From Stopwatch to Continuous Insight
Background
At its Singapore facility, a global medical technology manufacturer produces high-precision components for life-sustaining medical devices, including the fine tubing used in cardiovascular and heart surgery. The parts are costly and the tolerances are unforgiving, so every second of cycle time shows up on the cost sheet. Yet on one of its most mature lines, the industrial engineering team was still timing the work the way factories have for a hundred years: with a stopwatch and a clipboard.
The challenge
A stopwatch study only ever catches a sliver of reality. An engineer stands at one station for a short window, and that handful of readings then has to stand in for weeks of production. Nothing gets recorded to go back to, there is no easy way to hold a slow day up against a fast one, and two engineers timing the same job rarely write down the same number.
On a mature line, that gap matters most. The obvious efficiency gains here were banked years ago. What is left are small, stubborn pockets of lost time buried inside the averages. Digging them out, and building a case strong enough to justify a line-balancing or process change, needs better data than a stopwatch can ever give you.
What PowerArena did
PowerArena started small, with a single camera on a three-station work cell, and grew from there. The Human Operation Platform (HOP) pairs AI vision with the line’s own machine signals: IoT triggers from the equipment mark the exact start and end of each work cycle, so the timing is read straight from the line rather than estimated. The result is a cycle-accurate time study that runs on its own. Every cycle, every operator, around the clock, and nobody has to hold a stopwatch.
- Continuous time study: cycle and step-level timing is captured 24/7 and presented as charts the engineering team reviews on a regular cadence.
- Rewind any cycle: when a day shows an abnormally long cycle or step, the team replays the exact video of that moment to find the root cause instead of guessing.
- Automated alerts: email notifications flag anomalies, so issues surface quickly instead of waiting for the next manual study.
- Built to scale: multi-camera, multi-stream inference and per-line model fine-tuning let the same approach extend from one cell to an entire line.
Before and after
| Before HOP: the stopwatch era | After HOP: continuous AI time study |
| Manual stopwatch studies, one station and one observer at a time | Every station and operator measured automatically, in parallel |
| A short sampled window: minutes of data standing in for weeks of production | 24/7 capture of every production cycle, continuously |
| Subjective readings that shift with whoever holds the stopwatch | Objective, cycle-accurate data triggered directly by machine IoT signals |
| No historical record to revisit or audit | Any abnormal cycle can be replayed on video to see the root cause |
| Bottlenecks on a mature line stay hidden in the averages | Cycle and step-level charts surface the few seconds that still matter |
HOP turns time study from an occasional manual exercise into an always-on record of how the line really runs.
The results
The best proof is what the team did next. The one-camera pilot turned into a multi-station rollout, then jumped to a second and busier line, with more cameras bought along the way. Reviewing HOP charts is now part of the weekly routine, the team pulls up video whenever a cycle runs long, and they have started setting up and running studies themselves. They have even asked to load their own data into the dashboards. Once a customer wants to put more of their world inside your tool, you have stopped being a pilot and become the source of truth.
The day-to-day change is just as telling. Instead of a sampled guess, engineers now work from continuous, objective numbers. On a line this precise, that move from “we think” to “we can see” is what every line-balancing call now stands on.
Why HOP, versus other smart-factory approaches
| Capability | MES & counters | PLC / sensor-only | Generic camera analytics | HOP |
| Counts output and cycles | ✓ | ✓ | ◐ | ✓ |
| Cycle-accurate timing from the line | ✗ | ✓ | ✗ | ✓ |
| Sees operator motion | ✗ | ✗ | ✓ | ✓ |
| Explains why a cycle ran long | ✗ | ✗ | ◐ | ✓ |
| Replay any cycle on video | ✗ | ✗ | ✓ | ✓ |
| Tuned for a precision medical line | ◐ | ◐ | ✗ | ✓ |
Only HOP combines IoT-defined cycle timing with vision, so it both counts the cycle and explains why it ran long.
The honest trade-off: HOP is not a sensor you bolt on in an afternoon. It needs cameras, IoT integration, and a model tuned to each line, and its value compounds as fine-tuning matures and the team adopts self-service. For a mature line where the remaining gains are small but real, that investment is what makes the gains visible in the first place.
What’s next
With continuous timing now running in the background, the next step is to hand over more of the controls. The team is tuning models and setting up new lines on its own, HOP is spreading to more work cells, and its data is becoming the starting point for line-balancing and continuous-improvement work. What started as one camera on one cell is turning into the normal way this floor gets run.
Download Full Case Study
Interested in applying AI Vision for Manufacturing?


