“Target the German Mittelstand” sounds specific until someone has to build the account list. It can refer to an EU-sized small or medium-sized enterprise, an owner-managed company, a private business below a particular revenue ceiling, or simply a well-known industrial supplier. Those groups overlap, but they are not the same market.
For a B2B team, that distinction is operational. A size filter can return a local subsidiary of a global group, miss a large family-managed manufacturer, combine a brand with the wrong legal entity, or assign a headquarters contact to a plant-level buying process. The problem is not a lack of records. It is a market definition that has not been translated into evidence.
That is also a sales problem. If research cannot explain why a specific legal entity or operating site belongs in the market, more contact data only moves uncertainty downstream. The useful output is not the largest possible list. It is a reviewable account map in which every accepted record carries the evidence, entity boundary, owner, and permitted next action a sales team needs.
This article uses the latest relevant German structural data available when the sources were reviewed on 8 August 2026. It does not turn those statistics into a claim about which companies should buy a product. Instead, it shows how the structure of the market changes account discovery, entity resolution, qualification, buying-centre research, and regional expansion.
<img src="/assets/blog/mittelstand-1.jpg" alt="Industrial company locations considered during German account mapping" />
Turn the market thesis into a qualification workflow
Leadbase does not measure German competitiveness, determine whether a company is truly Mittelstand, or guarantee coverage of a DACH segment. Its strongest fit is narrower and more useful for a revenue team: turning an ambiguous market thesis into an evidence contract, a candidate set, and an accepted-account queue that can survive review.
The distinction matters because each stage produces a different object:
This is not automated truth. A person still defines the evidence rule and adjudicates conflicts. Leadbase makes that process inspectable: Sheet permissions distinguish who can view, edit, manage, or own the work; named versions and history support recovery; and Assistant actions can remain in Ask for approval while the team tests an unfamiliar research task. See the product documentation for focused enrichment, Sheet history and recovery, and Assistant approvals.
If “industrial Mittelstand” is still a slide rather than a reviewable market, bring one narrow segment, several known-fit companies, and the subsidiaries or sites your current list gets wrong. Ask Leadbase to turn that real segment into a written qualification contract and a reviewable sample—before anyone buys contact data for the wrong entities.
Separate the definitions before counting companies
Three commonly used definitions answer different questions:
Sources: Destatis SME definition, IfM Bonn Mittelstand definition, and the methodology on page 30 of the KfW SME Panel 2025.
The practical rule is simple: attach the definition, reference year, and covered population to every market-size number. Never combine a KfW company count with a Destatis employment share or an IfM ownership claim as if they described one denominator.
What the official numbers establish—and what they do not
For 2023, Destatis reports 3.2 million SMEs, equal to 99.3% of enterprises in its covered business-economy sections. Those SMEs accounted for 53.3% of persons employed, 26.2% of turnover, 41.5% of gross investment in tangible goods, and 40.9% of gross value added at factor cost. The source uses an enterprise-size classification; it does not measure owner management or the narrower idea of an industrial Mittelstand. See the Destatis 2023 size-class table, published with methodological notes and last updated on 11 September 2025.
The pattern matters more than the slogan “99% are SMEs.” Company count, employment, turnover, investment, and value added have different distributions. A market model that treats every SME as an equally valuable account ignores this concentration.
KfW provides a second, broader view. Its 2025 SME Panel estimates 3.87 million private enterprises with up to €500m annual turnover for 2024. Of that KfW-defined population, 83% had turnover below €1m and 77.4% operated in service sectors. Manufacturing represented about 5.5% of companies but around 15% of employment, with an average of roughly 24 persons employed per manufacturing company. These are weighted survey estimates from 13,079 respondents, not a list of named target accounts.
This is the useful correction to the usual industrial narrative: the broad Mittelstand universe is mostly very small and service-oriented by company count. An industrial GTM team is looking for a segment inside that universe, not the universe itself.
Why industrial account discovery fails at the entity boundary
Official classifications assign an enterprise to the main focus of its economic activity. A manufacturer may also operate software, field service, distribution, and engineering units; a holding company may own the relevant factory while its own classification says little about the product made there. Industry codes are a starting point, not final proof of operating fit.
The distinction between a legal entity and an operating site is equally important. In its 2024 business register tables, Destatis reports 209,815 legal units and 220,606 local units in manufacturing. The legal-unit table covers economically active natural persons, legal persons, and associations with a German seat and qualifying employment or VAT activity; the local-unit table counts locations under its own stated criteria. Compare the official legal-unit table with the local-unit table.
The difference is not a count of “multi-site prospects”; the two statistical units follow different rules. It does prove why a sales record needs an explicit entity type. A legal entity, headquarters, factory, service branch, brand, and corporate domain can refer to related but non-interchangeable objects.
For each candidate account, preserve at least:
- legal name, legal form, and registration evidence where relevant;
- operating name or brand without replacing the legal name;
- registered seat and each relevant operating site;
- canonical company domain plus local or brand domains;
- parent, subsidiary, and ownership relationships as separate fields;
- the reason the specific entity or site belongs in the market.
Do not merge records merely because names are similar. Do not split every location into an independent sales account. The right model depends on where the buying authority, budget, technical need, and implementation responsibility sit.
Translate the ICP into an evidence contract
“Industrial Mittelstand” is not an executable ICP. Rewrite it as conditions that a reviewer can confirm, reject, or mark unknown.
Consider this hypothetical brief:
Find German manufacturers of configurable production equipment that operate their own service organisation, sell to regulated production environments, and appear large enough to support a dedicated aftermarket workflow. Map the legal entity, relevant German sites, and the commercial and technical buying roles. Do not look up contact data until the account passes review.
