Sourcing · 9 min

A partner walks into Monday's meeting with a thesis that took two years and three deals to refine. It is specific, defensible, and genuinely differentiated. Then an associate tries to turn it into a target list, opens SourceScrub or PitchBook, and discovers the thesis does not fit in the boxes. The industry dropdown has “Medical Devices” but not the therapeutic area the firm is deep on. It has “Environmental Services” but cannot separate non-hazardous liquid waste from residential septic. The thesis that makes the firm money has no checkbox.

Across 45 prospect interviews, the same complaint surfaced in six different shapes from six different firms, none of them prompted. The pattern was clear enough to recognize it as a core problem, not an edge case.

Why thesis-driven sourcing breaks standard database filters

Thesis-driven sourcing generates a target universe from a firm's specific investment thesis rather than from broad industry and financial screens. It breaks standard filters because those filters are built on fixed industry taxonomies — SIC codes, NAICS, vendor-defined verticals — that describe what a company is on paper, not the operational or strategic nuance that defines a real thesis. A filter can answer “industrial services companies in Texas with three million of EBITDA.” It cannot answer “companies whose revenue is driven by recurring compliance-mandated inspection work, regardless of how they describe themselves.” The first question is a screen anyone can run. The second is a thesis, and it's where lean firms actually compete.

Six niche theses that break the filter

• Medical devices, organized by therapeutic area: the database offers one “Medical Devices” line item with no concept of therapeutic area — no way to separate a cardiac-rhythm device maker from an orthopedics supplier.
• Life-science services, defined by who the company serves: “anything that generates cash by servicing pharma, biotech, and diagnostics” spans a dozen SIC codes with only the customer relationship in common — invisible to any dropdown.
• Environmental services, needing to exclude the obvious neighbors: a search for non-hazardous liquid waste companies returns solid waste and residential septic — wrong neighbors a coarse taxonomy can't exclude.
• Infrastructure maintenance in the ‘tank space’: inspection, testing, maintenance, and compliance-driven services — so targeted that the hard part isn't data, it's qualification no filter can express.
• The ‘weak industry categories’ problem, named directly by a partner: the classification layer itself is too coarse to carry a real thesis.
• ‘Garage door repair companies in Alabama’: a thesis invisible to structured filters, forcing a fallback to open-web search that returns data that's half wrong.

Why industry taxonomies fail more spectacularly in the lower middle market

Classification systems like SIC and NAICS assume a company describes itself consistently, files structured data, and fits a category built for a larger, more legible market. The $5–50M founder-led company breaks all three assumptions. The categories are too coarse; the underlying data is unreliable, which compounds the first problem — leading sourcing tools estimate private LMM revenue with 60–80% error in this segment; and the most differentiated theses are defined by attributes the taxonomy never captured in the first place, like which customers a company serves.

Rigid filters versus thesis-native screens

A filter operates on a company's stated classification. A thesis-native screen operates on the firm's articulated thesis and reasons about whether each company fits it, the way a sharp associate would if they had time to read every company one by one. With a rigid filter, the firm's edge — the specific, hard-won thesis — gets discarded at the first step, because the tool has no place to put it. The associate then spends a week rebuilding by hand the qualification the thesis already contained. A thesis-native screen keeps the edge in the loop from the start.

The bottom line

The fact that your best thesis doesn't fit a database filter is not a tooling inconvenience. It's the clearest sign you have an edge worth protecting, because an edge that fit neatly in a dropdown would be available to everyone running the same dropdown. The right response is to make the tool work from the thesis, not flatten the thesis to match the tool.