Most confectionery manufacturers are collecting increasing amounts of data. Modern lines capture everything from tempering profiles to checkweigher readings. The difference between those who benefit and those who don’t comes down to how well that data is used.
That was the message from our Director of FMCG, Dr Stuart Gilby, when he spoke to Kennedy’s Confection for its July feature on data-driven manufacturing. Here’s a summary of his thinking.
Start with what you can control
Stuart’s view is that data only earns its keep when it’s linked to a clear outcome and to a lever you can actually pull. As he put it, it needs to be tied to an outcome, but also to something you can control.
Take customer returns. Knowing which batches came back is useful, but it’s only the first step. To reduce returns, you need traceability back to the production run, an understanding of why products failed, and a clear link to the process changes that would fix it.
Capture the full picture
Targets closer to the line, like uptime or scrap rate, are easier to measure in real time. But the detail matters. Are ambient conditions recorded? Do you know which mould or position a reject came from? When an operator changed a setpoint, do you know what prompted it? Without that context, it’s hard to tell which changes made a real difference.
Use the right tools in the right places
Statistical process control helps keep production consistent. Machine learning can spot patterns people might miss, such as seasonal humidity affecting sugar crystallisation. Used together, they help pinpoint what’s really driving quality.
Bring operators with you
Insights are only useful if people act on them. Use clear, real-time visuals and simple decision rules for frontline teams, backed by training that explains the reasoning. If the data seems to contradict experience on the line, it won’t be trusted.
Four steps to get started
Stuart’s advice for manufacturers is to set up good data governance early, start with a pilot on one line before scaling, bring in data science expertise that also upskills your own people, and build one team where operators, data specialists and R&D share the same language.
As published by Kennedy’s Confection (July 26 issue)