Keyword search and filters can only match the exact words a company uses and the fields someone set up ahead of time, but your best audiences are rarely that neat. They are usually defined by meaning, like "companies building AI tools for developers," or by resemblance to accounts you already trust, like "more companies like our ten best customers." Searching by meaning and finding lookalikes are how Landbase builds those audiences, so you can reach the right companies even when they do not use your exact words and even when you could not name them one by one.
Keyword search is literal. Search for "developer tools" and you get companies whose descriptions contain those exact words, and you miss the ones that call themselves software development kits, an engineering productivity platform, or LLM infrastructure. Search by Meaning reads for the idea instead, so a request about developer tools also returns companies that describe the same thing in completely different words.
It works where categories fail. "B2B SaaS with a product-led growth motion" is not an industry, and filtering Financial Services to find mid-market fintechs that process payments would also pull in banks, insurers, and wealth managers. Describing the idea gets you the companies you actually meant.
Meaning works on top of filters rather than replacing them. You can ask for AI developer tools and still limit the results to Series A through C, 50 to 500 employees, and US or UK headquarters. The meaning part decides which companies match, and the filters narrow down the pool. A single description grows into a large, scored, pre-ranked audience, and the same request returns the same result every time.
A lookalike search starts from a handful of example accounts, such as your best customers or a closed-won list, and finds companies that are similar to them. It matches on what those companies do and how they describe themselves, not just their industry and size. That is why it can surface a company in a different category that shares your buyer, your budget, and the same problem.
You can work at two sizes. A quick preview returns about 25 of the closest companies in seconds, and it shows how many of your examples each result matches and which one it is most similar to, so you can see why it showed up. Use it to confirm your pattern looks right before you commit. Then build the full audience, which grows into a large scored list ranked by confidence. The habit to build is simple: preview a small set first, confirm it, then build the full list.
The similarity score rewards a broad fit. A company that is moderately similar to eight of your ten examples scores higher than one that is very similar to a single example. You are matching the overall pattern of your ideal customer profile, not one account, which is why a careful set of examples matters more than any single one.
Because everything is built from the examples you start from, the quality of those examples drives the quality of the list. Choose accounts that truly represent the kind of company you want more of, and pick a set that covers the real variety of your best customers rather than ten near-copies. A focused, representative set produces lookalikes you recognize, while a messy one produces a list you do not trust.
When a first audience comes back too small, grow it on purpose rather than starting over. Add closely related job titles when the core list runs thin, since the person next to your target often belongs in the same audience. Extend into closely related industries that share your customer's traits when a single category is too narrow. Grow the list one step at a time and watch the count, so quality holds as the list grows.
Each of these is ready to adapt to your own market.
Companies building AI tools for developersB2B SaaS companies with a product-led growth motion, 50 to 500 employeesShow me 25 companies similar to Gong, Outreach, and Salesloft, then build the full listFind companies that resemble our ten best customers, focused on US mid-market, with a VP of Sales contact for eachSame audience, but include adjacent job titles and related industries