Having clear objectives for data-driven change

When considering a data-driven change, the first question that should be asked is what the end goal is for the project.

Data acquisition insights

  • Effective data-driven decision-making requires a meticulous approach—starting from understanding available data to thorough validation, cleaning, processing and clear communication to stakeholders.
  • Data, when aligned with specific objectives, empowers manufacturers to optimize working capital, reduce waste, cut costs, and adapt to market changes, fostering informed decision-making and competitiveness.

As processors and manufacturers face a number of challenges, including inflation driven material/energy price rises and meeting sustainability targets, the data organizations have access to can be used to positively impact the business.

To ensure data driven insights are accurate, however, a bottom-up approach needs to be taken – firstly working with data owners and users to understand what data is available, then focusing on the data that is needed. Not all data collected will necessarily be relevant to the problem that needs to be solved.

When gathering data, it needs to be comprehensive and relevant to the objectives. For example, an objective of improving the inventory position might require analysis of inventory, forecast, sales and production data. Undertaken successfully, this would enable food manufacturers to optimize working capital, reduce waste and improve service levels.

Alternatively, a specific objective may be to reduce delivery costs, where analysis of customer order and delivery data will highlight those customers that drive higher costs. With each objective data the requirements will vary.

Once the relevant data is collected, there are important steps to be taken.

  1. Validate the data with those familiar with the processes and activities the data describes – Does it feel right to them, has the data been impacted by short comings or a lack of process adherence.

  2. Understand the context – In what context was the data captured, what biases are likely to be present in the data. When presenting the insight, setting it within a context is important. Like a book or a movie, data only makes sense once the context is known and understood and looked at from multiple angles.

  3. Cleaning – Not all data is clean. It may require some work to correct inconsistencies, remove duplicates and handle missing values. Understand the scale of these issues and make clear to stakeholders how these issues have been addressed.

  4. Explore the data and get comfortable with it – Get an initial sense of the data’s distribution, patterns, and potential outliers. Visualizations like histograms, scatter plots, and box plots are useful tools – make sure that everyone understands what they are looking at, and don’t make it too complicated.

  5. Processing – The data might not be at the required level for analysis, it might require aggregation or the addition of more data to provide a richer picture of specific products, locations, and customers.

Once the data is processed, it is important to revisit what needs to be achieved. Are there specific hypotheses that need to be tested. If so, use statistical techniques such as regression analysis, hypothesis testing, or clustering, depending on the goals. Look for hidden patterns or even build predictive models using machine learning (ML) techniques to provide insight and move towards the overall organizational goals. Interpreting the results will help draw actionable insights to inform decisions, formulate strategies, or take action.

Finally, it’s no good having great insight if it isn’t communicated clearly to stakeholders and decisions makers. There are many tools available that create powerful visualizations. Use these tools, but think about the audience, and how are they going to consume the visuals.

Data plays a vital role in helping manufacturers ensure product quality, meet regulatory requirements, manage costs, adapt to changing market conditions and deliver products that meet consumer expectations. It is a valuable resource for making informed decisions and staying competitive today, but having clear objectives, collecting, validating and processing the relevant data is key.

– This originally appeared on Control Engineering Europe. Edited by Chris Vavra, web content manager, CFE Media and Technology, [email protected].

Written by

Paul Derbyshire

Paul Derbyshire is senior manager at BearingPoint.