By Until now, if you wanted to research an audience with SparkToro, you needed to choose a particular website, a search keyword, or a URL to analyze…. No more.
As of today’s launch, we’re now providing the ability to describe, in natural language, any audience, with any restrictions or specificity you want, and get data about their demographics and behaviors. You can describe:
- Demographic attributes like gender, age, geography, job titles, roles, responsibilities, and combinations of any of these features.
- Behavioral attributes like websites they visit, social accounts they follow, brands they’re interested in, products they own (or want to own), services they’re seeking, or any combination of these.
- Intent attributes like places they plan to travel, products they might buy, or activities they may engage in
- Competitive data like who they’ve previously transacted with, visited, used, or considered
- The only limitation is the country — we’re still US, UK, and Canada only for now, but we’ve got plans in 2027 to be able to support many more countries 🤞.
And when I say “you can get specific,” I mean VERY SPECIFIC, e.g.
Yup. You can combine gender, age, income, interests, services sought, and purchase intent, no matter how nerdy and niche. Even *I* am impressed at how accurate this data appears 👇
(click image to enlarge)
You can go to town with deep granularity on B2B audiences, even including multiple non-overlapping groups and still get great audience data, e.g.
Even though this example audience likely includes only a few thousand (perhaps fewer?) individuals across Canada, SparkToro’s analysis data looks 🔥🔥🔥
Normally when people say “the possibilities are endless,” they mean you can choose from fewer options than most of us have fingers. But when we say it… it’s literally accurate. The process works by using this fancy new thing called Large Language Models (LLMs). You might have heard of them 😉. We…
- Send the description to an LLM (currently ChatGPT 5.2, but this may change as we test other models)
- Request the LLM send us back matching profile characteristics we can use to query our dataset (hundreds of millions of LinkedIn profiles we get from Leadfuze)
- Also request the LLM send us back domains likely visited by this audience, then expand based on our clickstream panel from Datos, which gives us visitation behavior across the web.
- Recursively match domain visitation to demographics and vice versa
- If we have enough matching profiles and visitation info (which is true for 99%+ of audiences we’ve seen so far), we run the full report and show you the data, usually within <90 seconds 🤯
If you want to learn more about the process, this short video has the details. But, in the short term, we beg of you, please, challenge us! Throw your toughest, most obscure, challenging, hard-to-analyze subsegments of ICPs and aspirational audiences at us. We’d love to hear if you stump our systems or find data that doesn’t match your expectations via: support@sparktoro.com.
You can try this new way of searching from your audience research home if you already have an account, or start right from the SparkToro homepage and get a free analysis right now.