How Leading Manufacturers Use Data to Resolve Production Bottlenecks
Production process bottlenecks are costly and impede throughput, which is why manufacturers consistently aim to identify and eliminate them. Insights from line managers and work centers can inform these efforts, but to effectively spot and continually monitor bottlenecks, manufacturers need a reliable system to visualize channel production data. This article covers how leading manufacturers use a modern data platform and business intelligence tools to spot and monitor bottlenecks in the production process.
How to Build a System to Monitor Production Bottlenecks
Start with a simple goal.
It is imperative to start with a simple goal. For example, the initial goal might be to develop an automated dashboard that allows managers to see actual and theoretical throughput capacity. This goal is not overly broad, and notice that you are not initially focused on the outcomes of improving throughput or eliminating bottlenecks. While those are the ultimate goals, you need to first set a reasonable goal that allows you to start leveraging data for an automated view into production capacity.
Find and collect the pertinent data.
With an initial goal in mind, consider the available data pertinent to your efforts. You can expect this to vary considerably based on how the organization is currently capturing production-related data. Some manufacturers have shop floor data collection systems to consistently capture data based on the production activities on the floor. Others have data captured through their ERP, whether it is manually input or automatically ingested through an integration with a shop floor data collection system.
Think about the systems you have in place and ensure you have a basic understanding of where the key data resides. Once you have that list, talk with line managers and other knowledge workers to understand any limitations. Key questions to ask: how complete is the data, what is the longitudinal coverage, and are there any critical data points not currently being captured that are essential to understanding capacity and throughput?
Bring all of the necessary data into a data lakehouse.
With an understanding of where the key data resides, you are ready to consider how to deliver insights. At this stage, it is critical to design a dashboard solution that is as automated as possible. By automating the process of channeling raw data into curated insights, organizations preserve time for analysis, decision-making, and action. These are the fundamental ways that value is created through data initiatives. Commonly, time is wasted through relying on a manual process to refresh reports with the most recent data.
A performant, automated business intelligence system rests on three pillars: a modern data platform to house raw data, data modeling to build in metric definitions, and a dashboard platform like Power BI to serve insights. The data platform is of particular importance. This could be a data warehouse or a more modern data lakehouse. Think of it as a landing spot for raw data, with data pipelines connected to each source system: shop floor collection system, ERP, CRM, and so on. These pipelines allow the most recent data to automatically flow into the data lakehouse on a regular schedule.
Crucially, the data platform also allows businesses to combine data from disconnected systems. Consider a scenario where you need to pull data from two separate systems to get an accurate view of throughput. That is what the data platform provides. Through data modeling, you can codify specific metric definitions that take the most recent raw data and convert it into insights. With all three pillars in place, companies can implement a process where no time is spent preparing and serving insights. The BI system presents insights each day, allowing the business to focus on what the data is saying and how to make adjustments to improve key metrics.
Determine and define the metrics that will be used to measure success.
With a data platform in place and all the raw data available, businesses must determine the logic for specific metric definitions. Chances are these exist already, but if not, time will be required to talk with subject matter experts and confirm how key metrics should be defined. Once this is complete, these definitions are coded into the data models, which serve to update dashboards with the most recent data.
Returning to the goal of visualizing throughput, you need to determine an accurate theoretical throughput capacity for each machine to compare against actuals, and verify that the definition of throughput you will use is correct. A simple definition is:
Throughput = Total number of acceptable units produced / specified time frame (hour, shift, day, etc.)
Be sure to discuss your specific definitions and account for any nuances based on how your company measures throughput. You may also want to consider additional metrics beyond throughput. Work in process, downtime, overall equipment efficiency (OEE), total effective equipment performance (TEEP), capacity utilization, and first-time right (FTR) are all examples of metrics that can help pinpoint where bottlenecks are impeding production. That said, do not try to boil the ocean. Your initial goal is to provide automated reporting showing actual and theoretical throughput capacity. Bringing in additional metrics requires more time and risks overwhelming users. Start small, limit your first dashboard to one to three metrics, and expand as you collect feedback from users.
Build automated reports to help you track against goals.
At this stage, you should have the data platform set up, pipelines connected to each source system, and data models with key metric definitions developed in the data warehouse or lakehouse. With these prerequisites in place, you arrive at the crucial last mile of a data initiative: determining how to visualize the data for comprehension, clarity, and action.
There are many best practices for developing dashboards. The most important ones are to start small and not overwhelm users with too many metrics on one page, to provide metric definitions within the dashboard so everyone understands how the numbers are determined, and to go from big to small by highlighting key metrics at the top followed by more detailed views. This allows users to keep the main goal in mind while they look at more granular data. The dashboard should be built around the problem your team is trying to solve rather than serving as a passive report. You can find many more design principles in the Dashboard Design to Ensure Adoption guide. The overarching idea is to keep the team aligned around the goal, limit noise and increase signal, and ensure that everyone understands how metrics are being measured so there is one version of the truth.
Monitor adoption and iterate continuously.
Now that you have an automated dashboard, take time to collect feedback and practice using this new tool to improve business results. There are various ways to monitor usage statistics for dashboards, but informal feedback is equally critical. After releasing a new dashboard, talk with users to understand how they are using it, what they like and do not like about the layout, and what they want to see included.
Developing business intelligence is not a one-time effort. It is a continual process of analyzing how the business is leveraging available data to focus teams on the key metrics that determine profitable growth. Priorities and key initiatives change, and sometimes metric definitions change. As priorities shift, expect the specific needs for data to shift as well. The most valuable manufacturing KPIs are often the ones that get refined over time as the organization matures in its use of data.
Guiding Principles for Data Projects
Data is a fundamental element in improving processes, whether it be spotting and monitoring bottlenecks or improving customer loyalty. Companies usually have the data available but are not properly leveraging it to drive improvements. The principles that consistently produce results are the same across project types: start small and prove out a use case before expanding, invest in a proper data platform to provide a foundation for analytics, remember that dashboards are built for humans and should not overwhelm or distract from what is most important, and treat dashboards as living entities by measuring adoption and collecting feedback that drives iteration. A solid feedback loop with users is what turns a technically correct dashboard into one that actually changes how the business operates.
If you would like to explore how Blue Margin can help your manufacturing operation build the data foundation to monitor and resolve production bottlenecks, contact our team here.
