Companies are pouring money into dashboards, data platforms, and AI models. But only those who can prove real financial impact secure long-term budget and support for their data analytics projects. That’s why one question needs to be addressed upfront: the ROI. Many ask themselves: How do I actually calculate it? Which KPIs matter? And when does data analytics really pay off? We sat down with Dr.-Ing. Michael Haub, data scientist and engineer at Device Insight, to talk about how ROI in data analytics works in practice.
When looking at the bigger picture, data analytics is less about ROI – and more about a true “Return on Data.”
Once deployed on the shopfloor, a positive ROI from data analytics is typically achievable within 12 to 24 months.
Between 15 and 30 percent scrap reduction and ROI in under 12 months – a real-world outcome from a data analytics project owned by Device Insight.
Technology should never be the starting point. Companies need to begin with a clear business problem and their most pressing pain points.
Facts build trust: every piece of value created through data analytics should be made visible.
Michael Haub: “The ROI of data analytics is the measurable value created when data models and analytics are applied in business processes – minus the necessary investments in technology, talent, and implementation. In plain terms: it’s about saving costs. To get there, you need to bring transparency into processes that were previously guided by gut feeling or spot checks.
To make the value measurable, goals must be clearly articulated from the start of a project, along with the KPIs that will be influenced. Unlike investments in physical assets that typically require long amortization periods, the ROI of data-driven improvements can often be realized faster and with greater flexibility.”
Michael Haub: “With a machine, ROI is usually straightforward: you invest a certain amount, it produces a predictable number of units, and the return can be calculated from the sales margin. Data analytics works differently. The value emerges not from physical output but from improved information that drives smarter decisions.
Another fundamental difference is scalability. Once developed, data models can often be applied to additional processes or departments with relatively little extra effort. Viewed over the lifecycle of a product, the total benefit of all data-driven decisions can far exceed the initial scope. In fact, you could speak of a ‘Return on Data’ – a concept that goes beyond financial ROI to include scalability and the reusability of data models for entirely new use cases.”
Michael Haub: “In the short term, ROI in Data Analytics often reveals itself through very practical efficiency gains – and ultimately through cost savings. The key indicators here are familiar: less scrap on the shopfloor, higher machine availability, reduced downtime, and lower energy consumption. These metrics can usually be tracked quickly and precisely, making them excellent early markers of whether a project is on the right track.
Over the long term, the picture broadens. Incremental improvements accumulate into greater competitiveness, whether through higher degrees of automation, greater process transparency, or the company’s overall digital maturity. In that sense, ROI is not just about immediate savings but about building sustainable advantages – strengthening resilience, agility, and the ability to respond effectively to new challenges.”
Michael Haub: “The expected ROI is something we calculate right at the beginning of a project. Smaller initiatives – such as anomaly detection or targeted efficiency gains – can sometimes pay for themselves within three to six months, provided the data maturity is sufficient and the goals are clear. Once analytics solutions are integrated into the shopfloor, we typically see ROI achieved within 12 to 24 months.
Of course, this requires upfront work: connecting data sources, building a reliable data infrastructure, and training machine learning models. These steps lay the foundation for long-term optimizations. At the same time, not every use case delivers measurable value. In some cases, despite thorough analytics, the business case simply doesn’t justify investment – and then it’s equally important to advise against proceeding.”
Michael Haub: “Absolutely. The best approach is to start with small, manageable use cases that deliver quick wins, and then scale up from there. This allows organizations to build experience while already creating visible results.
Choosing the right partners is critical – partners who not only understand analytics but also the industrial domain and can bring best practices from other projects. Just as important is to involve business stakeholders from the start. They know the processes, the dependencies, and the pain points. Their buy-in is essential, both for prioritizing use cases and for convincing internal decision-makers of the value.”
Get inspired by real project examples. Facing similar challenges and looking to harness the potential of Data & AI? Explore our case stories and insights.
Michael Haub: “One compelling case is a project with a manufacturer struggling with excessive scrap rates. The issue was straightforward but costly: too many defective parts were being identified too late in the process. Working closely with IT, OT, and the production team, we developed a machine learning model for anomaly detection, based on real production data.
Within just a few weeks, we were able to identify patterns that had previously gone unnoticed. Projections now show a reduction in scrap rates of 15 to 30 percent, with ROI expected within nine to twelve months.
The key success factor was the tight collaboration between all stakeholders, combined with rapid visualization of insights for management.”
Michael Haub: “Only to a limited degree. ROI is always tied to the specific use case, and these vary enormously from one industry to another.
In manufacturing, the metrics are clear: machine uptime and scrap rates can be directly measured. In retail, the focus is on conversion rates, customer satisfaction, and loyalty. In the energy sector, forecasting accuracy and grid utilization matter – complex challenges that require sophisticated models, but with a long-term payoff. In healthcare, compliance with strict data protection regulations is often the dominant factor, which extends timelines but can yield tremendous impact when predictive models improve outcomes.
So, while the logic of ROI applies everywhere, the specifics are always unique to the sector and the business.”
Let’s talk about where you are today and where you want to go – in a first, no-strings-attached session.
Michael Haub: “The most important advice is to begin with a clear business problem, not with the technology. Identify the use case that addresses the most urgent pain point, then check whether the right data exists and how it can be consolidated. The clearer the target, the easier it becomes to measure success.
Second, involve the business units early. They understand the challenges best and can quickly recognize opportunities, preventing projects from being detached from operational reality. Even a small, cross-functional team that combines business expertise with IT and data science can make a big difference.
Third, start building a solid data foundation early – this is half the battle.
And finally, include the skeptics. Make sure analytics is seen as a tool for support and transparency, not as a threat.”
Michael Haub: “The answer lies in maintenance and scaling. Analytics solutions must be regularly reviewed, updated, and adapted to changing conditions such as seasonal demand shifts or new production processes. Successful models shouldn’t remain isolated, but rather be rolled out to other parts of the business. Just as crucial is to make the results visible. Dashboards, reports, and success stories help create acceptance internally and showcase achievements externally. As we like to say: facts speak for themselves.”
