Managing director Riasha Naidu sees it as a significant change in how Unistar is positioning itself for the future, with the processes behind the business becoming more data-driven, connected and predictive.
Central to this approach is the use of artificial intelligence (AI) and machine learning, using more than two decades of technical and customer data to strengthen quality control, accelerate troubleshooting and improve responsiveness to converters.
Over the past 18 months, Unistar Inks has developed an in-house machine-learning system that draws on its spectrometry and production data to identify patterns, correlate variables and speed up problem-solving. The system is used across the company’s quality-control processes and is also being deployed at converter sites.
Instead of treating every print defect as a new problem, Unistar can interrogate years of correlated data to establish which variables are likely to have caused it and what corrective action is most appropriate.
‘The system doesn’t simply flag a colour as out of specification,’ explains Riasha. ‘It uses historical production and spectrophotometric data to help identify why. If a cyan falls outside the customer’s Delta E tolerance, the model analyses similar deviations and recommends the most likely corrective action.’
The aim is to reduce troubleshooting time. Less time is spent working through assumptions, fewer variables are overlooked and problems can be resolved faster, which is particularly important because downtime has a direct cost.
The system is also being adapted to account for the realities of South African printing. Unlike many European and US operations, local converters typically do not operate in tightly controlled production environments. Temperature and other environmental variables can influence ink behaviour and print quality. A temperature change, for example, could influence the solvent blend required during ink dilution, prompting an adjustment that may prevent a print defect before it occurs.
‘Our model is learning from real South African production conditions, where temperature, humidity, press technology, substrates and operating environments can differ significantly from more controlled international environments,’ says Riasha. ‘Over time, that could create a predictive model specifically informed by the realities of flexographic and gravure printing in our market.’
The company says the technology is intended to support, not replace technical expertise. ‘AI is only as good as the person who’s driving it,’ says Riasha. ‘You still need to apply your own human intelligence in order to get out of it what you need.’
The company’s technical and finance teams remain essential because someone still needs to interpret the information, challenge the outputs and make business decisions.
The same philosophy is shaping the company’s customer offering. Having initially deployed the system internally, Unistar has extended its application to converter sites, where the objective is to understand the wider set of variables affecting print performance, from inks and raw materials to plates, rollers and environmental conditions.
One example illustrates the value of looking beyond the obvious. When Unistar encountered an unexpected change in ink performance, the usual formulation parameters – pigment, resin and solvent specifications – were unchanged. By analysing the wider production data, however, the system identified a raw-material substitution from an alternative supplier as a potential contributing factor.
‘Sometimes we become so focused on the immediate problem that we overlook changes elsewhere in the process,’ says Riasha. ‘AI allows us to interrogate those variables collectively and identify relationships that may not be immediately obvious.’
That ability to correlate multiple variables could also improve repeatability and colour consistency. By capturing the conditions behind successful print results and identifying deviations, Unistar can build a more predictable approach to colour across production runs, presses and converter sites.
We’re now using that knowledge to anticipate outcomes and develop solutions tailored to our customers’ needs.’
The longer-term ambition is to move from correcting colour deviations to predicting them. As the dataset grows, the system could identify when a colour is likely to move outside a customer’s Delta E tolerance, allowing adjustments to be made before the deviation occurs.
‘The objective is not simply to identify why a colour has deviated, but eventually to predict when a deviation is likely to occur,’ says Riasha. ‘That would allow adjustments to be made before the colour moves outside the customer’s Delta E tolerance.’
Unistar’s technology ambitions extend beyond quality control. A next phase, planned for 2027, will explore using AI and machine learning to forecast cash flow, identify anomalies in expenditure and predict customer payment behaviour and collection risks.
From data to development
The ‘2.0’ approach is not confined to internal processes. It is also shaping product development. Unistar has developed more affordable security-ink solutions for flexible packaging, targeting counterfeiting without requiring expensive specialist detection equipment. The approach uses a UV-detectable solution that can be checked with relatively inexpensive UV lights, making the technology more accessible to retailers and the broader market.
The company has also revived work on RFID and conductive inks, first pioneered a decade ago by Riasha’s late father, Dasson Pillay, Unistar’s founder and self-taught scientist.
With the South African market now showing greater readiness for RFID technology, Unistar is commercially manufacturing a silver conductive ink locally for RFID applications.
Riasha says these developments are part of the same strategy: using technology to solve practical problems.
‘After more than 20 years in the South African market, we’ve built up a wealth of local knowledge,’ she says. ‘We’re now using that knowledge to anticipate outcomes and develop solutions tailored to our customers’ needs.’
