The Plutobee Enterprise Delivery Framework.
Eight phases. Documented at every gate. Signed off by both engineering and the customer. This is how a Plutobee engagement actually runs, from a first email to year-three support.
Discovery.
Two weeks. Two senior engineers and a designer in your context. No deck-writing, no proposal theater.
Deliverables: Problem statement reviewed by your stakeholders, success criteria with measurable thresholds, scoped MVP backlog, identified constraints (regulatory, infrastructure, data, integrations), risk register with mitigations, and a written SoW with fixed-price or T&M commercial structure.
Signed off by: Customer product owner, Plutobee delivery lead.
What is real: We will tell you in week two if the project is the wrong shape, the wrong scope, or the wrong moment. We have walked away from $400k engagements at this gate. The cost to you is zero.
Architecture.
Two to four weeks. The C4 model gets drawn. The data model gets normalized. The non-functionals get teeth.
Deliverables: System context diagrams (C4 Level 1 through 4), data flow diagrams aligned to your regulatory regime (GDPR, HIPAA, PCI DSS, FedRAMP Moderate, SOX as applicable), API contracts in OpenAPI 3.1 with example payloads, infrastructure-as-code skeleton in Terraform, threat model using STRIDE, latency budgets per request path, capacity model for 12 months out.
Signed off by: Customer architecture review, Plutobee principal engineer, customer security team.
What we will not do: Skip threat modeling because the launch date is tight. Recommend microservices when you have one team. Use a vector database because it sounds modern. We pick boring, proven tools where they fit.
Development.
The squad is on your repo. Daily stand-ups, two-week sprints, weekly stakeholder review with shipped pixels.
Practices: Trunk-based development with feature flags. Required peer review on every change. Conventional commits and signed commits in protected branches. CI on every PR with required green builds. Test pyramid enforced by coverage thresholds in the pipeline. Documentation is a PR requirement, not an afterthought.
Pace metrics we publish weekly: Lead time for changes, deploy frequency, change failure rate, mean time to recover. We adopt the DORA four. We hit elite tier on every committed engagement past month two.
What we deliver every sprint: Working software in your staging environment, a demo recording, a retrospective with action items, and an updated risk register.
QA & Validation.
Tests are an asset. We invest in them like the production code they protect.
Layers: Unit (Vitest, Pytest, Go test), Integration (Postgres testcontainers, in-memory queues), Contract (Pact for service boundaries), Component (Playwright for UI, Detox for mobile), End-to-end (Playwright with sharded parallel runs), Visual regression (Percy or Chromatic), Performance (k6 with thresholds wired to CI), Accessibility (axe-core + manual screen-reader sweep), Mutation (Stryker for critical paths).
Coverage gates: 80% for new code as a hard CI gate. Critical paths must hit 95% with mutation score > 70%. Visual diffs require human approval. a11y violations block merge.
Output: Pre-production test report with reproducible run, defect log, and risk acceptance for any deferred items, signed by your QA lead.
Security Review.
Stinger Security, our in-house red-team practice, runs against your build before production. Findings are remediated, not waived.
Standard package: Static analysis (Semgrep with custom rules), dependency CVE scan (Snyk + GitHub Advanced Security), container image scan (Trivy), SBOM in CycloneDX format, dynamic analysis on staging (OWASP ZAP automated, Burp manual), secrets scan, IAM policy review with Cloudsplaining, and a 5-day manual pen-test against the live staging system.
Compliance overlay: ISO 27001 control coverage mapping. SOC 2 evidence collection for trust-center publication. NIST 800-53 control alignment if Federal. HIPAA Security Rule mapping if PHI is in scope.
Output: Executive summary, technical findings, severity-weighted remediation backlog with SLAs, attestation letter you can include in your own customer trust packages.
Deployment.
Zero-downtime. Documented runbooks. Rollback rehearsed before launch, not after.
Approach: Blue/green deploys for stateless services. Online migrations with shadow reads and dual writes for stateful ones. Feature-flag-gated rollouts at 1% / 10% / 50% / 100% with metric-based promotion. Pre-warmed connection pools. Database migrations decoupled from application releases.
Observability ready on day one: Distributed tracing across services (OpenTelemetry to Datadog, New Relic, or Honeycomb), structured logging with correlation IDs, golden-signal dashboards (latency, traffic, errors, saturation), alerting that someone is paid to respond to. We do not ship to production without runbooks for the top 10 alerts.
Launch day: War-room channel, named on-call rotation across both teams, hourly synthetic checks, communication plan to your customers in case of degradation.
Support & Operations.
The build does not end at launch. The maintenance contract is where most agency relationships fail. We design ours to survive year three.
Support tiers:
Standard: Business-hours response within 4 hours, fixes within 5 business days for non-critical defects. Quarterly architecture review. Annual security re-test.
Enterprise: 24/7 on-call with 15-minute response for P1, 30-minute for P2. Monthly architecture review with your CTO. Continuous Snyk and dependency scanning. Quarterly internal red-team. SLA-backed uptime.
Strategic: Embedded SRE on your team. Custom SLAs. Co-developed roadmap. Stinger Security on annual retainer with named-day pen-test.
What every tier includes: Named engineer who knows your system end-to-end. No outsourced L1. No reading from a script. The person who shipped it is the person who responds.
SLAs & Commercial Terms.
Numbers we put in writing. Not aspirations. Standard contractual commitments.
Availability: 99.9% for Standard tier, 99.95% for Enterprise, 99.99% for Strategic. Measured monthly. Credits apply automatically without a customer-side ticket.
Response time: P1 (production down) under 15 minutes Enterprise+, under 1 hour Standard. P2 (degraded) under 1 hour Enterprise+, under 4 hours Standard.
Resolution time: Mean time to recover under 60 minutes for P1 across the last 18 months on retainer customers. Published quarterly.
Communication: Status page incident open within 5 minutes of detection. Customer notification within 30 minutes of confirmed material incident. Postmortem within 5 business days, published if requested.
Money: Net 30 invoicing standard. SLA credits computed automatically. No reactivation fees. No mid-contract repricing. Volume discounts after year one.
Want the full delivery manual?
130 pages, written by our principal engineers. NDA-bound, free to qualified prospects.
Request the manual →