Semantic search and lookalikes

6 min Intermediate

What you'll learn

  • How semantic search differs from keyword matching and when meaning wins
  • Choose strong seed accounts, since a lookalike list depends on your examples
  • Generate scored lookalikes at scale and read the scores to prioritize
  • Widen an audience to adjacent job titles when the core list runs thin
  • Extend into adjacent industries that share your best customers’ traits

Searching by keyword returns the records that contain the words you typed, which is fine until the accounts you want describe themselves in other terms. Semantic search reads for meaning instead, so a request about developer productivity also brings in companies that talk about engineering efficiency or shipping faster. This lesson pairs that with lookalike search, which starts from accounts you already trust and finds more that resemble them, and it shows how to widen a list when your first audience proves too small.

Prompt
Companies that resemble Acme, Globex, and Initech, scored by similarity, focused on US mid-market

Searching by meaning

Keyword search returns records that contain the exact words you typed, which fails when the accounts you want use different language. Semantic search reads for meaning, so the concept matters more than the wording. Reach for it whenever a phrase could be said many ways and you care about the idea behind it.

Choosing strong seeds

A lookalike search starts from accounts you already trust and finds more that resemble them, so the examples you pick decide the quality of the list. Choose seeds that share the traits you actually care about, such as your best closed-won customers, rather than a random sample. A few strong, consistent seeds beat a long, mixed one.

Scored lookalikes at scale

Landbase can generate many lookalikes and score each one by how closely it matches your seeds. Read the scores to prioritize, working the top matches first and using the score as a cut line when the list is larger than you can action. Because the score reflects why an account resembles your seeds, you can trust the ranking.

Widen to adjacent titles

When a core list runs thin, reaching nearby job titles recovers useful volume without losing fit. If a list of one title is too small, adjacent titles often belong in the same motion. Widen one step at a time and watch the count so quality holds.

Extend to adjacent industries

The same idea extends to industries. Companies in related sectors often share the traits that make your best customers a fit, so adding an adjacent industry can recover volume while keeping quality. Add one related industry at a time and review whether the new accounts still resemble your seeds.

Try it in Landbase

  1. Pick three of your strongest customers and ask the agent for companies that resemble them.
  2. Review the scored results and note why the top matches earned their place.
  3. If the list feels small, widen it once by adding an adjacent job title or a related industry.
  4. Keep the accounts that still match your seeds and drop the rest.

Prefer the CLI? The same semantic search runs from your terminal:

landbase-cli search "companies similar to Acme, Globex, and Initech in US mid-market SaaS"
Tip
Start narrow with your best seeds, then widen one step at a time. Small moves keep fit high while you recover volume.