A major food manufacturing company approached us with a deceptively simple question:
“How many operators are really required to run this production line?”
Here’s our take:
Traditionally, the only way to measure operator efficiency was through manual time studies – brief snapshots of performance often tainted by observer bias or the “Hawthorne Effect,” where operators work differently because they know they are being watched. This creates a massive operational blind spot.
Without a way of measuring operator productive vs. idle time continuously and objectively, you are either overstaffing to manage “just-in-case” scenarios or understaffing and wondering why your OEE is plateauing.
In this article we’ll explore the benefits of using camera AI to enhance human operations in a food manufacturing environment. Where production efficiency is critical to business success, AI brings hidden wastages, stoppages and losses to the table in a world where they’ve never been quantified before.
Why manufacturers can’t measure operator efficiency
The business wanted to optimize staffing levels on a high-throughput packaging line. This meant looking at ways to reduce operator numbers without sacrificing output or quality and creating promising cost savings.
The problem? Traditional manufacturing data sources could not provide a confident answer.
Here’s some common industry challenges they experienced:
- PLC data could show line speed and stoppages, but not why they happened.
- Manual time studies captured only brief snapshots and were prone to observer bias.
- Audits and sampling offered limited visibility into real operator behaviour over full shifts.
In short, there was no continuous, objective view of how people interacted with the line in real-world conditions.
This created a blind spot: the organization could not determine true operator performance ceilings, how long peak performance could be sustained, or which losses were caused by people versus process constraints.
How camera AI measures and improves operator efficiency
The solution was clear: the manufacturer was missing out on key operator analytics which they could derive by plugging AI into their existing CCTV system.
Using camera AI allowed them to generate continuous, unbiased data from existing production cameras. Instead of relying on assumptions or short-term observations, the system would watch the line continuously, classify key objects and operator states, and convert visual information into operational metrics.
The goal was not “surveillance” – it was insight. By quantifying operator activity and line conditions in detail, it’s possible for manufacturers to expose the hidden factors that are not otherwise seen through manual inspection or IoT.
How to improve operator efficiency metrics in four steps
The food manufacturer needed a guide on what to prioritise, how to measure KPIs and how to report on outcomes. We define success by this 4-step process:
1. Define project objectives
From the outset, it’s important to understand what’s important to your business and project. There are four guiding principles which can be used to define your objectives:
- Speed: Deploy rapidly using existing infrastructure, generating actionable insight in days and validated conclusions within weeks.
- Trust: Deliver objective, repeatable analytics validated against the manufacturer’s own operational metrics.
- Scalability: Prove a solution that could extend across multiple lines, shifts, and sites with minimal reconfiguration.
- Flexibility: Ensure adaptability to different products, tasks, and line layouts without long redevelopment cycles.
These principles ensure that the initiative would deliver immediate value while laying the foundation for broader transformation.
2. Methodology: Measuringreal performance in real conditions
In order to analyse operator activity directly on the production line, a streamlined methodology is required to ensure the environment is controlled and results can be accurately derived. The analytical framework focuses on three critical dimensions:
- Maximum operator efficiency
Measuring the highest achievable productivity under real operating conditions – not idealised time-study assumptions. - Sustained peak performance
Determining how long operators can maintain peak output before fatigue, interruptions, or process limitations reduce performance. - Root cause analysis
Identifying and quantifying sources of performance loss, such as:
- Operators present but unable to work due to empty or full trays
- Waiting caused by downstream congestion
- Non-productive movement or interruptions
These are factors that traditional systems cannot reliably detect, yet they directly determine throughput.
By correlating operator actions, object states, and line conditions, it’s possible to separate people-driven inefficiencies from process-driven constraints – a crucial distinction when making staffing or capital investment decisions.
Deployment timeline: From rapid start to deep insight
Like any project, it’s critical to lay out milestones that correlate with the objectives that have been set. In the case of this specific food manufacturer, the project was structured in progressive milestones:
Week 1 – rapid start
Key objects and states (such as empty trays, full trays, and scale interactions) were classified. Visual data began streaming automatically, with minimal disruption to operations.
Week 4 – validated insight
Sufficient data had been collected across multiple shifts and operators to establish:
- Peak and average cycle times
- Sustainable performance durations
- Early root-cause indicators
Importantly, the visual AI outputs were validated against the manufacturer’s internal metrics, ensuring alignment and credibility.
Week 12 – deep operational insight
The system delivered:
- Quantified root-cause data linking operator behaviour and line conditions
- Operator counts over time
- Frequency and duration of performance losses
This provided evidence-based input to staffing models and process improvement recommendations.
Outcomes: from assumption to evidence
By the end of the initial phase, the manufacturer gained:
- Objective data on true operator performance ceilings
- Clear understanding of sustainable output levels
- Quantified impact of line constraints and interruptions
- Evidence to answer the original question:
“What number of operators are required to service this line at any given time?”
Crucially, the data also revealed where performance losses were driven by process limitations rather than staffing – guiding targeted investments in equipment or workflow redesign instead of unnecessary labour changes.
A scalable foundation for continuous improvement
Beyond the initial use case, the system established a scalable platform for:
- Tracking operator efficiency trends by shift or product
- Monitoring constraint frequency and severity
- Expanding to additional lines and facilities
- Integrating data directly into existing production reporting systems
The result was not just a one-time study, but a continuous performance intelligence capability embedded in daily operations.
Conclusion: Making the invisible visible
This project demonstrated how computer vision can close a critical data gap in manufacturing – revealing the real interaction between people, equipment, and process constraints. By replacing assumptions with objective evidence, the manufacturer gained confidence in staffing decisions, uncovered hidden inefficiencies, and built a foundation for long-term operational excellence.
For visionAI, it was a clear example of transforming existing camera infrastructure into a strategic performance engine – delivering trusted insight where traditional systems fall short.