Why Data-Driven Operator Capacity Planning is Better

Why is the manufacturing industry still so reliant on process managers’ gut feelings?

Although “93% of companies believe that AI will be pivotal in driving growth and innovation in the manufacturing sector”, actual adoption of data science tools faces resistance, according to Deloitte.

This is especially noticeable when it comes to operator capacity planning, where process managers and HR draw up the budget for the following year. Much of the process is data-driven, apart from one crucial thing: Planning the number of plant operators. Here, the process reverts to the unscientific: Going with the process manager’s gut.

Why is such a time-consuming and error-prone approach still so widely used across the industry? Are manufacturers simply resigned to mistakes happening? Are managers concerned that new methods won’t produce the expected results?

When weighing up the pros and cons of adopting a new approach, manufacturers typically evaluate the relative advantage of the innovation compared to the competition and how easily it can be implemented. If they determine that their company is profitable the way it is, why change a winning team?

However, the effects of the COVID-19 pandemic are making it clear that manufacturers need to think about how to manage operations more efficiently and with greater agility.

According to the Deloitte report, there is still a widespread belief among manufacturers that technology cannot be trusted — even though intuition produces errors. On experience-led staffing models, planners are often overly cautious, and fears that supply won’t meet demand leads to overmanning of production lines.

Without a data-driven process in place to validate gut feelings, people tend to assume their intuition was right. This means the process manager and HR will continue to have their annual meetings, operator capacity planning will continue to be time-consuming, and production lines will continue to be over- or under-manned. The upshot: The plant will continue to unnecessarily lose money.

Using the plant’s data to build an operator capacity plan won’t just produce “a” plan, but the best plan. When predictive analytics are applied to the data, planners get the insight to optimize operator capacities ahead of events, bringing real benefits in time- and cost-savings.

Implementing machine learning models to analyze and perform predictive analytics on the specific data in the plant enables managers not just to optimize operator capacity planning but also:

• Minimize production line bottlenecks with predictive employee attrition

• Improve the balance between capacity and demand with product demand forecasting

• Reduce unplanned production line downtime with predictive maintenance

Using a modern data science tool makes operator capacity planning agile. What was previously planned annually can now be planned, reviewed, and optimized in good time, whenever changes in the plant data indicate optimization is necessary.

This is also where the experience of the plant’s process manager is better put to use. Using an easily-learned low-code, data science tool, the process manager can fold in their expertise, monitor, and optimize that employee attrition forecast — no coding required. Boosting planners’ skills to interpret the data and draw insights also further up-values their expertise.

The CEO of a Czech carmaker suffered headaches every September: It was time for the annual capacity planning sessions.

The problem: Sometimes there weren’t enough operators on the line; sometimes managers realized they had too many in the headcount and had to send them home. Plus, the plant was on the border with Germany, meaning it often faced labor shortages with workers increasingly preferring to earn more abroad.

The source of the headache: Planning of the approx. 1500 operators at the Czech plant was done manually, based on experience.

Seeing the inherent unreliability in an experience-led staffing model, the CEO took critical action: To be able to make reasonable data-driven decisions about operator capacity, a data science team was brought in.

A cross-departmental team identified the features that influence the number of operators, e.g.:

• Net sales / car type

• Maintenance-related downtime

• Sickness rates

• Employee turnover rates

The team took those features as well as numbers of operators for each month going back over a three-year period and fed them into their model.

“We were able to include more parameters in comparison to the former planning method. We enriched our model by additionally integrating forecast sales for each car type, expected sickness rates, and operator efficiency (how good operators are at their task). This made our model not only more robust but also more accurate in comparison to the previous approach,” commented a leading data scientist on the team.

Using KNIME Analytics Platform, they built a machine learning model that would predict the required number of operators for the year. The model was trained to make predictions based on a subset of the plant’s data and then measured for accuracy during tests. The test predictor returned an accuracy score (R²) score of 92%, already indicating high accuracy.

Next, these predictions were validated against four months of real operator numbers, planned values of net sales, and a number of other features. After applying the newly built model they were able to report a prediction accuracy of over 96%!

This data-driven solution now enables the company to plan operator capacity effectively. Intuition is a thing of the past. Instead, the data-driven solution enables the plant to compare quickly between the plan and reality and make accurate predictions. The process manager’s time is freed up to spend more meaningfully injecting expertise into the forecast where necessary. Planning and decision making are now based on data and ensure there are no big gaps between the actual and required number of operators.

Resistance to innovation is an operational bottleneck. But this can be overcome by equipping teams with a tool that is easy to learn. KNIME, as a low code/no code tool, means they “don’t need to spend half a year learning how to code before finally moving on to the data science”.

The additional benefit of KNIME is that it’s open source. Teams can download it and start evaluating how useful it can be for your company straightaway – no upfront investment required.

It’s time to rethink and see if a data-driven planning process doesn’t make more sense. Should something as important as the operator capacity planning of your production line be left entirely to intuition?