This article is for plant managers, production managers, continuous improvement and methods engineers in high-mix, low-volume plants who are preparing a business case for a cobot, automated loading or a more flexible cell. It explains why the first step is not choosing the equipment but measuring, on the current machines, where the available hours actually go: running time, changeovers, short stops, waiting and breakdowns. Without that breakdown, the automation project risks solving a loss that is not the main one.
- Map the process, then measure the hours at each machine.
- Changeovers, short stops and waiting call for different projects.
- Pick the automation pilot on the loss that really dominates.
- Measure before and after with the same rules and instrument.
Understand the current cell before automating it
Maryland MEP, the Manufacturing Extension Partnership center that supports manufacturers across Maryland, published Automating Small Batch Production: Process Innovation Strategies for Maryland’s Specialty Manufacturers. The article describes the pressures specialty manufacturers face, the technologies that suit low volumes (collaborative robots, flexible manufacturing systems, smart sensors and analytics) and a set of practical steps. The first of those steps is process mapping, introduced with a clear principle: “You must fully understand your current operations before you can automate them.”
The article then recommends starting with a single pilot, such as automated material handling, proving the return, and training the existing workforce to program and maintain the new equipment. It also points to the cost of switching between custom products and to the downtime that changeovers create. This article does not repeat that method. It adds one practical layer to the mapping step: a measured time baseline at the machine, so that the pilot is chosen on the loss that really dominates, and its result can be checked afterwards on the same basis.
Why a process map is not enough on its own
A process map shows the sequence of steps, the handoffs and the queues. It rarely shows how many hours of a shift each machine spends cutting, waiting or being set up. In small-batch production, that distribution is precisely what decides whether automation will pay off, and it is where perception and reality diverge most.
- Changeovers are underestimated. Teams usually remember the standard setup time, not the full window from the last good part to the first good part of the next batch, including program loading, first-part checks and adjustments.
- Short stops are invisible. A machine that pauses for 20 seconds to clear chips, wait for a part or reset a door interlock never appears in a manual log. Repeated dozens of times a shift, these stops can weigh as much as one long breakdown.
- Waiting is spread across causes. Waiting for material, for an operator who runs two machines, for a program, a fixture or a quality release is often recorded as “idle” or not recorded at all.
- Averages hide the mix. In a high-mix cell, one product family can carry most of the setup time while another runs almost unattended. A plant-level figure does not show this.
Each of these losses calls for a different answer. A cobot that loads parts addresses waiting for an operator. Automated material handling addresses waiting for parts. Neither shortens a changeover that is mostly fixture adjustment and first-part approval, and neither removes short stops caused by chip evacuation or part presentation. That is why the hours need to be measured before the solution is chosen.
A five-step time baseline on the current machines
- Pick the machines the project is meant to change. Usually the constraint of the cell, plus the machine that feeds it. Measuring a non-constraint machine can show a large gain that never reaches shipped output.
- Define the time categories with the team. Running, planned changeover, unplanned stop, short stop, waiting (split by cause: material, operator, program or fixture, quality release), planned downtime. Write down where each category begins and ends, for example that a changeover runs from the last good part to the first good part.
- Measure automatically over several weeks. Cover the full product mix and every shift. Two or three days often catch an unusual week, and a week without the longest setups gives a misleading picture.
- Qualify the stops at the machine. Operators assign a reason to each stop when it happens, from a short list. The duration comes from the measurement; the cause comes from the people who saw it.
- Convert the result into hours per week and per product family. Hours speak to the finance team and to the automation supplier better than percentages, and they make the expected gain of each option directly comparable.
The output is a simple table: for each machine, how many hours per week go to each category, with the OEE that results. That table becomes the first page of the automation business case.
Reading the baseline: which loss calls for which project
The table below links the dominant loss to the type of project it usually justifies. It is a reading guide, not a rule: the right answer still depends on the parts, the volumes and the team.
| Dominant loss in the baseline | What it usually means | Project it tends to justify | What automation alone will not fix |
|---|---|---|---|
| Changeovers (long or frequent) | Setup, program loading and first-part approval consume the hours | SMED work, standard fixtures, offline programming, then automation | A robot does not shorten a manual fixture adjustment |
| Waiting for an operator | Machines stop at the end of a cycle until someone is available | Cobot or automated loading, staggered cycles | Waiting caused by missing material or programs |
| Waiting for material | Parts, blanks or tools do not arrive on time | Material handling, kitting, supermarket at the cell | Unstable schedules upstream |
| Repeated short stops | Chips, part presentation, sensors, interlocks | Root-cause work on the process, robustness improvements | Automating a cell that stops every few minutes adds a robot that waits too |
| Breakdowns | Reliability of the existing machine | Maintenance plan, retrofit of critical components | Adding automation to an unreliable machine multiplies the stops |
For example, imagine a five-axis machine in a specialty cell that is available 80 hours a week. Suppose the baseline shows 38 hours running, 18 hours of changeovers, 12 hours of waiting (eight for an operator, four for material), 8 hours of short stops and 4 hours of breakdowns. These figures are invented for illustration. In that case, a loading cobot targets at most eight hours a week, and only if the operator is the real cause of the wait, while changeovers represent more than twice that. The baseline does not say “do not automate”; it says what to fix first, and how many hours the chosen project can realistically return.
Keep the same measurement for the pilot and after
The Maryland MEP article recommends starting small with a pilot and expanding once it shows a solid return. That return can only be shown if the before and after are measured the same way: same machines, same time categories, same rules for what counts as running, waiting or changeover. If the baseline was estimated by hand and the post-project figure comes from the new cell’s controller, the comparison mixes two instruments and the gain becomes a matter of opinion.
A measured baseline also protects the project after go-live. In small-batch production, a new cell often moves the loss rather than removing it: the robot is fast, but it now waits for fixtures, programs or first-part approval. Keeping the same measurement running shows where the hours went after the change, and gives the team the next problem to work on.
Measuring the hours on existing machines without a heavy project
Many specialty plants run machines of different ages and brands, some without an accessible data interface, which is one reason the baseline is often skipped. TEEPTRAK measures OEE, machine stops and short stops in real time with a module wired in the electrical cabinet and a sensor at the edge of the line, without connecting to the PLC. Installation takes about an hour per machine, and OEE is available within 48 hours. Every stop is timestamped automatically, operators qualify the reasons at the machine, and the hours can be read by machine, shift and product. TEEPTRAK is not a MES: it measures what the machines do, which is the input an automation decision needs. It is used in more than 700 plants in over 30 countries, where it delivers on average 10 to 15 OEE points of improvement and 30 percent fewer unplanned stops.
Further reading
- Maryland MEP, Automating Small Batch Production: Process Innovation Strategies for Maryland’s Specialty Manufacturers
- Changeover frequency and batch size: when shorter changeovers really release capacity
- Machine retrofit: measure OEE before and after to quantify the gain
FAQ
How long should we measure before deciding on automation? Long enough to cover the full product mix and every shift, which in small-batch production usually means several weeks. A short measurement can miss the longest changeovers or an unusual week and distort the business case.
Is OEE meaningful in a high-mix, low-volume cell? Yes, provided the ideal cycle time is set per product and changeovers are recorded as their own category. The breakdown of lost hours is often more useful than the OEE percentage itself, because it shows which loss an investment should target.
Can a cobot fix changeover losses? Rarely on its own. A cobot mainly removes waiting for an operator during loading and unloading. Changeover time depends on fixtures, programs and first-part approval, which usually need SMED and standardization work before or alongside automation.
0 Comments