Topic Breakdown
About Cobber App
Cobber App is an Australian peer‑to‑peer delivery platform dedicated to moving large or bulky second‑hand goods sustainably. Users simply open the Cobber mobile apps to request a vetted driver with transparent, instant pricing. Drivers with utes, vans or trailers earn flexible income by accepting jobs that fit their schedule.
Business Challenge
The integration bottleneck
Cobber planned to partner with online marketplaces, furniture recyclers and charity shops. Each partner wanted an embedded booking widget so buyers could arrange delivery without leaving the host site. Unfortunately, Cobber’s monolithic back end made external integrations risky and time‑consuming—averaging 3 – 4 weeks of bespoke coding per partner.
Unpredictable traffic & rigid data
Weekend garage‑sale campaigns generated huge, short‑lived spikes that the single VM struggled to absorb, hurting user experience. The existing SQL schema also resisted rapid changes to partner‑specific metadata, slowing iterations and growth.
Compliance & revenue share
Splitting card payments between Cobber, drivers and partners required compliant marketplace payouts—adding security scope Cobber’s small team wanted to avoid.
Solution Architecture
Lanex Australia designed a containerised, serverless architecture centred on AWS and open‑standard tooling:
| Layer | Technology | Reason |
|---|---|---|
| Edge | AWS API Gateway | Secure entry; throttling & usage plans per partner. |
| Compute | Docker containers on Amazon ECS with AWS Fargate | Auto‑scales by task; no servers to patch. |
| Micro‑services | Node.js Booking, Auth & Payment services | Each shipped as an independent image for rapid CI/CD. |
| Data | MongoDB Atlas global cluster | Flexible documents; auto‑scales storage & tier on load spikes. |
| Payments | Stripe Connect | PCI DSS out‑scoped; automated revenue‑share payouts. |
| Portal Widget | 40 kB JavaScript bundle in an <iframe>; parent‑iframe communication via postMessage with origin checks. | |
| CI/CD | GitHub Actions → Amazon ECR → ECS blue/green | Image promotion in < 10 min; zero‑downtime. |
| Observability | Amazon CloudWatch & AWS X‑Ray | p95 latency alerting; root‑cause tracing. |
| Security | IAM least‑privilege, Secrets Manager, WAF | Meets Cobber’s ISO‑27001 roadmap. |
Implementation Timeline
| Phase | Duration | Activities |
|---|---|---|
| Discovery & design | 1 week | Current‑state review, load‑profile modelling, widget UX wireframes. |
| MVP build | 4 weeks | Containerisation, Atlas cluster, Stripe Connect onboarding. |
| Pilot partner rollout | 2 weeks | iframe SDK docs, API rate‑limit tests, security penetration test. |
| Full launch | 1 week | DNS cut‑over, CloudWatch alarms, data migration to Atlas. |
Total elapsed time: 8 weeks from kick‑off to first production booking.
Outcomes & Success Metrics
| KPI | Before | After |
|---|---|---|
| Partner onboarding time | 3 – 4 weeks | < 1 day self‑serve |
| Peak traffic sustained | ≈ 200 RPS | 1 000 RPS @ < 200 ms p95 |
| Booking growth from partner sites | 0 % | 27 % of all bookings in 90 days |
| DevOps hours / month | 30 | 8 |
| Infra cost / delivery | AU $0.19 | AU $0.11 (42 % saving) |
Technical Deep‑Dive
1. Container scalability
Docker’s application‑level containers standardise dependencies, letting ECS place tasks anywhere in the cluster for horizontal scale. Fargate removes the need to manage EC2 hosts, so Cobber scales purely by CPU/RAM metrics or scheduled promos.
2. Document DB flexibility
With MongoDB Atlas, new JSON fields for partner campaigns require zero migrations—Atlas simply re‑indexes on demand. Auto‑scaling tiers absorb weekend peaks then down‑size to save cost.
3. Secure revenue share
Stripe Connect vaults cards, handles KYC and deposits partner commissions automatically, aligning incentives and minimising Cobber’s PCI scope.
4. iframe SDK ergonomics
The widget uses window.postMessage() with strict origin whitelists and JWT‑signed payloads, mitigating tampering risks highlighted by the OWASP community.
Total‑Cost‑of‑Ownership (TCO) Analysis
Lanex compared Cobber’s legacy VM against Fargate + Atlas over three years:
- CapEx avoided: AU $120 k hardware & licences.
- Compute savings: 70 % on non‑production via Fargate Spot; 40 % overall due to auto‑scale to zero at night.
- Ops FTE: 0.5 → 0.15 (automation & managed services).
- NPV improvement: AU $188 k.
Lessons Learned
- Simple iframe plus good docs trump custom SDKs for partner velocity.
- Serverless containers handle unpredictable marketplace peaks cost‑effectively.
- Document databases accelerate fast‑changing integrations.
- Payments‑as‑a‑service speeds compliance and revenue sharing.
- Observability first: tracing saves future debugging hours.
Next Steps for Cobber
- Multi‑region deployment to serve New Zealand classifieds with < 100 ms latency.
- Background image optimisation off‑loaded to AWS Lambda for cost‑per‑use.
- SEO quote API so partners can pre‑render delivery prices, improving search rankings.