warehouse insider
The problem starts when a facility handles a high number of SKUs and fulfils high-volume mixed outbound orders which is more common in food distribution, automotive parts, and general merchandise. Due to the high SKU count, even pallets that are stored at higher levels of the rack storage are used as pick locations. As a result, a significant fraction of the pallet locations in the facility are partial pallets, and it becomes important to accurately track the remaining number of cases remaining on any given pallet.
Counting partial pallets is fundamentally harder than counting full pallets for three reasons:
● With a full pallet, and a solid receiving process, a cycle count process may only require verifying the Pallet ID: no interaction needed with the case count (because the incoming case count on a full pallet is already known). But if the case count has changed for these pallets, it’s no longer accurate to just track the Pallet ID.
● At upper rack levels, a human counter needs a lift to get close enough to count accurately.
● The count changes every time a picker pulls cases, which in active warehouses can happen continuously throughout the shift.
Partial pallet counting is one of the most labor-intensive, error-prone tasks in cycle counting, and it becomes increasingly difficult to manage as warehouse operations scale.
Vimaan’s StorTRACK approaches partial pallet counting using AI & computer vision, which can cleverly combine 2D visual information with depth mapping. StorTRACK is able to recreate the entire inventory digitally and measure how many cases have been removed from a full pallet. Here’s how:
Step 1: Scanning the Pallet Face to Create a 3D Map
As StorTRACK moves along the aisle, it does not just take flat images of rack faces. It builds a spatial depth map: a precise map of every point on the pallet surface relative to StorTRACK’s position.
A complete, untouched pallet appears as a flat, uniform plane in the depth map, with consistent depth across the entire face, representing an unbroken wall of cases. The width and height of this plane become the reference point against which everything is measured in the steps that follow.
Step 2: Detecting Removed Layers
When a full layer of cases is removed from the top of a pallet, the pallet face is still a uniform plane, just shorter. StorTRACK compares the current pallet height against the original case count baseline stored in the WMS and calculates how many complete layers are missing. Since it knows the case structure, i.e. cases per layer multiplied by the number of layers, it calculates remaining cases by applying the same logic to what’s still on the pallet.
For more on how case count baselines are established in warehouse operations, see Ti-Hi Pallet Counting Automation.
Step 3: Detecting Partially Picked Layers
When a picker removes individual cases from a layer rather than a whole layer, the depth map becomes uneven across the pallet width or height. The positions where cases have been removed show increased depth, while remaining cases stay at the original depth. StorTRACK reads this variation across the pallet face and counts how many case positions are missing from any given layer, then calculates what is left.
Step 4: Combining the Count
StorTRACK adds cases in the complete layers still present to the remaining cases in any partially picked layer to arrive at a total case count for the pallet. This count is compared automatically against the WMS record, and any discrepancy is flagged immediately with photographic evidence for the inventory team to review.
StorTRACK’s partial pallet counting delivers accurate results when the following four conditions are met:
1. Single-deep storage: StorTRACK counts cases on pallets visible from the aisle. Pallets in position 2 or beyond cannot be visually identified or counted, although most cycle counting operations typically don’t count behind position 1 either.
2. Known original case count: A baseline is needed for accurate counting, typically from WMS data or a count established at receiving. In some cases, this can also be inferred from the pallet’s Ti-Hi structure.
3. Single SKU, uniform case dimensions: Mixed SKU pallets or varying case sizes complicate the depth map inference.
4. Front-pick SOP compliance: Pickers must pull from the aisle-facing side. Back-picking disrupts the depth map logic.
Real warehouses do not always behave as expected, and StorTRACK accounts for this:
● Partial shrink wrap remaining over picked areas can distort the depth map. These locations can be configured to be flagged for manual review or ignored per warehouse policies and procedures.
● SOP violations (picking from the back) create irregular depth patterns that do not match the expected model. These are flagged automatically to an Inventory team
● Very small cases relative to pallet size can push against the limits of dimensional measurement accuracy. While this is rarely seen in warehouses, it can also be flagged for review.
In all of these scenarios, StorTRACK surfaces the exception with an image rather than estimating a result it cannot confirm. The inventory operator can often resolve it visually on a tablet. Genuinely ambiguous cases require a physical bin visit. For most warehouses, partial pallets represent a disproportionate share of cycle counting effort. StorTRACK does not eliminate that complexity altogether, but significantly reduces the manual effort involved, concentrating on the cases that genuinely need a human in the loop.
For a deeper technical breakdown of how StorTRACK’s depth mapping works and the criteria your operation needs to meet, read the full application note.
To see how StorTRACK handles partial pallet counting and other cycle counting use cases in practice, refer to our Cycle Count Automation Kit that includes demo videos, real-world case studies and technical documentation.
Q. How does computer vision improve inventory accuracy for partial pallets?
Computer vision builds a depth map of the pallet face and calculates remaining cases by measuring what has been removed, rather than trying to directly count every visible case. This works consistently across rack levels without requiring a human counter.
Q. How does automated partial pallet counting reduce labor dependency?
Manual partial pallet counting requires a dedicated counter, often with a lift for upper rack levels, to physically verify every location. Automated solutions using computer vision like Vimaan’s StorTRACK handles the count during a standard forklift mission, with a single operator covering significantly more ground. Human effort is only needed for the exceptions the system flags, not for every partial pallet in the warehouse.
Q. What is the role of the original case count baseline in automated partial pallet counting?
The baseline tells the system how many cases should be on a full pallet, which is what partial counts are calculated against. If this data is accurate in the WMS, the system can calculate remaining cases precisely. If it is missing or outdated, a baseline count needs to be established at the time of receiving before automated counting can work reliably.
Q. How does automated cycle counting eliminate manual inventory scanning for partial pallets?
Instead of sending a counter with a handheld scanner to each partial pallet location, the system captures depth map data during a forklift mission and calculates the case count automatically. Manual scanning is only required for the small percentage of locations the system flags as exceptions it cannot resolve.
Q. What makes partial pallets harder to count than full pallets?
Full pallets have a known, fixed case count tied to the Pallet ID in the WMS. Partial pallets have an unknown and changing quantity that must be physically verified each time. Add upper rack positions and hidden rear cases to that, and it becomes one of the most time-consuming and error-prone tasks in a manual cycle counting program.