WZ codes in M&A sourcing

The M&A market has changed radically over the past 15 years. Business models have become more hybrid, more technical and more complex. Yet while investment theses are becoming ever sharper, the fundamental tool of deal sourcing is often still based on Germany’s “Classification of Economic Activities, 2008 edition” (Klassifikation der Wirtschaftszweige, WZ 2008) or on NACE codes.

For analysts and M&A advisors, reaching for the industry code is the classic starting point when building a longlist. You filter by “62.01 Software development” or “28.99 Manufacture of other machinery” in the hope of mapping the market. Yet in today’s market environment, this static approach carries a significant risk. If you filter rigidly by codes in a dynamic economy, you are working with an apparent precision that does not exist in reality.

The problem is not just the inefficiency of manually weeding out unsuitable hits. The real risk lies in the false negatives – those hidden champions and attractive targets that slip through the net simply because they have been filed in the wrong pigeonhole. In a competitive environment where access to proprietary deals determines success, relying on WZ codes is no longer a dependable standard but a competitive disadvantage.

The problem: why reality no longer fits into codes

Why are traditional industry codes increasingly falling short? The reason lies in the structure of modern companies. A WZ code forces every company to choose one primary pigeonhole when it registers. Economic reality, however, can rarely be reduced to a single string of digits. Three structural problems make code-based sourcing inefficient today:

The hybrid trap

Modern companies often combine traditional industries with new technologies. A company that develops AI-driven drones for agriculture is neither a pure agricultural machinery manufacturer nor a classic software company. If the founder chooses the code for “Mechanical engineering” at registration, the company will be overlooked by every investor searching for “software” or “tech”. Today, innovation happens at the intersections – precisely where WZ codes are blind.

The “other” catch-all

Anyone who has worked with databases in depth knows the phenomenon: the most interesting niche players often hide behind generic codes such as “Provision of other business services” or “Manufacture of other goods”. These categories are vast catch-alls with no meaningful distinction between the businesses they contain. Exclude these codes and you lose potential targets. Include them and you drown in the noise of irrelevant hits.

The holding company problem (a DACH-specific issue)

Particularly in the German-speaking Mittelstand – the region’s mid-sized, often family-owned companies – operating units are frequently organised under holding structures. The WZ code of these parent companies is often “Management of own assets” or “Holding companies”. A conventional code search leads to a dead end here: either you filter out holding companies and lose access to the operating asset, or you end up with thousands of pure asset-holding vehicles that are not operating targets.

What is more, a code assigned when the company was founded ten years ago says nothing about the pivot it made two years ago. While the company evolves, the code stays static – and the target remains invisible to the traditional sourcing process.

The consequence for deal flow: noise instead of relevance

In practice, relying on static classifications leads to two costly inefficiencies in the deal origination process:

First, code-based search generates a great deal of noise (false positives). If, for example, you search for companies in “e-commerce” and select the corresponding WZ code, you get a list containing everything from global online shops to local retailers with a simple brochure website. As a result, analysts spend hours manually striking irrelevant hits off the longlist. Resources tied up in research are, in the end, unavailable for outreach and deal initiation.

Second – and more seriously – it creates systematic blind spots (false negatives). The most attractive targets are often those that cannot easily be categorised. A hidden champion developing a niche technology for the automotive industry may historically be listed as an “engineering consultancy”. If this company slips through the industry-code net, it effectively does not exist for the buyer. In a market where alpha is generated through an information advantage, these missed opportunities are the real cost of outdated search methods.

The fallacy of “intelligent mapping”: why these approaches fall short

In an attempt to make the rigid WZ logic usable, some providers layer semantic mapping on top of the codes. At first glance, the idea sounds pragmatic: the user enters a modern keyword (e.g. “cybersecurity”) and the tool then suggests the matching WZ codes.

But this approach is at best a cosmetic fix and does not solve the fundamental structural problem. It merely digitises the signpost to an outdated pigeonhole. The problem here is the disconnect between search intent and reality:

The “wrapper” effect

A tool that merely translates keywords into WZ codes ultimately remains shackled to the limits of the code. If an innovative fintech start-up is wrongly classified as “Other software development”, even the best keyword mapping will not help. The tool will unerringly lead the analyst to the wrong category or to a completely diluted list.

Lack of granularity

The directories can only reflect what is in the register. They do not validate what the company actually does. If you search for “3D printing for medical technology” and the tool directs you to the code for “Manufacture of medical devices”, you get thousands of hits. Identifying which of them actually use 3D printing remains manual work.

The validation problem

Such systems blindly trust the information the company itself provided when it was founded. They do not check whether the company still does today what the code suggests. Semantic matching that takes place only at the level of the classification, rather than at the level of operational company data (website, news, products), merely reproduces the mistakes of the past in a modern interface.

The paradigm shift: semantic beats static

The solution lies in moving away from metadata towards actual content. The “modern commercial register” has long since ceased to be the registry court’s database – today, it is the company’s own website. That is where companies describe precisely what they do, which markets they serve and which technologies they use.

A real-world example illustrates the difference: if you run a conventional search for “EdTech”, WZ codes will usually turn up only a handful of publishers or software houses. A semantic search, by contrast, reads the websites and recognises connections. It finds companies that use terms such as “e-learning”, “LMS”, “digital education” or “remote training” – regardless of whether the company is listed in the register under “Data processing” or “Education”. You are no longer searching for a company’s category, but for its business model.

The solution: AI-powered sourcing with ProxDeal

This is exactly where ProxDeal comes in. Instead of forcing users through rigid filter trees, the platform uses AI-powered semantic search. M&A professionals describe the target they are looking for in natural language – just as they would explain it to a colleague.

The decisive advantage is contextualisation. The algorithm understands the difference between a company that manufactures solar panels and a trade business that installs them. It identifies holding structures by checking whether a website points to an operating business or merely addresses investors. ProxDeal translates search intent (e.g. “SaaS provider of HR solutions”) into a precise longlist by matching it against the content of millions of websites and register records in real time.

The result is a longlist based not on outdated self-reported information but on the companies’ current digital footprint. This eliminates the “other” problem and makes niche markets transparent and accessible in the first place.

Conclusion: a new operating system for M&A

WZ codes were the right tool for an analogue, clearly delineated industrial economy. In the age of digital business models and hybrid value chains, however, they have become a limiting factor. If you still search solely on the basis of industry codes today, you accept unnecessary noise and systematically overlook niches and companies.

The future of M&A sourcing is not static but semantic. Technologies such as ProxDeal make this qualitative leap scalable. For M&A advisors and investors, this means less time cleaning up data in Excel lists – and more time for what really matters: the transaction itself.

Frequently asked questions