Analytics Manager · 8 Years · FinTech, CPG & Healthcare
I turn complex, messy data into decisions executive leadership can act on. Specializing in collections, workforce, and performance analytics for U.S. lending and banking — with hands-on domain experience across healthcare and CPG.
Each project below reflects actual career work — analytics that moved numbers and changed how leadership thought about the business.
When executive leadership suspected agent underperformance behind a 9.3pp drop in Promise-to-Pay rates, I built a waterfall decomposition across two years of data — isolating IVR automation (480 bps), hardship diversion (50 bps), portfolio strategy shifts (340 bps), and macro factors (70 bps). The analysis reframed a "performance problem" as strategic success.
Mortgage delinquency waterfall decomposition for a leading real estate platform — same methodology applied to HMDA public data. Paired with a Custom GPT trained on the findings.
Coming soon ↗Despite rising call volumes, recoveries had plateaued — and leadership doubted the data. I validated all pipelines end-to-end, then used SQL and Python cohort analysis to identify concurrent-loan customers as an under-addressed segment. A t-test confirmed significantly lower PTP rates. Targeted agent coaching drove a 15% PTP lift per customer.
Multi-product BNPL borrower segmentation for a leading buy-now-pay-later lender — concurrent obligations analysis using LendingClub public data. Paired with a Gemini Gem configured as a Collections Analyst AI.
Coming soon ↗Built a program-level Tableau dashboard from scratch for a major LOB with 100+ users across executive, manager, and associate tiers. Started with stakeholder workshops to define KPIs, unified fragmented data pipelines, introduced LOD expressions for performance optimization, and ran phased UAT before enterprise rollout.
3-tier Tableau dashboard architecture applied to a leading U.S. communications company — unified exec, manager, and agent views from a single pipeline. Paired with a Claude Skill for automated KPI dictionary generation.
Coming soon ↗Speech analytics revealed 50%+ of Debt Settlement Company calls weren't handled effectively. I built a Tableau dashboard tracking three recovery outcomes (Promise-to-Pay, Hardship enrollment, CCCS), redesigned the agent knowledge base, introduced WalkMe real-time guidance, and launched a gamified incentive contest — shifting agent behavior at scale.
Customer churn intervention analytics for a leading U.S. telecommunications company — speech analytics to NPS signal mapping, agent coaching effectiveness, and retention gamification using telecom churn data.
Coming soon ↗Rebuilt the workforce capacity model for 5 ops teams using an Erlang-C formula combined with Cresta AI AHT projections. Modeled best, expected, and worst-case volume scenarios — reducing overstaffing by 30 FTE and delivering $1.4M in annualized savings. Separately, Cresta NLP transcript analysis quantified IVR and AI deflection opportunities worth ~$370K annually.
Interactive Erlang-C staffing calculator built as a Claude Artifact — input volume, AHT, and SLA targets, output FTE requirements by shift and scenario. Paired with a Gemini Live walkthrough of the model logic.
Coming soon ↗At Fractal Analytics for a Fortune 500 CPG client, applied the Hook Model framework to diaper category buyers — segmenting customers into Hooked, One-Timers, Delayed Repeaters, and Switchers. Sample and coupon customers were 45% more likely to repurchase within 30 days. Built a Power BI dashboard adopted by 200+ users across the analytics team.
Hook Model applied to travel rebooking behavior for a leading online travel platform — segment analysis using public travel data. Audio deep-dive generated via Gemini NotebookLM for hiring managers to listen on commute.
Coming soon ↗The existing incentive structure rewarded total balance volume only — so agents chased easy accounts and ignored high-delinquency buckets. I redesigned the framework using a weighted rank model: Balance Moved (70%) + Buckets Cured (30%), with higher delinquency buckets (B4, B5) earning more points. Top 70% of agents eligible, with a 70/30 bonus split.
Interactive incentive rank calculator built as a Claude Artifact — input agent performance data, toggle the 70/30 weighting, and see rank shifts live. Visitors interact with the actual model logic.
Coming soon ↗Work Experience
How I Work
Every project follows the same discipline — start with the business question, not the data.
Every analysis starts with a stakeholder workshop — not a data pull. Question → Metric → Data. Measure what matters, not what's easiest to query.
Metrics lie when context is missing. Validate pipelines, isolate variables, never present a finding I can't reconcile back to its source. Trust — then verify.
Insights that don't change decisions are just reports. Designed for the 30-second scan — clear KPI hierarchy, one implication, one recommended action.
Let's Connect
Whether you're building something from scratch, rethinking how your team uses data, or just want a second pair of eyes on a thorny BI problem — reach out. I'm actively exploring senior analytics, BIE, and BIE Manager roles in Seattle. Good analytics conversations are always worth having.