# Stack Teardown — sample (synthetic) ## Context - **Industry:** B2B SaaS (fictional sample) - **Headcount:** 50-150 - **Stage:** Series B - **System of record:** Ledgerly ## Stack map ### 1. Ledgerly → MailMotion - **Sync:** one-way, nightly batch - **Flow:** billing accounts → marketing contacts - **Notes:** Billing is the only system with reliable identity; the nightly batch means marketing works with day-old truth. ### 2. StreamPipe → Glacier - **Sync:** one-way, streaming - **Flow:** product events → warehouse - **Notes:** Glacier is the analytics system of record — but TrackPanel ingests the same client-side events in parallel, so there are two versions of every funnel number. ### 3. MailMotion → PushPilot - **Sync:** none (manual CSV, ad hoc) - **Flow:** email audiences ↔ push audiences - **Notes:** Two messaging tools, two contact stores, no sync. Duplicate and conflicting contact records are structural, not accidental. ## Failure modes ### No single customer identity across billing and engagement **Impacted:** Ledgerly, MailMotion, PushPilot A paying customer in Ledgerly can be two different contacts in MailMotion and PushPilot. Churn-save campaigns and dunning emails routinely target the wrong record. ### Double-counted product events **Impacted:** TrackPanel, StreamPipe, Glacier TrackPanel and StreamPipe both fire client-side. Activation and funnel metrics disagree between the product team's dashboard and the warehouse — and both are defended in meetings. ### Marketing operates on day-old billing truth **Impacted:** Ledgerly, MailMotion The nightly batch means upgrade/downgrade signals reach messaging up to 24h late; win-back sends fire at customers who already renewed. ## Cost leakage ### Overlapping messaging (MailMotion + PushPilot) - **Range:** $18,000 – $36,000 / year - **Assumptions:** both tools priced on contact volume; ~40% audience overlap held in both - **Redundancies:** duplicate contact storage, two audience-builder teams-of-one ### Redundant event ingestion (TrackPanel + StreamPipe) - **Range:** $24,000 – $48,000 / year - **Assumptions:** event-volume pricing on both; same client-side events shipped twice - **Redundancies:** parallel pipelines for identical events ### Manual reconciliation labor - **Range:** $30,000 – $60,000 / year - **Assumptions:** ~0.25–0.5 FTE across RevOps and finance; monthly board-metric fire drills - **Redundancies:** spreadsheet exports bridging Ledgerly ↔ Glacier every close ## Metric impact The metric everyone argues about — activation rate — differs by up to 14 points between TrackPanel and Glacier because of double-fired client events. Until ingestion is unified, neither number is wrong; both are unaccountable. ## Target architecture Ledgerly stays the identity anchor. One event pipeline (StreamPipe → Glacier) with TrackPanel reading FROM the warehouse rather than ingesting independently. Messaging consolidates to a single contact store with push and email as channels, not silos. ## Migration sketch Phase 1: reconcile identities read-only and quarantine every ambiguous merge (no writes). Phase 2: dual-run messaging on the unified contact store with a completeness ledger proving every record's fate. Phase 3: cut over channel by channel behind human approval; decommission the duplicate pipeline last. ## Hook You have two versions of every customer and two versions of every number. The fix is not another tool — it is an accounting of the ones you have.