Supply Chain · Network Design · Optilogic Cosmic Frog · 13-Week Engagement

North American Network Design:
Finding $2.3M in Annual Savings

A 13-week end-to-end network study for a major logistics distributor — spanning data validation, SQL pipeline engineering, Tableau dashboards shared directly with the client, and scenario analysis across a 25-facility network processing 1.2B lbs. of annual freight.

Optilogic Cosmic Frog SQL Server / SSMS Tableau Python SQL Network Optimization Data Validation LTL / TL Rating
$2.3M
Projected annual savings from recommended network
1.2B
Pounds of freight modeled in baseline
25
Facilities across US, Mexico, and Canada
<5%
Variance between actual and modeled baseline costs
The Problem

A 25-facility network built around a different business

A major North American logistics distributor had grown its facility network over time across the US, Mexico, and Canada — but the network's structure reflected historical decisions, not current volume flows. With significant business wins and losses reshaping demand, and lease expirations creating natural inflection points, the question was: what does the optimal network actually look like, and how much is the current one costing?

The engagement covered 13 weeks — from raw shipment data through validated baseline modeling, adjusted demand forecasting, nine distinct network scenarios, and a final recommended roadmap presented to executive leadership.

My Contribution
My focus was on the data foundation and client-facing deliverables: validating and cleansing 456K+ shipment records, building SQL pipelines in SSMS to structure raw data into model-ready inputs, and developing the Tableau dashboards used by client leadership to interrogate results throughout the engagement. I also contributed to baseline model flows and worked closely alongside the scenario analysis in Optilogic Cosmic Frog, gaining hands-on exposure to how network constraints, routing rules, and facility assumptions are configured in production optimization models.

Project Workstream

13 weeks, end to end

01

Data Collection & Validation

456K+ shipment records cleaned, matched, and validated against actual costs within 5%

02

Baseline Modeling & Dashboards

SQL pipelines in SSMS, historical baseline flows, Tableau dashboards shared with client leadership

03

Adjusted Baseline & Scenarios

Demand adjusted for wins/losses; 9 network scenarios modeled in Optilogic Cosmic Frog

04

Recommendations & Roadmap

5-step prioritized roadmap with facility sizing, lease timing, and greenfield opportunities


Data Foundation

The baseline is only as good as the data

Before any scenario could be run, the underlying shipment data had to be validated, cleansed, and structured. This was the critical first gate — a model that can't reproduce history can't be trusted to predict the future.

Shipment validation

472K raw shipment records processed. Irregular flows, circular routes, misclassified direct/consolidation shipments, and shuttle flows identified and excluded systematically.

96.7% retained

SQL pipeline & data prep

Raw shipment records queried, joined, and restructured in SQL Server and SSMS — cleaning lane classifications, matching consolidation stops, and producing aggregated model inputs across origin, destination, mode, and product dimensions.

4,684 origins

Tableau dashboards

Modeling results linked directly to interactive Tableau dashboards for drill-down by facility, customer, lane, flow type, and direction. Shared with and used by client leadership throughout the engagement.

1,002.5MM lbs. visualized

Demand adjustments

Adjusted baseline accounted for 14 lost customer programs and 2 major customer wins, plus mid-year go-live annualization for 14 partial-year accounts and growth assumptions at a key border facility.

−141M lbs. net

Cost validation

Baseline transportation and facility costs validated against actual 2024 operating data. Modeled total came within 5% of actual across both transportation and facility cost categories.

<5% variance

Aggregation & dimensionality

Origins compressed from 11,419 to 4,684 ZIP-level clusters. Destinations from 2,980 to 1,642. Products defined across four dimensions: source ZIP, country, flow type, and top customer.

10,871 products

Scenario Results

Nine scenarios, one recommended network

Each scenario tested a targeted network change against the adjusted baseline — opening new facilities, closing underutilized ones, or consolidating overlapping locations. Individual scenarios were then combined to find the optimal portfolio of changes.

Total Supply Chain Cost Impact vs. Adjusted Baseline (indexed to 100%)
Adjusted Baseline
0.00%

Add Midwest Hub
−0.61%
Add Northeast Hub
−0.49%
Close Small Border Facility
−0.11%
Add Secondary Midwest Site
−0.25%

Close West Coast Facility
+1.73%
Consolidate Ohio Valley Sites
+0.34–0.41%
Relocate Southeast Hub
+0.02%

Scenario Total Cost Impact Recommendation
Add Midwest Hub −0.61% Proceed
Add Northeast Hub −0.49% Proceed
Close Small Border Facility −0.11% At lease expiry
Add Secondary Midwest Site −0.25% If Midwest Hub insufficient
Close West Coast Facility +1.73% Do not close — downsize to ~13k SF
Consolidate Ohio Valley Sites +0.34–0.41% Keep both — reevaluate on lease
Relocate Southeast Hub +0.02% Only if strategic reason
Recommended 5-step roadmap

Business Impact

From data pipeline to executive recommendation

$2.3M
Annual savings from recommended network changes — closing an underutilized border facility and adding two partner hubs in underserved regions
1.4%
Theoretical maximum total supply chain cost reduction identified across all network optimization levers
<5%
Variance between actual 2024 costs and the historical baseline model — the validation threshold that unlocked scenario analysis
Why the data foundation mattered

A network optimization model is only credible if it can reproduce history. Getting the baseline within 5% of actual costs — across 456K shipment records, 25 facilities, and a complex mix of consolidation, direct, and cross-dock flows spanning the US, Mexico, and Canada — required substantial upfront data work. The SQL pipelines and Tableau dashboards weren't supporting artifacts; they were the mechanism by which the client could interrogate and trust every number the model produced.