How much of a company’s success is down to the industry it operates in? It is one of the oldest questions in strategy research and for forty years the answer has rested on the assumption that we can correctly organise firms into their appropriate industries.
In new research forthcoming in the Strategic Management Journal, UCL School of Management Associate Professor Bart Vanneste and PhD student Alireza Eshraghi challenge that assumption. They find that the systems used to file firms into industries — SIC, NAICS, GICS and others — disagree substantially about where firms belong and that those disagreements quietly distort what we think we know about why some firms outperform others.
Consider a simple example: A UK startup develops software to manage energy use in commercial buildings. To customers, it would be viewed as a climate technology firm. Its founders might consider it sitting somewhere between software, energy services and real estate. But when the firm files accounts, raises institutional investment or appears in financial databases, it must be classified. Is it a software company, competing with other software developers? Is it an energy firm, grouped with utilities? Or is it part of property services? Each classification has consequences, placing the firm alongside a different set of peers, exposing it to different comparisons and shaping how its performance is judged.
Industry labels are typically assigned by government agencies, data providers or financial index compilers using formal classification systems and once assigned, those labels can ripple outward. Investors use them to build portfolios. Regulators use them to decide which rules apply. Researchers use them to study competition and performance.
Dr. Vanneste and Eshraghi’s study shows that these labels are not neutral. Using older classification systems can place very different firms into the same industry or split highly similar firms across different ones. When that happens, the shared features of an industry are blurred. Performance differences that are really driven by market conditions can end up looking like firm-specific success or failure.
Their research systematically explores this problem across publicly listed North American companies. The authors compare several widely used industry classification schemes — including the long-established Standard Industrial Classification (SIC), and more recent alternatives such as the Global Industry Classification Standard (GICS) — and show that they often disagree substantially about where firms belong.
The paper then introduces a new approach that uses machine learning to group firms based on the language used in their annual reports. Rather than relying on fixed categories, some frozen since 1987, this method clusters firms that describe genuinely similar activities and the difference is clearly noticeable.
When firms are grouped using older systems such as SIC, industry explains around 7% of performance differences. But when firms are grouped using the new machine learning approach, that figure rises to 25% — with a corresponding decrease in what appears to be firm-specific performance. What can look like the result of unique strategic choices is often better explained by the market conditions firms operate in.
Being classified in one industry rather than another affects who a company is compared to, which investors take interest and how regulators see its activities. For fast-moving sectors such as digital platforms, clean technology or artificial intelligence, outdated classifications can leave firms squeezed into boxes that no longer fit.
Speaking about the paper Dr. Vanneste said:
“Strategy research has long estimated that industry plays a relatively small role in firm performance. Our results suggest that conclusion is largely an artefact of how firms have been classified, not a feature of the world. With more accurate groupings, industry’s role looks far larger and important.”
The study emphasises that some of the most influential decisions in business are not made in boardrooms or on shop floors, but in the quiet machinery that decides how firms are grouped and understood.
