Sales and marketing folks have been talking about lead scoring for years, so we often get asked “what’s different” about Infer’s way of doing things. One of the reasons Infer’s models perform so much better than traditional lead scoring is that our system pulls in several thousands of external signals, going well beyond what most organizations track in Salesforce.com or other CRM and marketing automation tools.
Broadly speaking, Infer gets these signals from three sources: crawling the web, purchasing data, and inferring signals from raw data sources. The last of these is the most subtle, and it’s our equivalent of Google’s secret sauce for web search. Consider the relatively simple example of figuring out if someone is a “Manager” or not, just by using their raw job title information, which could be SVP, Associate, Director, EA — it’s not actually so simple. That’s why defining and deriving the inferred signals that turn out to matter is a key piece of Infer’s proprietary data science.
Specific examples of the external data we use in Infer’s models include:
Firmographic Data: Things like a company’s revenue, employee count, industry, etc. In many cases we get this data from multiple data providers so that we can provide maximum coverage and triangulate the right answer.
Web and Social Presence: We look at how many online sources are talking about the company. Do they have a Facebook profile or Twitter handle? These signals can tell us how legitimate/mature the company is.
Public Data: We use numerous public filings such as patents, location and financial information that are all available if you know where to look. These can often be leading indicators of whether a company is going to be a good fit for your product.
Company Reputation: We also glean insights from whether and how a company appears in various lists, from Fortune 500 to Crunchbase. Also, external postings such as their job openings and advertising campaigns are often important predictors of a company’s priorities.
Crawling the Company’s Web Domain: Our models use many distinct data points from a company’s domain, ranging from whether they’re displaying ads, to what kind of company they are (B2B vs. B2C, or SaaS vs. traditional software), to if they require login for any portions of the site, etc.
For each customer, our machine learning algorithms dig through all of these external signals (and others), and pick out the ones that matter in their specific situation. So while Infer has thousands of signals to draw from, any given customer model typically uses a couple hundred at varying weights.
And while some signals give us binary Yes/No answers, we also look at many in ranges — e.g. for revenue or employee count. As part of the process of determining which signals apply to a particular customer, Infer’s algorithms derive the appropriate break points and impact of different ranges. Certain revenue bands may be strongly positive, other bands weakly positive, and some revenue bands strongly negative.
You can see how companies who use Infer are at a distinct competitive advantage over those who don’t. It’s lead scoring on steroids. They’re able to effortlessly tap into the known universe of external signals and distill it into a simple score. A score that tells them where to focus their energy to maximize revenue. And just like Google, Infer is attracting incredible engineering talent to work on this problem and identify new signals. With each customer we add, the platform becomes stronger and stronger!