ParallelHQ

Data scientists and analysts at mid-sized banks and NBFCs today must manually aggregate, cleanse, and visualize credit portfolio data from disparate core banking and CRM systems. This process—often involving Excel, Tableau, and custom Python scripts—consumes an average of 11 hours per analyst per week on report generation alone (source: internal interviews with 8 target institutions, Feb 2025). The manual workflow introduces a 4-7% error rate in key risk metrics like PD and LGD, leading to mispriced loans and regulatory exposure.

Business case: 250 targetable mid-sized institutions (source: RBI 2023-24 bank list, filters applied) × 4.2 average analysts per institution (source: LinkedIn headcount sampling) × 11 hours saved weekly (source: validated baseline) × $42/hour fully-loaded cost (source: Regional Cost Benchmarks, India B2B SaaS) × 48 working weeks = $2.33M/year recoverable analyst capacity. If adoption is 40% of estimate: $932K/year. This MVP must prove we can capture at least 20% of that time-savings.

ParallelHQ is a no-code dashboard that ingests raw credit data and renders regulatory-grade visualizations in 4 clicks. It is not a core banking system, a loan origination platform, or a replacement for a data warehouse—it is a read-only analytics layer on top of existing systems.

Primary Metrics:

Guardrail Metrics (must NOT degrade):

What We Are NOT Measuring:

• Total registered users (vanity; we care about weekly actives)
• Number of visualizations created (could be low-quality test clicks)
• Social media mentions (irrelevant for deep B2B workflow tool)

Decision: Authentication & Data Isolation Model Choice Made: Email/password per institution (no SSO), database-level row isolation by org_id. Rationale: SSO (Okta, Azure AD) adds 3 weeks of integration time; password auth gets pilots faster. Row-level isolation is simpler than schema-per-tenant for PostgreSQL. ──────────────────────────────────────────────── Decision: PDF Export Fidelity vs. Speed Choice Made: Prioritize 100% data accuracy over pixel-perfect RBI formatting. Rationale: Pilot institutions will accept "draft watermarked" PDFs if numbers are correct. Perfect formatting is a scaling problem, not validation problem. ──────────────────────────────────────────────── Decision: Industry Benchmark Data Source Choice Made: Use synthetic benchmarks based on RBI annual reports, labeled "Illustrative". Rationale: Licensing real benchmark data (CRISIL, CIBIL) requires 6-month negotiations. Synthetic data proves the UX; we can replace with real data post-validation. ────────────────────────────────────────────────

Pre-Mortem: It is 6 months from now and this feature has failed. The 3 most likely reasons are:

1. Compliance officers at pilot institutions blocked production usage due to data residency concerns, despite NDAs, because we couldn't offer a local Mumbai region VPC fast enough.
2. Analysts found the 8 pre-built metrics insufficient and demanded custom calculations, but our no-code builder was too rigid, causing them to revert to Python after 2 weeks.
3. FICO responded by bundling basic analytics into their existing score subscription, neutralizing our price advantage before we signed 10 pilots.

Success looks like: Pilot analysts stop their weekly "Python Thursday" scripting ritual. Compliance managers reference ParallelHQ screenshots in board meetings. The sales lead reports, "They're asking how to add their second loan book, not if the tool is secure." We have a 6-month product roadmap co-created with 4 paying institutions.

Competitors solve the credit analytics problem in two ways: FICO Dashboard provides entrenched but generic risk scores requiring expensive customization ($250K+ engagements). Tableau/Custom Python is hired for unlimited flexibility but requires scarce data science talent, creating a 3-6 week lag for new reports.

Our wedge is no-code simplicity for regulatory reporting because analysts need compliance-ready views today, not in 6 weeks, and cannot wait for IT or data science backlogs.

We hypothesize that credit analysts will replace their manual weekly reporting workflow (Excel → Python → Tableau) with ParallelHQ if it delivers audit-ready visualizations in under 15 minutes from raw data upload, with zero SQL or code required.

Business case math: 11.2 hours × $42/hour × 48 weeks = $22,579/year/analyst recoverable time. 1,050 analysts (250 inst × 4.2) × $22,579 = $23.7M total addressable time cost. Our 20% capture target = $4.74M/year.

P0 (Weeks 1-6): Core validation engine.

