Investing in robots requires a thorough calculation of Return on Investment (ROI). In practice, however, organizations tend to rely only on the most obvious parameters, which means that key benefits often go unnoticed. That’s why Fizyr, with input from its partners, has mapped out in detail the factors that determine the real ROI of robotics.
The initiative stemmed from a sense of puzzlement, says Tibor van Melsem Kocsis, CCO of Fizyr, a developer of deep-learning vision software for automated picking and placing. “We work with integrators who offer our solution to end customers. Occasionally, we notice that these solutions are perceived as expensive. We wanted to understand why the ROI turns out negative in such cases, but we never received a consistent explanation. So, we started systematically collecting data from integrators and experts to understand how they approach ROI calculations.”
The next step was to gather feedback from the actual users of robotic solutions. “We started speaking with the end customers of our integrators—not to sell to them directly, but to understand why they see the solution as expensive or why they don’t perceive sufficient value. That was our starting point. And after a few of these conversations, it became clear: ROI calculations often fall short because companies base them on an overly simplistic formula.”
Incomplete Picture
He explains: “The ROI is often calculated based on the upfront investment and the number of FTEs that can be reduced. That saving is subtracted from the investment cost, using an arbitrarily chosen payback period of two or three years. But this doesn’t tell the whole story. End users also face seasonal work, peak demand, employee turnover, human error, labor availability, and a lack of flexibility to scale quickly. They must also comply with increasingly strict ergonomic regulations, such as lifting restrictions on heavy boxes.”
Many of these factors are excluded from typical ROI models. “But if you consider what a company really pays to do all of this with human labor—not to mention the difficulty of finding staff—it adds up. We wanted to make that clear in our research so the conversation could change. Even our partners admitted they weren’t sure what exactly should be measured.”
Cross-Functional Blind Spots
Van Melsem Kocsis stresses this is not due to incompetence. “People are often aware of these aspects. But because many organizations are structured in silos, they assume it’s someone else’s responsibility. The innovation team looks at different things than operations, and finance—who ultimately approves the investment—has yet another perspective.”
“There is some change happening, but in many companies the final decision is still made based on purely financial metrics. That’s not wrong, but the calculation is incomplete. That’s why we added extra parameters to those financial models. You get very different results. The goal is not a 100% accurate calculation—some factors are difficult to quantify—but rather to create better insight.”
Labor: More Than Just Headcount
Labor, for instance, has multiple dimensions. “Robots are especially valuable when work is repetitive, unpleasant, dirty, or dangerous. Some companies struggle to find workers and face increasingly stringent regulations. If a company has a seasonal peak from May to August and needs to hire 30% more staff during that period, it impacts their cost structure. These temporary workers need training, make more mistakes, and often cost twice as much as permanent staff. That’s rarely factored into ROI models.”
“A robot works 24/7, doesn’t take breaks, and isn’t affected by motivation. In sectors like e-commerce or manufacturing, a human error might require a costly correction process—credit notes, returns, reshipments. If robots can prevent that, the financial benefit is indirect but significant. Plus, unlike wages, power costs are far more stable and predictable.”
Payback Time: A Real-World Example
He shares a concrete example: “We did an analysis for a parcel induction operation, where packages are unloaded from a plane onto a conveyor and then sorted. For one site, we calculated everything: hourly throughput, labor costs, staffing levels, training, and so on. The originally estimated payback time was 2.8 years. After recalculating using broader parameters, it came down to 1.1 years. The cost saving jumped from about €800,000 to €1.8 million.”
This example illustrates the value of advanced computer vision in automation processes—not only to guide robots and conveyors but also for quality control and inspection tasks.
He concludes, “This new understanding changes the conversation. We’ve shared this with our entire ecosystem—partners, integrators, and end users. Not because we claim to have all the answers, but to show that the picture is more complex. We support our partners with this. Many are technical experts, and we equip them with the tools to help explain the value of technology in a business context. That leads to better decisions and, ultimately, the best solution for the customer.”
See also
