The founders of Infer Inc. have long been in the business of figuring out what consumers will buy next. With Infer, they’ve been applying those same techniques to businesses that sell to other businesses.
To that end, the company last week announced a partnership with Terminus to extend predictive analytics to B2B advertising by combining the Infer Profile Management platform with Terminus’ Account-Based Marketing platform. Moving into the B2B represents a “greenfield” opportunity for Infer, according to Vik Singh, CEO. Businesses already keep their data in the cloud. With the right APIs, that data can be accessed easily for Infer’s predictive software to crunch.
Infer followed up its partnership with Terminus by announcing support for Oracle’s Eloqua platform at Oracle’s Modern Marketing Experience in Las Vegas today. New features include top event and activity dashboards, which can take marketing and sales data and chart it out to show client activity, like e-mail response, web page visits, and content downloads. That information can be used to flag opportunities to turn prospects into conversions.
Infer also uses the Eloqua API to empower its predictive platform to gather Eloqua data into a historical record, again analyzing those records for “signaling”–signs that a prospect is interested in purchasing a product or service.
B2B is catching up with the “business to consumer” (B2C) marketing, Singh noted. Models analyzing data in the B2C space are very product-specific and tailored in-house, he said, while B2B firms rely on cloud-based data and SaaS-based CRM applications like Salesforce. That results in more standardized data, easily accessed via APIs. Predictive models can be built more quickly, and sales people can reach customers more quickly as a result, he noted. “B2B can be a role model for B2C in how we do predictive.”
Customer signals are also easier to spot in the B2B space, and also more consistent. If a prospect visits the firm’s pricing page, that is indicative of interest, Singh said. Infer can evenly distinguish between a real prospect and a “spam prospect”, based on their keystrokes. An interested person will take time filling out the fields correctly, while a casual prospect will just fill the fields with random words, Singh observed. Even hiring can tip off the direction a potential client’s investment, showing what market they will be staffing up to reach.
Every firm takes a different approach towards predictive modeling. Infer’s approach relies heavily on machine learning. “Machine learning allows us to personalize the modeling for that customer,” Singh said. Specifically, this means “personalizing” the Infer platform to handle the customer’s data. Infer can construct a model that will comb through a client’s data, but it will also pay attention to external signals that may signal interest, like registering for an app’s demo period or downloading a white paper, Singh explained.
Leads can be scored on a scale of 0-100. To make that information useful to a sales person, Infer will recast those scores into broad categories—A, B, C, D. Each category acts as a shorthand reference to indicate which customers are more likely, or less likely to convert from a sales pitch to a sale.
Users are sometimes tempted to “hyper-segment” their markets, but “scoring is a way to solve that,” Singh said. By simply measuring the likelihood of a sale, a user can focus on prospects instead of constructing a myriad of rules-based categories and sub-categories. (Hyper-segmentation can be valuable too.)
As potential customers “signal” their intent, one must be able to sort through correlation and causation—that a signal will lead to a sale. This is pretty difficult, Singh said. Infer tries to stick to a correlation-based approach, confident that the signals selected are the right ones to pursue. “If the signals are not adding value, filter them out,” Singh said. “You don’t want to overthink this.”
