DOSSIER — FILE 04 / PROJECTS

Case files.

A few of the systems I've built, broken, and rebuilt — chosen because each one left the organization measurably better off than the last version of itself.

Eleven-Instance CRM Consolidation

Program owner

Merged eleven regionally customized CRM instances into one governed platform without freezing the sales org for a quarter — the usual cost of this kind of migration.

Data GovernanceChange ManagementSalesforceMigration

OUTCOME

One platform, one data model, and a 40% drop in "which CRM has the real numbers" support tickets.

Every region had bent the CRM to its own habits, which meant eleven different definitions of "qualified lead" and no shared source of truth. The migration itself was the easy part; the negotiation over whose workflow became the standard was not. Ran the consolidation region-by-region with a two-week parallel-run window each time, and made every workflow decision reversible for the first month. That reversibility clause is the reason the program didn't get vetoed in week three.

ERP–CRM Event Bridge

Lead engineer, then handed off to a dedicated integrations team

Retired a fragile nightly batch sync between the ERP and CRM in favor of an event-driven bridge, eliminating the "why doesn't sales see this order yet" complaint almost entirely.

IntegrationEvent-Driven ArchitectureERPAPI Design

OUTCOME

Data-lag escalations fell from roughly weekly to fewer than three in the following year.

The batch job ran at 2 a.m. and failed silently often enough that "just re-run it" had become a documented step in someone's onboarding guide. Replaced it with a message-driven bridge that pushed order and account changes as they happened, with a dead-letter queue instead of a shrug. The harder part wasn't the code — it was convincing finance that real-time sync wouldn't destabilize month-end close. Ran both systems in parallel for six weeks with a reconciliation report before finance agreed to cut over.

Intent-Weighted Lead Scoring Engine

Product owner and lead architect

Replaced a static, demographic lead-scoring model with one weighted on behavioral intent signals — page depth, reply latency, and pricing-page returns — cutting response time by more than 80%.

CRMLead ScoringSQLSales Ops

OUTCOME

First-response time dropped from 14 hours to 87 minutes across 11 regional sales teams.

The old model scored leads on job title and company size — signals that were easy to capture and mostly useless. Rebuilt the scoring layer around what a lead actually *did*: how many pages they read, how fast they replied, whether they came back to the pricing page uninvited. Rolled it out region by region rather than all at once, comparing response-time and conversion numbers against a control group at each stage. By the third region, the case made itself and the rollout stopped needing to be defended.