Seniority and function classification

The tiers pass 1 applies. The full regex patterns live in title-taxonomy.json, which build.py loads — you don't need to read it. Consult it only to extend the taxonomy or when someone disputes a classification.

Seniority — first match wins

Evaluated top-down, which is why Chief Executive Officer resolves as founder/C-level rather than matching "executive" further down.

Tier	Matches
founder_c	Founder, Co-Founder, Fondateur/Fondatrice, CEO/CTO/CMO/CRO/COO/CFO, Chief … Officer, President, Owner, Managing Partner, Managing Director
vp_head	VP, SVP, EVP, Vice President, Head of, Director, Directeur/Directrice, General Manager, Country Manager
manager_lead	Manager, Lead, Responsable, Supervisor, Principal, Founder's Office
ic	Account Executive, SDR, BDR, Specialist, Coordinator, Analyst, Consultant, Engineer, Intern, Junior

Ordering is load-bearing. A flat set of patterns would classify VP of Engineering and Engineer identically.

Function — all matches collected

Unlike seniority, functions accumulate. General Manager of Sales and Marketing carries both sales and growth_marketing, and satisfies an ICP containing either.

Key	Matches
sales	Sales, Vente, Commercial, Account Executive, SDR/BDR, Business Development, New Business, Pre-Sales
growth_marketing	Growth, Marketing, MarTech, Demand Gen, Acquisition, Brand, Content, SEO, Paid
revops	RevOps, Revenue Operations/Systems/Strategy, Sales Ops, GTM, Go-to-market, CRM, Automation, Enablement
partnerships	Partnerships, Alliances, Channel, Affiliate
product_tech	Product, Engineering, Software, Technology, Data, Security, Platform, Architect
other	HR, Recruiting, Finance, Legal, Customer Success, Support, Coaching, Editorial

Collecting rather than picking matters for cross-functional titles, which are common at the size of company most ICPs target. Forcing a single function on a Head of Growth & Partnerships throws away half the signal.

Title first, bio as fallback

Detection runs on the job title. When the title carries no seniority signal, or no function signal, pass 1 falls back to the bio.

A Managing Director whose bio reads "Développement commercial" matches on that bio. This is worth more than it sounds: on a real 150-lead list, reading the bio cut the review bucket by roughly 60%.

Every bio-inferred match is flagged as such in its reason and enters the pass-2 queue, because inference from free text is exactly where a pattern is most likely to be confidently wrong.

Case is part of the pattern

Detection runs against the original string, not a lowercased one, and a pattern written with an uppercase letter is matched case-sensitively.

This exists because short acronyms are indistinguishable from ordinary words once case is discarded. Matched case-insensitively, IT matches the English pronoun — so a bio reading "making it happen" scores as a technology function, and if the ICP doesn't include that function the lead is dropped. Three patterns rely on this: IT, R&D, and EI in the agency list.

The practical rule when extending the taxonomy: write a pattern in lowercase unless the capitals carry meaning. \bproduct\b should match "Product", "product" and "PRODUCT" alike; \bIT\b should match only the acronym.

Multilingual titles

The taxonomy carries French alongside English — Fondatrice, Responsable, Directeur, Commercial, Vente — because a list drawn from a European market is routinely half non-English, and a taxonomy that only reads English routes all of it to review.

It doesn't cover every language. A title in one the taxonomy misses lands in review and gets resolved in pass 2, which is the correct failure mode: visible and recoverable rather than silently mis-sorted.

Noise titles

Student, intern, open-to-work, retired, investor and similar route to no_match rather than review. They aren't ambiguous — they're clearly not buyers — and putting them in review inflates the bucket that a human has to read.

The distinction is worth holding to generally: review means the engine declined to guess, not this lead is unusual.