From Static Logic to Execution Science
How we replaced flawed volume-based WMS logic with true 3D mathematical optimization to eliminate 184 bad pack instructions and downsize 20% of daily volume.
The Flaw in "Volume + Buffer" Math
For a high-volume 3PL handling premium branded merchandise, accuracy at the pack station is critical. However, their legacy WMS system was creating major bottlenecks because it didn't truly understand physical space.
Like most WMS platforms, it relied on simple fluid volume math. It would add up the cubic volume of the items in an order, add an arbitrary "safety buffer" to guess if they would fit, and output a box size. This led to massive inefficiencies. If the buffer was too small, packers were told to stuff long poster tubes into flat boxes. If the buffer was too large, the system suggested massive boxes filled mostly with air. Every time a physically impossible suggestion hit the floor, the packer had to stop, manually override the system, guess a better box size, and reprint the label.
Auto-Tagging the Physical Reality
To solve this, we bypassed the WMS logic entirely and routed a dataset of 626 historical orders through the TCS-i engine. Because maintaining master data is notoriously difficult, we utilized our Auto-Tag Technology. By programmatically reading the raw SKU descriptions, the engine instantly applied complex execution science to their 13 packaging options:
- Flexibility Logic: Items containing words like 't-shirt' or 'tote' were instantly modeled as foldable, pliable objects.
- Cylinder Math: SKUs matching 'bottle' or 'tube' triggered exact cylindrical nesting algorithms, treating them as round objects rather than generic rectangles.
- Fragility Rules: Hard constraints were automatically placed on items like premium 'flasks', strictly preventing the engine from ever placing them into unpadded bags.
Outcome 1: 100% Mathematically Verified Plans
The TCS-i engine doesn't act as a filter for bad WMS logic; it replaces it with true 3D bin packing. Because our engine mathematically places every item in a virtual 3D space prior to outputting a decision, it guarantees physical fit.
It inherently understands that a 33cm poster tube cannot be angled into a 30cm box, or that combining a rigid keyring with a foldable tote bag exceeds the volumetric stress limits of a small mailer. By generating packing instructions based on actual 3D geometry rather than fluid volume, the engine naturally eliminated 184 physically impossible instructions that the WMS had previously generated. When suggestions are 100% viable, the packer override rate drops to zero.
Outcome 2: Aggressive Downsizing Without Buffers
Because legacy WMS systems don't know how items will actually nest together, they are forced to use those generic safety buffers, leading to oversized packaging. The TCS-i engine eliminates the need for buffers entirely. If the 3D math proves the items fit, the engine selects the smaller box with absolute confidence. In total, 280 orders (over 20% of the dataset) were successfully downsized.
- Box-to-Box Downsize: By properly orienting a water bottle alongside a notebook in 3D space, the engine proved they could shift from oversized 49cm wide boxes into tightly packed 30cm standard boxes.
- Box-to-Bag Downsize: By recognizing that apparel can compress, orders containing notebooks and tote bags were safely downgraded from costly rigid cardboard into inexpensive polybags.
The Bottom Line
This sandbox pilot proved that upgrading from static WMS logic to execution science does two things simultaneously: it reclaims lost margins through material and shipping downsizing, while massively boosting warehouse throughput by eliminating the friction of manual packer overrides.