The phrase “appear large enough” still needs a defined proxy. It could be an accepted employee range, documented service locations, a hiring pattern, or another observable condition. Revenue or headcount alone should not stand in for operating complexity without a reason.
Retain the source URL, observation date, excerpt or evidence note, and reviewer decision for every mandatory criterion. “Unknown” is a valid result. Converting missing evidence into “no” lowers recall; converting it into “yes” destroys precision.
Build the market map in layers
A defensible market map is assembled, not downloaded in one pass.
1. Create the candidate universe
Use several discovery paths because each has a different blind spot: official registers and classifications, industry associations, exhibitor or certification lists, company product language, local site pages, and known-account lookalikes. Record the discovery source so you can later see which path produced accepted accounts rather than merely the most rows.
Keep the query version and collection date. A candidate is not yet a qualified account, and an unreviewed directory category is not ground truth.
2. Resolve entities before adding people
Create stable account and site IDs. Link aliases, brands, domains, parents, subsidiaries, and locations without flattening them into one string. Route conflicts to review: two websites can describe the same legal entity, and one website can cover several entities.
Resolve which object the sales team will own. For a centrally procured product, that may be the parent or headquarters. For plant-level maintenance, each operating site may need its own qualification and buying-centre context.
3. Apply mandatory gates and reason codes
Classify every candidate as accepted, rejected, duplicate, or unknown. Use narrow reason codes such as distributor_only, wrong_product, foreign_entity_only, scale_not_met, entity_unresolved, or evidence_stale.
A “hidden champion” label should never substitute for those checks. It is a useful research hypothesis, not an official industry or ownership field. A little-known specialist can be a strong account, but only the evidence contract explains why.
4. Research the buying centre after account acceptance
The account definition determines the roles to research. An industrial buying centre may involve operations, plant management, service, engineering, procurement, finance, IT, or executive management, but those roles are not mandatory in every company. Define role families from the product and the expected change:
Do not infer a person solely from a translated title. Preserve the original title, map it to a role family, and keep the evidence date. Only request the contact fields required for the intended channel after the role and current employment are sufficiently supported.
Measure sample quality before expanding the list
Before activating a full market, freeze a labelled sample. For example, take 80 candidates across the discovery sources and deliberately include borderline industry matches, possible group duplicates, and multi-site companies. Have a second reviewer independently assess at least 20 records.
Set thresholds before reviewing results. An example contract—not a universal benchmark—could require:
- at least 85% discovery precision: accepted accounts divided by all accounts with a decided fit outcome;
- 100% of accepted accounts with current evidence for every mandatory criterion;
- zero unresolved duplicates or entity conflicts in the activation set;
- no more than 10% unknown outcomes on the criterion that most often blocks a decision;
- at least 90% agreement on the double-reviewed subset before resolving disagreements;
- 100% of people research restricted to accepted accounts and mapped role families.
Publish raw counts with every percentage. If 44 of 50 decided candidates pass, report both 44/50 and 88%; also disclose the rejected, duplicate, and unknown counts. Track results by discovery source, industry stratum, region, and size band so a strong overall average cannot hide a weak segment.
When a threshold fails, revise the definition or source mix and label a new sample. Do not quietly loosen the criterion after seeing which companies a preferred method returns.
Treat DACH expansion as a new mapping exercise
German evidence does not establish an Austrian or Swiss market size. Legal forms, registers, industry coding, language, site structures, and available financial data differ. Preserve the common commercial definition, then rebuild the entity and evidence rules for each country.
The same applies within Germany. A regional cluster may use specialised vocabulary or supplier relationships that a national query misses. Measure incremental accepted accounts and duplicate overlap when adding a new region or source rather than presenting the larger raw list as expanded coverage.
Move only accepted accounts into the sales conversation
The commercial value of the market map appears at the handoff boundary. A sales owner should receive the accepted entity or site, the inclusion reason, relevant sources and dates, known exclusions, the likely role family, and the next action the reviewer approved. A company name plus an email address is not an equivalent handoff.
Leadbase keeps that packet close to the research decision. A team can select accepted rows for the next focused company or contact question, inspect the result before accepting it, and leave incomplete or uncertain research as no usable result. Where one field genuinely needs recurring review, scheduled enrichment belongs to that specific enrichment column and exposes its cadence, next run, and execution history; a completed run is not a guarantee that every row is correct or current.
The handoff can then be deliberate: grant the appropriate Sheet access or export the reviewed data for the downstream system. Leadbase does not replace the evidence contract, legal-entity judgment, CRM, or outreach workflow in this process. It gives Research, RevOps, and Sales one controlled boundary at which an account is either ready, rejected with a reason, or returned for more work.
To evaluate that boundary, freeze a real sample from one segment and include the uncomfortable cases: group companies, similar names, distributors, multi-site manufacturers, and sparse websites. Ask Leadbase to work through the sample with your team and show the accepted and rejected records—not a curated database demo. The useful sales conversation is whether the workflow produces explainable accounts your team would actually act on.
Source and definition notes
- Destatis: SME definition and 2023 size-class shares: enterprise-size statistics for the stated WZ 2008 business-economy coverage; latest table reviewed 8 August 2026.
- IfM Bonn: Mittelstand definition: qualitative unity of ownership and management; not a size-limited SME denominator.
- KfW SME Panel 2025: 2024 enterprise estimates from a weighted repeat survey; 13,079 participants in the 2025 wave; private enterprises up to €500m turnover, excluding the public sector, banks, and non-profits.
- Destatis: legal units by size and industry, 2024 and local units by size and industry, 2024: separate register concepts with separate inclusion and counting rules.
