Data science and visualization are about turning noise into narrative and extracting actionable truth from complexity.
The science is foundational and the visualizations and transformative.
A great visualization doesn’t just inform — it reshapes decisions by making the invisible immediately undeniable.
Identified $237 M in annualized cost-lost through statistically significant defects (Over $130 M Solved)
Created 4 Production Dashboards to track, monitor, and identify defects in Outbound Operations
Trained 30+ analysts and scientists in statistical inference, regression, and other scientific approaches
Combined data feeds from several ERP software to unify reporting in a one-stop-shop for various teams
Worked cross functionally with over 13 teams including Operations, Engineering, Data Science, Business Intelligence, Software Development, Finance, and others.
Root Caused 6 defects and provided business solutions that resulted in over $100 million in combined cost-savings initiatives
Report: Backlog Recommendation One-Sheet
Definitions
Pickable Backlog - The amount of work available to be planned for and picked
TUR - Target Unit Rate
MoW - Pickable Backlog converted into Minutes of Work (Pickable Backlog/(TUR/60)
Time LOW - Time Low on Work; Less than 3 Minutes of Work per Packer/Inductor
Methodology
Induct Rate, Pack Rate, Pick Rate, Time Under Buffer, and Time Low on Work were analyzed to define the backlog limits and identify where the most egregious step changes occur based on statistical and correlative significance in relativity to key performance indicators.
Summary
There is significant statistical correlation between Outbound processing rates and Time Under Buffer and available minutes of work in pickable backlog.
By increasing pickable backlog we are able to effectively increase processing rates, and decrease Time Under Buffer, and Time Low on Work.
Recommendation
There are three levels of backlog where significant step changes occur.
- 36 Minutes - This is the minimum minutes of work in backlog we should maintain at all times. Dropping below these values causes significant rate degradation and cost implications.
- 47 Minutes - This is the minutes of work in backlog where we effectively eliminate WIP related Time Out of Work.
- 61 Minutes - This is the minutes of work where all processes are able to operate healthily. Low Backlog symptoms are 95% relieved.
Financial Summary
Cost Per Minute of Work Under 60
$6.7 M Annualized (ARS Network)
Cost of Maintaining 30 Minutes of Work
$201.45 M Annualized (ARS Network)
Bottom Line Up Front
NA ARS is experiencing $154 M in annual Late Slam DPMO costs driven by two critical process failures:
· AFE shift handoff deficiencies leading to sorted dwells (PPW exceeding 80 units)
· Systemic capacity imbalances (24% over-pick rate vs. pack/dock capacity) causing scanned dwells
Resolution requires implementation of unified capacity metrics and standardized AFE handoff procedures, targeting 10% reduction in Late Slam DPMO ($15.4M savings opportunity).
The North American ARS network faces critical operational challenges due to Late SLAM DPMO, projecting annual costs of $154 M based on 41 M affected units at $3.80 per miss. This kaizen initiative targets a 10% reduction in Late SLAM DPMO, aiming to achieve savings between $7.7 M (minimum) and $15.4M (target).
Network analysis reveals systemic capacity imbalances driving Late SLAM DPMO:
· Multi's experience 103.6% higher defect rates than Singles processing
· Over-picking to plan by 24% (72 UPH per picker)
· Pack over-processing by 25% (54 UPH per packer)
· Induct under-processing by 7% while over-staffing by 10%
These imbalances result in a net over-pick of 28 UPH per picker, significantly increasing Pick-to-Slam cycle times - the primary driver of Late SLAM DPMO. Key contributors include pick-to-slam cycle time variations (42.42% of misses), inadequate pack staffing to plan (38.96%), and inaccurate induct rates (33%).
Critical sites exemplifying these issues include:
· ATL2: Late Slam percentage increasing from 0.8% to 1.1%, with 43.8% Time Under Backlog (worst in network)
· BTR1: 2.1% Late Slam (highest in network), extended pick-to-slam cycle times (107 minutes)
· DEN4: Late Slam more than doubling from 0.8% to 1.9%
Root causes are concentrated in two areas:
1. AFE Handoff failures during shift transitions (walls exceeding 80 PPW)
2. Pack vs. dock capacity misalignment
Our three-phased solution approach includes:
1. Immediate Actions (1-30 days): Standardizing processes and STP adherence protocols (3-5% reduction, $3.2M-$3.8M savings)
2. Medium-Term Solutions (31-180 days): Implementing integrated capacity planning and automated cycle time monitoring (5-8% reduction, $5.1M-$6.4M savings)
3. Long-Term Solutions (181-365 days): Deploying comprehensive balanced staffing models and predictive analytics (8-10% reduction, $7.7M-$15.4 M savings)
Key focus areas include developing unified dock capacity metrics, standardizing AFE handoff procedures, and implementing robust process controls for shift transitions. Success will be measured through reduced Late SLAM DPMO, improved pick/pack/dock balance, and enhanced process stability across all sites.
This initiative will be supported by a comprehensive communication plan, including daily business reviews via Slack, QuickSight reports, weekly internal reviews, and network-wide updates through PMO change digests. Regular office hours will ensure understanding and compliance as new programs are implemented.
The above image displays Minitab.
