Lead discovery is hard—even when you have the right data. Business attributes like industry codes, revenue, or location are helpful, but they often fail to capture what someone really means when they describe their ideal customer. "witch supply stores," "boutique hotels with a wellness focus," "startups with a minimalist brand aesthetic"—these aren’t easy to translate into precisely specified queries.
Enigma’s initial Lead Discovery tools gave our internal teams the power to work with business descriptions in a taxonomy, but it assumed deep familiarity with our data model to pull out nuance. Users needed help navigating it, and even for experienced users, surfacing the right set of leads was often a multi-step, trial-and-error process.
We asked ourselves: "What if you could describe your ICP (Ideal Customer Profile) in natural language and get back real, viable leads—instantly?"
When we built the Enigma Explorer, we weren’t just redesigning a UI—we were rethinking the experience of business discovery entirely.
That meant:
We realized pretty quickly that dropdowns and filters alone wouldn’t get us there. Our structured data is rich, but for many use cases, it’s not expressive enough. We needed a way to capture the feel of a business—not just its tax classification.
This is where embeddings come in.
At a high level, embeddings are a way to turn complex things—like websites, descriptions, or business profiles—into numbers. Not just any numbers, but numbers that reflect meaning. Similar businesses have embeddings that are close together. That lets us compare businesses in a way that reflects real-world similarity, not just shared NAICS codes.
Our in-house data model, graph-model-1, includes embeddings generated from a massive dataset of business websites and metadata. It doesn’t just look at the words on a site—it learns patterns about design, tone, structure, and more. It captures the latent characteristics of a business: is it tech-forward? Family-run? Sustainability-focused? Traditional or modern?
Using these embeddings, Enigma can compare a customer’s ICP description or reference list to every business in our database—and surface those that are semantically closest. Think of it as a similarity search for business identity.
Here’s what’s under the hood:
This system allows us to compute and update embeddings at internet scale, and search across the full set in milliseconds. When a customer enters a prompt like "luxury pet spas," we can instantly return the closest matches—even if those businesses don’t explicitly use that phrase.
With embedding-powered discovery, our customers can:
It also opens the door for more intelligent iteration. Users can explore clusters, refine their definitions, and uncover patterns they didn’t know to look for.
We see this as the foundation for a much more flexible discovery engine. Embeddings let us model not just who a business is, but how it evolves. That means:
Over time, we believe this will enable more proactive, personalized, and dynamic workflows—discovery that adapts to your strategy, not the other way around.
We’re just getting started, but embedding-powered business discovery is already transforming how our customers explore the world of small business. Want to try it for yourself? Sign up for the Enigma graph-model-1 research preview and get exploring.