To optimize a production process, start with the one station that limits output of the whole line. Measure how its time is spent, fix what blocks it, and test bigger changes such as a new layout or another operator in a simulation before you move anything on the floor. If you improve any other station first, you usually get more stock in front of the bottleneck and the same output at the end of the line.
Where does a production process lose time?
Most lines lose time in a few places that repeat every shift: parts waiting in front of a slow station, operators walking to fetch material, changeovers that take longer than planned, and machines that wait because the station before them stopped. You see them on the floor as queues, idle machines and overtime at the end of the week, while a monthly report rarely shows them.
Walk along the line and look for three signs:
- Work in progress piles up in front of one station and is thin after it.
- A later station often waits for parts while an earlier one is busy.
- Some machines run all shift while others stand idle for long stretches.
Where all three point to the same place, you have a first candidate for the bottleneck, which the data then has to confirm.
How do you confirm the bottleneck in a production process?
The bottleneck is the station with the least capacity compared with what the line needs. As a result, it sets the pace for everything after it. You confirm it with three numbers per station: cycle time, available time and how long it is busy.
Illustrative example: a line has four stations with cycle times of 50, 62, 48 and 55 seconds per part. Station 2, at 62 seconds, can make about 58 parts an hour (3,600 / 62). Even if the other three stations run perfectly, the line cannot go faster than that. Speeding up station 4 from 55 to 45 seconds changes nothing at the end of the line.
However, real lines are messier than this example. Cycle times vary by product, a station may be the bottleneck only on one product family, and short stops can make a fast station act like a slow one. That is why you need data over at least a few days, split by product and shift.
The DBR77 Digital Twin knowledge base covers this in more detail in how to find production bottlenecks before they happen.
What data do you need before you change anything?
You do not need a perfect data set to start. You need enough to describe how material moves and how long each step takes.
| Data | Why you need it | Where it usually comes from |
|---|---|---|
| Cycle time per station and product | Shows the capacity of each station | Time studies, machine counters, MES |
| Changeover times | Shows how much capacity product changes take | Shift logs, setup sheets |
| Stops and their reasons | Shows how often a station stops and why | Machine signals, stop logs |
| Transport times and routes | Shows how long parts travel between stations and stores | Layout drawings, observation, forklift or AGV data |
| Buffer sizes | Shows how much stock can sit between stations | Layout, racks and conveyors on the floor |
| Shift pattern and staffing | Shows available time and who serves which station | Shift plans, HR planning |
First, collect it for one line. If machine data is not available, manual time studies are enough for a first model. The DBR77 Digital Twin knowledge base lists what to prepare in what data you need to start a factory simulation. For stop data straight from the machines, see why machine data needs downtime reasons and a state model in the DBR77 IIoT knowledge base.
Why test changes in a simulation first?
Some fixes are small and cheap: a clear stop reason list, a better changeover routine, a material rack closer to the bottleneck. So do those directly on the floor. Others cost money and are hard to undo: moving machines, adding a buffer store, changing the number of operators, buying a robot. Those are worth testing first.
A simulation model of the line lets you compare several variants with the same data and see which one gives the most output per day. Two DBR77 Digital Twin projects show what that looks like, and in the first one the winning layout did not have the shortest flow time:
- In a production line design project, three layouts were compared. The best one placed a buffer warehouse between production and assembly. Its average material flow time was not the shortest, but daily productivity was 6% higher than in the linear layout without an intermediate warehouse. The analysis of all variants took one week [1].
- In a layout optimization project at a wooden furniture maker, the new layout removed a buffer. The case study reports working capital lower by PLN 300,000, departmental costs lower by 14% and a need for 1.5 fewer full-time positions in direct labor [2].
Test your next layout or staffing change before you move a machine. See DBR77 Digital Twin on your scenario: the team prepares a session on a layout and process close to yours.
How do you optimize a production process step by step?
- Pick one line and one goal, for example more parts per shift or fewer overtime hours.
- Find the bottleneck with cycle times, stop data and a walk along the line.
- Fix the cheap problems at the bottleneck first: stops, slow changeovers, material waiting.
- Build a model of the line with the data from the table and check that it reproduces today’s output.
- Test two to four variants in the model, for example a different buffer size, another operator at the bottleneck or a new station sequence.
- Choose one variant, implement it in stages and compare the line’s output with the model.
If you want a fuller method for step 5, see how to compare layout variants without guesswork in the DBR77 Digital Twin knowledge base.
What makes changes stick on the shop floor?
A better layout on paper does not help if the line goes back to the old way after a month. Operators need to understand why the change was made. For example, show them the before and after numbers from the model and from the line. People who run the bottleneck every day often know about problems no data set records, so ask them before you build the model, not after.
Then measure the result for several weeks. Output in the first week after a change is often lower while people learn the new routine. Compare like with like: the same products, the same shift pattern.
Also keep one owner. Someone in the plant needs to own the line’s numbers and decide what to change next. When the first bottleneck is gone, the next one appears somewhere else in the production process.
For the wider picture of losses on a single machine, read why OEE is low and how to find the causes on our blog.
FAQ
What is a bottleneck in production?
A bottleneck is the station or resource with the least capacity relative to demand. It limits the output of the whole line, so an hour lost at the bottleneck is an hour lost for the line. An hour saved at any other station often only adds stock in front of it.
Do I need a digital twin to optimize a production line?
No. In many cases, improvements such as shorter changeovers or fewer stops at the bottleneck need only good data and a team on the line. A simulation model pays off when you compare expensive or hard-to-reverse options, such as a new layout, a buffer store or a robot.
How long does a simulation project take?
It depends on the size of the line and how ready the data is. In the production line design case on our site, comparing three layouts took one week [1]. In fact, collecting data for a first line often takes longer than building the model itself.
Can I optimize a production process without machine data?
Yes. Manual time studies, shift logs and changeover sheets are enough for a first analysis and a first model. Still, machine data makes the picture more precise, especially for short stops, so it is worth adding later.
Ready to test your next change?
Compare layouts, buffers and staffing before you commit budget. Book a DBR77 Digital Twin demo and see a scenario close to yours, with the KPIs you care about.
