When you target “VP of Engineering,” you’re missing everyone whose LinkedIn says “Head of Engineering,” “SVP Engineering,” “VP Software Engineering,” or “Engineering Lead.” Same role, different words. Every title you miss is a prospect your competitors reach and you don’t.There’s no standard for how companies name roles. A 50-person startup calls them “Head of Growth.” A Fortune 500 calls the same role “Senior Director, Growth Marketing & Demand Generation.” Both are your ICP. A keyword match catches one.Expand Titles solves this automatically. Give it any job title and it returns a ranked list of similar titles — not just obvious synonyms, but titles that real companies use interchangeably for the same role.
Landbase analyzes millions of professional profiles to map which titles actually co-exist at the same companies, in the same departments, at the same seniority level. If companies routinely staff “Growth Marketing Manager” and “Demand Generation Manager” on the same team, the system learns those titles are functionally interchangeable — regardless of whether the words overlap.This is grounded in labor market data, not language. It captures how companies actually organize, not how a dictionary would define roles.
Show What's happening under the hood
The technique is called Shifted Positive Pointwise Mutual Information (SPPMI). It’s a well-established method in computational linguistics that measures how strongly two items are associated based on co-occurrence frequency. If title A and title B appear together at the same companies significantly more often than chance would predict, they get a high association score. The “shifted positive” correction filters out noise from rare or coincidental overlaps.
A language model trained specifically on job title data understands what titles mean, even when the phrasing is unusual. It knows “SWE” means “Software Engineer,” that “Dir.” is “Director,” and that “AI/ML Developer” and “Machine Learning Engineer” describe the same work — even if no company in the dataset uses both titles.This catches new titles, niche titles, and abbreviations that haven’t built up enough real-world usage data yet.
Show What's happening under the hood
The model is fine-tuned on job title and LinkedIn profile data. It converts every title into an embedding vector that captures the title’s meaning. Titles with similar meanings are positioned close together in this vector space. Similarity is computed as the distance between these vectors across millions of titles, using a dedicated vector database for speed.
Statistical patterns miss new or rare titles — a freshly coined role like “AI Revenue Architect” won’t have enough co-occurrence data to score well, but the language model understands what the words mean
Language understanding misses titles that use completely different words — “Evangelist” and “Developer Advocate” describe the same role, but the words have almost no overlap for a language model to latch onto. Co-occurrence data catches this because those titles sit on the same teams at real companies
The combination eliminates both blind spots. A title that scores high on both real-world usage and semantic meaning is almost certainly a genuine match.
Every result includes a similarity score (0.6–1.0) so you can see exactly how confident the match is. The system also knows the department and management level of each title — meaning it distinguishes between “VP of Engineering” (engineering leadership) and “VP of Sales Engineering” (sales leadership) even though both contain “VP” and “Engineering.”Results are pre-computed, not generated on the fly. Lookups are fast (sub-second for a single title) and deterministic — the same input always returns the same output. The similarity table updates as Landbase’s underlying person dataset grows.
Up to 200 similar titles per query, with a default of 20. Only titles with a similarity score of 0.6 or higher are returned — this threshold eliminates weak or coincidental matches.
