Enhance Your Goods To Person Operations With Artificial Intelligence

1. Equipment: Know Your Throughput Limitations

AutoStore provides flexibility for multiple Goods to Person order fulfillment models, pick station designs, and operational activity at each port. Each port has a mechanical maximum bin presentation capacity, as well an operational capacity. For example, a carousel port is capable of performing 400 bin presentations per hour (Table 1), while a less expensive conveyor port may only process half that rate. Knowing how these mechanical rates are capped by different operational processes is critical to understanding the building’s total throughput capacity.

Typical functions supported by AutoStore include:

• Decanting
• Inventory Control
• Empty Bin Retrieval
• Bin Consolidation
• Picking

Table 1: Hypothetical Bin Presentation Capacities by Function (Carousel Port: 400bph)

When different processes are performed in parallel, each with different throughput rates, the challenge evolves into a multi-variable function to determine the overall impact to operations.

Typical ways to measure and capturing throughput include:

• Labor Management System (LMS)
• Business Intelligence (BI) applications
• Shared directory files
• Automated email alerts from various systems
• Excel

These reports are almost always historical and not forward looking.

Best Practice
Management should understand the sub-system operational rates of a single process before attempting to forecast hourly throughput when ports/stations are performing multiple functions. Table 1 shows port utilization (a proxy for throughput) is impacted by various functions and their process time.

The Power of Artificial Intelligence
Artificial intelligence platforms baseline, analyze and identify patterns in the collected operational data. By capturing the total sub-system throughput capacities for any given combination of station processes (for example: 3 ports picking, 3 ports decanting), the CognitOps platform understands the complex interrelationships of profiles, functions, and priorities to estimate cycle times. When processes start exceeding these standard cycle times, Operations is notified that order SLAs may be at risk.

2. Goods To Person Orders: Execute The Right Work At The Right Time

Once the functional throughputs have been baselined, Operations can make informed adjustments to activity timings that support the goal of maximum order SLA achievement of on-time and in-full.  Executing the right work at the right time has the largest impact on overall order cycle time.

Picking is usually the most critical task, but also the least predictable, as order volume, priorities, and profiles may be unknown until a late-stage allocation is complete. Operational managers must plan for variable demand against both fixed machine throughputs and static cut-off times.

Typically, there are 2-3 major milestones within an operational 24-hr day:

1. When orders become available (driven either by allocation logic or internal approval)
2. Last order acceptance time for same-day processing (if applicable)
3. Ship cutoff time

Operations may delay releasing orders to leverage batching benefits and reduce duplicate travel, but condensing the pick window risks missing service levels and excess overtime. 

Best Practice

Operational managers should develop an operational clock for the building, working backwards from the ship cutoff, mapping out upstream / downstream processes, expected activity windows within AutoStore, and cutoff times with sufficient transportation buffers to create an operational clock for the building. Leveraging this operational clock will provide guidance on labor planning, fulfillment status, and if the building is ahead or behind schedule.

Graphic 1: Example high-level 24-hr operating schedule

The Power of Artificial Intelligence

CognitOps can dynamically forecast and plan the operating clock of the building, directing resources across all areas from receiving through shipping at the right times, with the right sized teams.  Leadership will spend less time analyzing spreadsheets and reports, and productivity will increase based on better inventory availability and work release logic. 

Graphic 2: Example detailed 24-hr operating schedule

3. Inventory: Position Inventory to the Right Fulfillment Engine

Orders typically do not take the same path through the building to arrive on an outbound truck.  With the exception of some unit-level only eCommerce Companies, most distribution centers  are designed with multiple fulfillment “engines” to process outbound orders. 

Slotting attempts to identify the correct pick processing area(s) for an item based on its attributes (seasonality, conveyability, etc.) and other operational trade-offs (redundancy/availability vs. total locations per SKU; reduced number of replen vs. days on hand/inventory turn strategy; dedicated areas for faster SKUs to minimize travel vs. congestion/bottlenecks).  All these trade-offs can benefit from the computational muscle of AI/ML systems. 

Often, operators will push for maximum utilization of the AutoStore at go-live with A-velocity SKUs to maximize ROI, which can actually reduce the largest density gains and create bottlenecks. Careful analysis of bin utilization and pick station capacity should guide the slotting of SKUs to the AutoStore, typically not the highest velocity SKUs.

Best Practice

To maximize the efficiency of the AutoStore, operations managers need to keep a close eye on the performance of their AutoStore, including:

• The velocity and inventory levels of slotted SKUs
• Overall bin utilization
• Appropriate Decant to Pick Volumes

The Power of Artificial Intelligence
Instead of monitoring and reacting, AI-driven forecasting can predict when demand will exceed AutoStore inventory, make recommendations to move fulfillment to an alternate engine, or prioritize replenishment. CognitOps connects ERP, WMS, WCS/WES, and Labor Management Systems to identify which SKUs are driving the most exceptions; calculate the complex decanting, bin utilization, and picking gain trade-offs in real-time to make recommendations on whether/when slot to other fulfillment engines.

4. People: Select The Right Person For The Right Job

Goods-to-person operations have now matched inventory to the appropriate storage media and understand when to execute the various AutoStore processes throughout the daily facility schedule based on the operational throughput capacities of the sub-system. With everything in place for flawless execution, leaders must now ensure resources are allocated to the correct departments to fully recognize the benefits of their upstream optimizations.

Most Companies will defer to their Labor Management Systems (LMS) when evaluating the correct candidate for a job. Knowing an Associate can pick at 110% is great, but as shown in Table 1, it is possible to out pick a machine, which causes more worker idle time and is an ineffective use of labor.

Best Practice
Labor moves are costly. Constantly readjusting staffing based on 5-minute demand shifts results in lower productivity, throughput and employee morale. Most operations build a fairly static staffing plan, and do not adjust it regularly.

Artificial Intelligence and Your Goods-To-Person Operation
AI-based applications can be configured to optimize based on the Company’s philosophy around labor staffing.

1. Should I cut my shift early based on achieving target or pull volume forward based on forecast demand?
2. Based on today’s profile, how much volume should be left in each of the processing stages to allow the next shift to effectively startup without downtime?
3. Assuming I cannot let people leave early, what is the recommended staffing across my primary departments to achieve the remainder of today’s SLA?

5. Repeat Success Intentionally

The best operations must continue to learn from changing order patterns, labor productivity, and equipment effectiveness. The recent COVID pandemic has taught us that assumptions and plans will change, and operators must be able to adapt, and also recognize when a similar volumes and profiles occurred previously.

The complete historical memory of the CognitOps AI platform immediately recognizes changing patterns, identifies historical relationships, and recommends operational changes faster than operators can on their own.

Conclusion

Advanced automation has enabled companies to sustain their aggressive growth by allowing supply chains to deliver exacting customer expectations. Achieving the next levels of growth will require effective and holistic workflow management vs. individually optimized solutions.

By connecting multiple independent data silos, applying constantly trained AI algorithms, and guiding decision making, CognitOps puts facility operations in the best position to achieve or exceed daily goals. With the CognitOps Warehouse Operating System, the decision-making aperture is widened, adjustments can be made proactively, and total facility performance can be optimized.