• CSV upload with 5-column validation (customer_id, loan_amount, default_flag, score, date)
• Three pre-built visualizations: Score distribution histogram, 12-month default rate trend, vintage cohort analysis. /``` ┌─────────────────────────────────────────────────────────────────────┐ │ Dataset Manager [Upload New] [API Sync] │ ├─────────────────────────────────────────────────────────────────────┤ │ loans_q1_2025.csv ✅ Validated 1.2M rows [Explore →] │ │ test_portfolio.csv ⚠ 3 errors 45K rows [Fix Errors→] │ │ corporate_jan.csv ✅ Validated 450K rows [Explore →] │ └─────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────┐ │ Credit Score Distribution [Export PDF] [←Back] │ ├─────────────────────────────────────────────────────────────────────┤ │ ████████████████████ 720-750: 24% │ │ ████████████ 680-710: 18% │ │ █████████ 650-680: 14% │ │ Industry Avg: █████████████ 690 mean │ │ Your Mean: 705 (+15 bps) │ └─────────────────────────────────────────────────────────────────────┘

P1 (Weeks 7-10): User workflow.
- Drag-drop metrics builder (max 4 metrics per report).
- Export to PDF with company watermark.
- Role-based access: admin (upload), analyst (view/build), viewer (read-only).

Phase 1 — MVP Pilot (10 weeks) Target: 10 paid pilot institutions ($1K/month, annual contract).

US#1 — Analyst Upload & Validate

• Given an analyst has a CSV from their core banking system
• When they upload it and our validator detects column mapping
• Then they see a validation summary within 60 seconds with ≤2 required manual corrections
• If story fails, analysts abandon during onboarding as mapping is too complex
• Validated by PM against 5 real customer CSV samples by Week 6

US#2 — No-Code Report Creation

• Given an analyst is on the dashboard with validated data
• When they drag 2 metrics (e.g., "NPA %" and "Provision Coverage") onto a report
• Then a compliant visualization renders in <5 seconds with 100% data accuracy
• If story fails, users revert to Excel, killing the core value prop
• Validated by Lead UX against 10 tasks with pilot users in Week 9

Out of Scope (Phase 1):

1. Custom SQL editor → Replaced with curated metric dropdown. Rationale: If analysts need SQL, they’re not our target user (non-technical).
2. White-labeling & custom themes → Standard ParallelHQ branding. Rationale: Compliance teams care about data accuracy, not colors.
3. Real-time alerts & notifications → Manual export triggers. Rationale: Email alerts are Phase 2 after proving core visualization value.
4. Multi-entity consolidation → Single portfolio view only. Rationale: 70% of target institutions manage one primary loan book.
5. Audit trail log export → UI-only view. Rationale: MVP auditors will accept screen recording; automated log export is compliance complexity.

1. Compliance Acceptance: Analysts will trust and use an external SaaS tool for sensitive credit data without requiring on-prem deployment.

   • Probability: High | Impact: High
   • Mitigation: Pre-sell MVP access to 5 pilot institutions with signed NDAs & data processing agreements. Offer 30-day data purge guarantees. Owner: CEO, due by Week 3.

2. Data Schema Universality: Our predefined data validators will handle 80% of customer CSV formats without custom mapping.

   • Probability: Medium | Impact: High
   • Mitigation: Build schema detection for 10 most common core banking exports (Finacle, Flexcube, TCS BaNCS). Fallback: manual column mapper in UI. Owner: Lead Data Eng, due by Week 6.

3. No-Code Sufficiency: The drag-drop report builder will satisfy 70% of ad-hoc analysis needs without SQL fallback.

   • Probability: Medium | Impact: High
   • Mitigation: Ship with 8 pre-built metrics (PD, LGD, EAD, NPA ratio, etc.) and test with pilot users in Week 8. If <70% satisfaction, pivot to curated template library. Owner: PM, validated by Week 9.

Kill Criteria — we pause and conduct a full review if ANY are met within 90 days of pilot launch:

1. <30% of pilot users (≤3 of 10) upload a second dataset after initial trial.

2. 50% of report building sessions require IT assistance (measured via support ticket tagging).

3. Median time from upload to first visualization >25 minutes (2x our target).
4. Any critical data breach or compliance violation (RBI data localization requirement not met).

Minimum Viable Experiment (4 weeks, $15K burn): Instead of building the full dashboard, create a static PDF generator that takes a standardized CSV template and emails back 3 pre-rendered charts with benchmark overlays. Manual onboarding via Zoom call.

• Test: Will analysts send their sensitive data to an email-based service for a 10x faster report?
• Metric: Conversion rate from inbound lead to data send >25%.
• Kill: If <10% of 30 contacted prospects send data, the compliance barrier is too high for the SaaS model; pivot to on-prem consulting.