Salesforce Interview Guide: Cloud, CRM & V2MOM Framework

Multi-Tenancy Approaches: 1. Shared Database, Shared Schema (Salesforce's approach): • All tenants share tables with a tenant_id column • Every query includes WHERE tenant_id = ? • Custom fields stored in flex columns or a metadata-driven schema • Pros: efficient resource usage, easy deployment • Cons: complex query isolation, noisy neighbor risk 2. Shared Database, Separate Schemas: • Each tenant gets their own schema within a shared database • Better isolation but harder to manage at scale (10K+ tenants) 3. Separate Databases: • Maximum isolation but highest cost and operational complexity Key Design Components: • Metadata Engine: - Store tenant-specific configurations (custom fields, layouts, workflows) as metadata - Compile metadata into runtime objects for performance - This is Salesforce's core innovation — the 'metadata-driven architecture' • Governor Limits: - Enforce per-tenant resource limits (API calls, query rows, CPU time) - Prevent any single tenant from degrading service for others - Example: max 100 SOQL queries per transaction, max 50K rows returned • Data Isolation: - Row-level security enforced at the platform level, not application level - Encryption per tenant with separate key management - Audit logging for compliance (SOC 2, HIPAA, FedRAMP) Scaling Strategy: • Shard by tenant groups (pods/superpods) • Route tenant requests to the correct pod via a tenant-routing service • Rebalance tenants across pods based on usage patterns • Separate compute for heavy batch jobs (reports, data loads)

Lead Scoring Pipeline: 1. Signal Collection: • Demographic data: company size, industry, job title, location • Behavioral data: website visits, email opens, content downloads, form submissions • Firmographic data: revenue, employee count, technology stack (from third-party enrichment) • Engagement recency: how recently the lead interacted 2. Scoring Model: • Point-based system with configurable weights per customer • Demographic fit score (0-50): how well the lead matches the ideal customer profile • Behavioral engagement score (0-50): how actively the lead is engaging • Decay function: reduce score for leads that go inactive • Threshold tiers: Hot (80+), Warm (50-79), Cold (<50) 3. Real-Time Processing: • Event-driven architecture: each lead action triggers a score recalculation • Use a stream processor (Kafka Streams or Flink) for real-time scoring • Cache current scores in Redis for fast retrieval • Persist score history to a time-series database for trend analysis 4. Lead Routing: • Round-robin within territory for equal distribution • Skill-based routing: match lead industry to rep expertise • Capacity-aware: don't route to reps who are at their lead limit • SLA enforcement: hot leads must be contacted within 5 minutes CRM Integration: • Sync scores back to the Lead/Contact record in Salesforce • Trigger workflow rules: auto-create tasks, send alerts, update lead status • Dashboard for sales managers: conversion rates by score tier, rep performance Salesforce Context: This is essentially what Salesforce Einstein Lead Scoring does. Mentioning Einstein, Marketing Cloud engagement tracking, and Pardot integration shows domain awareness.

Why Salesforce Asks This: Trust is Salesforce's #1 value — they display real-time system status on trust.salesforce.com. They need engineers who understand that innovation cannot come at the cost of reliability for enterprises that depend on the platform. Strong Answer Structure: • Context: 'Our team wanted to migrate a critical authentication service from a monolith to microservices to enable faster feature development. The service handled login for all enterprise customers — downtime was unacceptable.' • The Tension: 'A full cutover would be faster to implement but risky. A gradual migration would be slower but safer. I advocated for the gradual approach despite pressure to move fast.' • Method (V2MOM Thinking): - Vision: Modern auth architecture enabling 10x faster feature delivery - Obstacle: Zero-downtime requirement for 2,000+ enterprise customers - Method: Strangler fig pattern — route 1% of traffic to the new service, validate, increase gradually - Measures: Error rate < 0.01%, p99 latency within 5% of baseline, zero auth failures • Outcome: 'Migration completed over 8 weeks with zero customer-impacting incidents. The new architecture enabled us to ship OAuth 2.0 support in 2 weeks vs the estimated 3 months on the old system.' Salesforce Signal: You prioritized trust over speed, used data to manage risk, and still delivered innovation — just on a responsible timeline.