Overcoming bottlenecks: the key to successful job shop scheduling

    Elmar Karlowitsch
    February 8, 2024

    The theory of constraints (TOC) by Eliyahu Goldratt is a management philosophy that focuses on identifying and addressing the most limiting factor (constraint) in a system to improve overall efficiency and productivity. It emphasizes optimizing the entire workflow by systematically managing and alleviating constraints rather than sub-optimizing individual components.

    Bottlenecks are such limiting factors in the system. They increase throughput time and therefore reduce productivity. Identifying bottlenecks in a high-mix low-volume production environment can be quite challenging due to the dynamic and agile nature of the production process. In this article, I will delve into the process of identifying bottlenecks and implementing effective strategies to overcome them.


    Experience has shown that – with good reason! – throughput and/or on-time delivery are the core scheduling goals of job shops or any high-mix low-volume manufacturing companies in over 95% of cases. Accordingly, it makes sense to keep the throughput time of the jobs as short as possible in the context of target improvement. To achieve this, what determines the throughput time must first be recognized. In other words, the bottleneck or the bottlenecks in the current situation must be identified.

    Before I explain how bottlenecks can be identified, it makes sense to first clarify what exactly is meant by a bottleneck:

    A bottleneck is any production unit whose capacity is equal to or less than the demand at a given point in time.”

    As you can see, it is all about supply and demand. If the demand for capacity at a certain time is higher than the given supply of demanded capacity at that time, then a bottleneck occurs. In many essays on the subject, the temporal aspect is neglected. In my opinion, this is fatal in the HMLV case. Precisely because both the supply (finite capacities) and the demand (order backlog) at HMLV manufacturers fluctuate greatly, bottlenecks have the unpleasant habit of jumping from one area to another over time.

    Identify bottlenecks through Gantt chart visualization

    Now let's get back to the practical side of things: How can you determine if your company has bottlenecks and pinpoint their exact locations? Just like with Automatic Finite Scheduling (AFS), the schedule can be visualized through Gantt chart views, making identification a breeze. By delving deeper into the data using two separate views, you can gain a comprehensive understanding of the schedule's situation:

    • In the resource-centric Gantt chart view look for time windows in which operations on the assigned resources are lined up one after another (almost) without idle times.

      Identify bottleneck in a HMLV production: resource group view
      This resource group view reveals a clear lining up of operations on the Edge Polishing machine, which will result in a delay for PO 14.

      Bottleneck resource group view delay

    • In the job-centric view, you can see where there are gaps in the throughput of the individual job routings. Then look at the resource of the operation after the gap (when forward scheduling).

      Bottleneck jobs view

    As you can see here in the jobs view, it becomes apparent that there is a significant time gap between two operations in PO 14. Upon further scrutiny of the routings for PO 14, it becomes evident that the Edge Polishing operation does not immediately follow the CNC operation but instead has a substantial time interval between them. This observation serves as a clue that the Edge Polishing resource is operating at maximum capacity.

    Bottleneck jobs view edge polishing detailed


    If corresponding candidates have been sighted in the schedule, then it is worth taking a second, more detailed look before actions are taken prematurely. In my view, the phenomenon of "damming up" of operations should not be equated per se with an emergency. First of all, this stands for high utilization of the corresponding resource (which some SMBs even strive for as an overall goal). It is, therefore, necessary to determine whether and what effects damming has. If the waiting times within a stage "before a resource" become too long, then this is generally reflected in delays or (if the due date is determined residually) in long throughput times (= high WIP). Some AFS also visualize waiting times, such as with a specific color scheme and the associated link from the predecessor to the waiting operation. Such functionality is very useful and helps immensely in spotting bottlenecks.

    Strategies to overcome bottlenecks

    A bottleneck identified in this way should then in the sense of the TOC be used to align targeted measures as follows.

