Developing a technology is one thing. Figuring out how companies can actually use it is quite another. For the CISPA startup InputLab, this learning process began with about 40 discussions with banks and insurance companies—and with a surprising realization: The biggest problem often isn’t the technology itself.
Since August 2025, the team has been talking to people from quality assurance, product management, and other departments. It gradually became clear that while the technical generation or anonymization of test data is challenging, it is often not the actual bottleneck.
“As scientists, we initially assumed that the central challenge was generating high-quality synthetic data with complex structural and semantic properties,” says Dominic Steinhöfel, CEO of InputLab. “That remains technically relevant. But in practice, many organizations aren’t even at that point yet: First, it must be clarified what properties the data should actually have.”
The Problem Begins Before the Technology
For a new testing project, quality assurance, development, product owners, business analysts, and—in some cases—database teams must work together to determine which customer profiles, edge cases, and domain-specific properties are needed. This coordination can take a long time. And even once initial data has been provided, it’s not uncommon to find that requirements were understood differently. The challenge, therefore, lies not only in generating data but also in bringing together the knowledge about it.
This has also changed product development at InputLab. “What surprised us most of all was how important collaboration features and integration into existing structures are,” explains Steinhöfel. This includes role and permission models, version control, integration with existing systems and formats, as well as a user interface where even non-technical users can contribute their domain expertise. The original focus on data generation thus evolved into a collaborative enterprise solution.
Data Knowledge as a Competitive Advantage
Another surprise: According to Steinhöfel, data landscapes in some companies are surprisingly poorly documented. Knowledge about dependencies and domain-specific characteristics often resides in the minds of individual experts or is only indirectly reflected in legacy software systems. When these individuals leave the company, relationships sometimes have to be reconstructed from the software itself.
This also explains why companies continue to use real data for testing: not necessarily because it’s the best solution, but because it already contains the required business characteristics. At the same time, this creates drawbacks in terms of data protection and targeted test coverage. The assumption that a better algorithm alone would solve the problem thus proved to be too simplistic.
From Spinoff to Company
The learning curve doesn’t just apply to the product. InputLab started as a spinoff at CISPA and was founded in May 2025. During its time at CISPA, the startup carried out three successful, unpaid pilot projects. The team is currently in concrete discussions about further pilot projects in the insurance sector.
The path to getting there is also part of the learning process. Contacts were established through conferences, trade shows, and events, as well as through direct outreach. Typically, there are two to five interactions with a potential customer before the first demo. With insurers, sales cycles can last 12 to 24 months, explains Steinhöfel. At the same time, InputLab is developing its offering for small and medium-sized enterprises, where the team expects shorter decision-making cycles.
How Research Leads to Technology Transfer
InputLab’s experience shows that technology transfer doesn’t end with a good technology. That’s when a new learning process begins. For deep-tech startups, this means expanding the research perspective to include a market and customer perspective. Conversations, networks, and pilot projects help test assumptions and further develop their own solution. Steinhöfel: “As a scientist, you try to understand natural or technical systems. As an entrepreneur, your main task is to understand people.”
