# From raw material to finished good: mapping the plant floor
A single steel coil, roughly the size of a truck tire and weighing several tons, arrives at a bearing plant. Six weeks later, part of it leaves as a pallet of finished ball bearings bound for a wind turbine gearbox. What happens in between (every cut, heat treat, grind, and wait) determines whether that bearing costs a little or a lot, and whether it ships on time.
Let's walk one bearing through the floor.
Every part has a routing: the ordered list of operations it must pass through, from raw material to shipping. Think of it as a recipe plus a mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition →.
For our bearing ring, a simplified routing looks like this:
1. Uncoil and blank (cut discs from the steel coil)
2. Turn (machine the ring shape on a lathe)
4. Grind (finish surfaces to tight tolerance)
5. Assemble (add balls, cage, seals)
6. Inspect
7. Pack and palletize
Each step happens at a workstation: a machine or cell staffed by an operator. The routing is not just a sequence. It encodes decisions. Do we grind in-house or send rings to an outside shop? Do we heat treat in batches of 500 or 2,000? Those choices move cost and lead time, as we will see.
At each workstation, the key number is cycle time: how long one unit takes to be processed. If the turning lathe takes 45 seconds per ring, that station's cycle time is 45 seconds.
The station with the longest cycle time is the bottleneck. It sets the pace for the whole line. You cannot ship faster than your slowest station allows, no matter how fast the others run.
Concrete example: if turning runs at 45 seconds but grinding runs at 90 seconds per ring, grinding is the bottleneck. Buying a second, faster lathe does nothing for output. The rings just pile up in front of the grinder.
This is the single most common mistake non-operators make: they optimize the wrong station.
Between workstations, parts wait. That waiting inventory is called WIP (work in process): material that has entered the factory but is not yet a finished, shippable good.
Why do buffers exist? Because stations do not run in perfect lockstep. Heat treating runs in batches (you load a furnace, cook for hours, unload). Grinding runs one ring at a time. Between them sits a rack of hardened rings waiting their turn. That rack is a WIP buffer.
Buffers are useful and expensive at the same time.
A famous relationship called Little's Law ties these together. In plain terms:
> Lead time = WIP divided by throughput.
If you have 3,000 rings in process (WIP) and you finish 500 per day (throughput), the average ring spends 6 days on the floor. Want to cut lead time? Cut WIP or raise throughput. You can read a clean explanation of Little's Law on Wikipedia.
This is why "less inventory" and "faster delivery" are often the same project.
Here is the counterintuitive part. On many plant floors, a part spends the vast majority of its total time waiting, not being worked on. Actual machining might take minutes. Total time on the floor might be days or weeks.
That gap between processing time and total lead time is the biggest lever most factories have. You do not shorten it by running machines faster. You shorten it by reducing waiting: smaller batches, fewer handoffs, better scheduling.
Batch size is one of the most powerful and least understood dials on the floor.
Say heat treating a furnace load takes 4 hours regardless of quantity. Run 2,000 rings per load, and the per-ring furnace time looks tiny. Efficient, right?
Not so fast. Big batches create three problems:
1. Long queues: grinding cannot start on a batch until the whole batch finishes heat treat. A ring waits for its 1,999 neighbors.
2. Big WIP: 2,000 hardened rings is a lot of tied-up cash sitting on a rack.
3. Slow quality feedback: if a heat treat setting is wrong, you may not discover it until grinding inspects the batch. Now 2,000 rings are suspect, not 200.
Smaller batches cut WIP and lead time and catch defects faster, but they mean more furnace setups and changeovers. The art is finding the batch size that balances setup cost against carrying cost and speed. Techniques like SMED (Single-Minute Exchange of Die: methods to make machine changeovers dramatically faster) exist precisely so that smaller batches become affordable.
Back to the routing. Two decisions there reshape the whole floor.
Make versus buy. If the plant sends rings to an outside heat treater, it avoids owning a furnace but adds transport time (parts leave the building and come back). That outside step might add days of lead time and a truck's worth of WIP in transit. Keeping heat treat in-house costs capital but tightens the loop.
Alternate routings. Suppose there are two grinders: an old slow one and a new fast one. A scheduler may route urgent orders to the fast grinder and standard orders to the slow one. That flexibility is valuable, but it also adds complexity: now two paths must be balanced, and the "mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition →" has a fork in it.
Every fork is a decision that a planner or a scheduling system has to make, thousands of times a day.
Knowledge check
1. What does a part's 'routing' fundamentally represent on the plant floor?
2. A line has stations with cycle times of 45s, 30s, 90s, and 60s per unit. Which determines the pace of the entire line?
3. If grinding (90s/ring) is the bottleneck and turning (45s/ring) is not, why does buying a second, faster lathe fail to increase output?
4. Select ALL correct answers about what the routing encodes beyond a simple sequence of steps.
Select all the correct answers.
5. Select ALL correct answers about cycle time and bottlenecks.
Select all the correct answers.
Practitioners capture all of this in a single diagram called a value stream map (VSM): a picture of every step from raw material to shipping, annotated with cycle times, batch sizes, WIP counts, and wait times.
A VSM makes the invisible visible. It typically shows a shocking result: a part that takes minutes to process may spend days on the floor. The mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition → draws processing time and waiting time as two separate lines, and the waiting line is almost always far longer.
You do not need software to start. Teams often build the first VSM by walking the floor with a pencil, a stopwatch, and sticky notes, following one real part from dock to dock. That walk (sometimes called "going to the gemba," a Japanese term meaning "the actual place") often reveals buffers and delays that no report captured.
For a solid free introduction, the Lean Enterprise Institute glossary defines value stream mapping and related terms in plain language.
Let's total up our ring, using illustrative (not actual) numbers to show the pattern:
Add the processing: a few minutes of actual work per ring. Add the waiting: days. The lesson repeats: the cost of the steel and machining is real, but the lead time is dominated by queues and batches, and lead time is what customers feel.
Improve the floor and you attack two things at once: shrink the buffers (free up cash) and speed the flow (win the order).