Your company's unique processes are likely what give you a competitive advantage in the marketplace. However, just because your processes are unique or different does not mean that they cannot be measured, benchmarked or engineered. We hear this statement repeatedly when working with potential customers. The real issue comes down to the cost of measurement.
The traditional approach in the warehousing industry has been to use engineered labor standards when developing production metrics and standards for employees. For the right type of organization, this can be an excellent approach, but for the majority of organizations, it is an expensive challenge to make it work. For those unfamiliar with engineered standards, these are production metrics developed by an industrial engineering (IE) approach. The IE maps in detail each step of a process, product locations, conducts time and motion studies, and industry standard comparisons. Up to 24 input variables can be used to model each specific process. If a process is changed, then the process and its input variables need to be re-calibrated for the change. For large (over 100 warehouse employees), consistent organizations such as food and grocery facilities, this approach can work well. But for small or more dynamic/seasonal operations, the cost of developing and maintaining the standards often exceeds the benefit received from them.
We have developed an approach that we call Employee Analytics that greatly broadens the application of performance standards both in industry type and organizational size, and from empirical experience, achieves the same or better results than engineered standards. Our average customer has increased productivity by over 40% within the first 7 months.
Our approach is relatively straight forward. When you look at the output data from any warehouse management system (WMS), or any basic data model of a warehouse process, you will have three to five variables around each process. For example, for devanning you may have # cases, # SKU's, case weight, and case cube. For picking, you may have # cases, # scans, # locations, case weight, etc. We have found 2 or 3 of the input variables affect 80-90% of the labor required to do that process. So, for devanning, #cases and # SKU's generally have the most impact on that process. For picking, it is scans, cases and locations. From your WMS or through manual tracking, you can collect the data for each process and the time it took to perform each process for each employee. At a minimum, you should look at 100 data points. The more the better. Then, using standard regression tools in a spreadsheet or database, you can build a performance model for each process where:
Expected Time = AX + BY + CZ + D (A,B, C and D are the calculated coefficients from your analysis and X, Y and Z are the actual inputs)
For example, let's look at the process of devanning a container where the two main input variables are cases (X) = 2500 and SKU's (Y) = 25. The container is floor stacked and offloaded by hand onto carts. The coefficients are in minutes A = 0.05 ; B = 5.00 ; C = 0 ; D = 15 where D is the fixed setup time per container. Based on these number then, the Expected Time = 2500*.05 + 5*25 + 15 = 265 minutes or 4 hrs 25 minutes. Now, let's assume a two person team completed the project in 2 hours for a total of 4 labor hours. The performance standard on the process is then 110%.
The key thing to understand with benchmarking is that the accuracy of your standards is not nearly as important as the consistency of your approach. You are not comparing your processes to hypothetical industry or engineered standards. Instead, you are benchmarking each employee against the historical average of all employees on specific processes and the relative performance of current employees. From this performance data, you can see who is performing well and who is not performing and then take appropriate action. You can also build accurate labor forecasting models, budgets, pay for performance programs, etc.
This approach is also very flexible and inexpensive. It generally takes less than a month to model all your processes and calibrate the performance standards with an effort of approximately 4 hours per week. Additionally, the system is highly adaptive so when a process is changed, the model for that process can be quickly modified to take into account the change.