shipped May 15, 2026
OCC cost reduction analytics program
Outlier detection, cohort journey analysis, and finance modeling that took an unfavorable Q1 OCC program to $28M+ favorable for the year — against a $300M annual spend.
$35.78M
favorable to budget
#python#outlier-detection#finance-modeling#cost-reduction
Context
A consumer wireless carrier needed to reduce Other Credits & Charges (OCC) — credits issued to customers that directly reduced revenue. The program was unfavorable in Q1 with no structured data-driven approach to diagnosis or intervention. Annual OCC spend was approximately $300M.
The problem
No automated outlier detection, no systematic field engagement cadence, and no finance outlook methodology. Performance management was reactive and manual.
Methodology
Outlier detection
- Built outlier detection methodology identifying rep-level OCC behavior vs. expected distributions, segmented by title, region, and call type
- Used statistical distributions to set fair, differentiated performance thresholds across a wide-ranging population
- Automated detection via Python scripts; daily flagging of reps crossing thresholds
- Created automated email alerts to supervisors delivering rep-specific OCC data directly to field managers — eliminating manual lookup and accelerating coaching response time
Cohort & journey analysis
- Redesigned cohort methodology to leverage new data from OCC Modernization
- Built subjourneys to trace customer OCC events back to upstream root causes (system issues, process gaps, behavioral patterns)
- Separated controllable vs. uncontrollable OCC spend; focused field intervention on controllable only
- Developed end-to-end automated KPI analysis: scheduled EDW pulls → organize → analyze → email to stakeholders
Finance modeling
- Created top-down OCC budget methodology: forecasted OCCs based on revenue drivers while accounting for planned pressures and savings
- Built OCC Incentive Pilot model:
- Constructed supervisor performance distributions
- Applied standard deviations to create equitable tiers
- Ran 3-scenario cost-benefit analysis across 4 supervisor tiers
- Designed pay-for-performance incentive structure
- Maintained quarterly finance outlooks with rolling pressure / savings tracking
Performance management cadence
- Launched bi-weekly field partner calls: reporting, coaching, accountability
- Built and distributed holistic performance packages to all stakeholders on a defined schedule
Stack
Python, SQL / Teradata EDW, Excel (financial modeling), automated email distribution (Python).
Impact
- $35.78M favorable to budget; $14.97M savings YoY
- User Level Controls: $2.6M/month cost savings
- Supervisor workload reduced 33%
- Outlier OCC/contact reduced 15%; 65% rep graduation rate
- Turned program from unfavorable in Q1 to $28M+ favorable for full year
- Manager: “proactively seeks opportunity… able to provide a custom solution based on the needs of the business”