    Increasing the utilization of the bottleneck

    • In HMLV, bottlenecks are unfortunately not always 100% utilized. This is fatal because an unused minute of a bottleneck affects ALL jobs (that have to go through the bottleneck) negatively. Or to put it another way: If the bottleneck is used better, then ALL jobs (that have to go through the bottleneck) will be positively influenced. Better use of bottlenecks can be made, for example, by cleverly prioritizing jobs.
    • In this context, it may also be worthwhile to place similar operations in the bottleneck one after the other to save set-up times and increase the proportion of productive time, if possible.

    Increasing the bottleneck capacity

    • Depending on what kind of resource represents the bottleneck in the time window under consideration, even short-term peaks can sometimes be compensated for. If, for example, manpower is scarce, temporary workers can be a solution to close the gap in capacity requirements. In this case, it makes sense in the medium term to also think about training internal employees in the required skill areas.
    • When it comes to machines, it is worth investigating carefully whether an existing – less busy – machine can also take over bottleneck operations. Do not shy away from high conversion costs here, but consider that every additional bottleneck minute will increase your throughput and thus your turnover. Just ask yourself if the increased costs of using the machine are higher than the additional sales in the long run or vice versa. In this context, the option of a new investment should also be mentioned. However, I expressly warn against hasty decisions. In my career, I have experienced relatively few cases in which a bottleneck remained constant and lasting and could not be adequately tackled by the other methods. Because only then does an investment make sense in the long term.
    • Last but not least, know that bottleneck operations can also be outsourced. In this case, it must first be ensured that the supplier can also guarantee the desired quality. Also, the same question needs to be answered: Are the increased costs of outsourcing lower than the additional sales in the long run or vice versa? Experience has shown that this is very often the case and outsourcing is a very good and effective method of tackling a bottleneck and "earning more money."

    According to the TOC, it is of course necessary to check after each measure taken to overcome a bottleneck whether it still exists or whether another area has become a bottleneck. In the latter case, analogous measures must then be initiated, and so on.

    Accepting bottlenecks

    The measures explained above all relate to tackling and trying to overcome a bottleneck. As already said, that also makes 100% sense with HMLV scheduling. However, one should not be under the misconception that bottlenecks can always be eliminated. It is then much more important to keep the given bottlenecks small, to use the corresponding capacity in a particularly responsible manner, and to align one's actions accordingly with the bottleneck. What does that mean in detail? Let´s take the example that there is one particular bottleneck work center:

    • Jobs can then be separated into bottleneck and non-bottleneck jobs that do not need any capacity from the respective work center. When doing forward and priority-based scheduling then all non-bottleneck jobs should get higher priority than all the other bottleneck jobs.
    • When scheduling a new bottleneck job, then derive a realistic due date. Based on this, finetune the WIP by productively setting the start date of operations before the due date to tighten the throughput time.
    • When there is the question of whether a new order should be accepted or not, a bottleneck-related key figure can be used. In this course, the margin of the order should be divided by the time taken in the bottleneck. The higher this quotient, the more attractive the order, because the most limiting factor is taken as the basis to maximize the resulting earnings. If it succeeds (and market demand allows), build an order backlog that maximizes this metric, which will also maximize its goal to "make money."
    • Last but not least: Forget about low utilization in non-bottleneck areas. I see again and again that this instruction causes much "pain." But what is certain is that doing anything else (other than activating the particular resource in the bottleneck area) is a worse alternative in the short run.

    Further readings

    Recommended book 

    New call-to-action

    For over two decades, Elmar Karlowitsch has dedicated his passion and expertise to assisting job shop owners in achieving greater efficiency and profitability. He is widely recognized as the author of a highly regarded production scheduling guide on the topic of Automatic Finite Scheduling.

    Blog posts 

    User roles, hierarchies, and responsibilities in job shop scheduling

    3 red flags when Excel is no longer a good fit for production scheduling

    Effective scheduling in an HMLV production environment-the jpi concept

    Subscribe by Email

    No Comments Yet

    Let us know what you think