so the price is fixed and the timeline is predictable. There is no day-rate, no hourly billing, and no open-ended "team" you have to manage.
The working rhythm for this pattern:
- Shared access to Slack, Linear, and GitHub so reviews and demos happen in the open
- A bi-weekly sprint demo plus async updates between demos
- 11:00 CET overlap with London (1 hour ahead) so feedback loops stay tight
- Every PR reviewed before merge — architectural coherence is owned by one person, not diffused across a rotating roster
Sprints 1–4: Tenant Portal
The first feature module shipped was the tenant portal:
- Rent payments — Stripe integration with automated reminders and receipt generation
- Maintenance requests — Photo upload, categorization, and real-time status tracking
- Document vault — Tenancy agreements, inspection reports, and correspondence in one place
- Mobile-responsive — 65% of tenants access the portal on mobile
The portal pattern is designed to reach activation quickly: in this reference scenario, 72% of tenants activated their accounts within the first month.
Sprints 5–8: Maintenance Automation
The highest-impact module replaces a chaotic email-based workflow with an automated system:
- Smart routing — Maintenance requests automatically categorized and assigned to the right contractor based on type, location, and availability
- Contractor portal — External tradespeople accept jobs, update status, and upload completion photos
- SLA tracking — Automatic escalation if repairs aren't acknowledged within 4 hours or completed within the agreed timeframe
- Cost tracking — Every maintenance job linked to the property with full cost history
In this reference scenario, property managers save 3–4 hours per day on maintenance coordination.
Ongoing: Infrastructure Optimization
Infrastructure work runs as an Iteration Sprint Plus module (feature modules plus ongoing CI/CD, observability, and performance tuning):
- Right-sized EC2 instances — Most were over-provisioned by 2–3x
- Reserved instances — 40% savings on compute costs
- Caching layer — Redis for frequently-accessed property and tenant data
- Database optimization — Query analysis and indexing reduced average response time from 800ms to 120ms
- CI/CD pipeline — Automated testing and deployment replaced manual processes
In this reference scenario, the monthly AWS bill drops from £8,200 to £4,900.
The Results
After 6 months of bi-weekly sprints, the target outcomes for this pattern:
- Feature releases: From 1–2 per month to 5–6 per month
- Tenant portal adoption: 72% of tenants active within 30 days of launch
- Maintenance efficiency: 15+ hours per week saved on coordination
- Infrastructure: 40% cost reduction, 85% faster average response time
The same architect carries context across every sprint, so the codebase stays coherent as it grows — and the engagement converts naturally into an ongoing Iteration Sprint or Production Subscription once the roadmap stabilizes.
Why It Works
-
Timezone alignment is underrated. Soatech operates CET — just 1 hour ahead of London. Demos, reviews, and Slack conversations happen naturally, not at awkward hours.
-
One architect, continuous context. A single senior architect owns the roadmap across every sprint. Architectural decisions compound instead of fragmenting across a rotating set of contractors.
-
Fixed-price discipline. Each feature module is scope-capped and priced before work starts. There is no hourly meter and no scope creep — out-of-scope requests roll into the next sprint, never silently into the current one.