Equinix goes partner prospecting with AI

Multinational data infrastructure company Equinix has been capitalizing on machine learning (ML) since 2018, thanks to an initiative that uses ML probabilistic modeling to predict prospective customers’ likelihood of buying Equinix offerings — a program that has contributed millions of dollars in revenue since its inception.

But as the company evolved since the launch of that project, so did its reliance on channel partners to accelerate customer acquisition and expansion. So, in 2021 Equinix revisited its prospecting platform to take it a step further by adding a data-driven sales prospecting approach that uses AI to identify partners best positioned to help the company drive new sales, both globally and within specific regions and countries.

That’s because, in certain geographies, sectors, and industries, Equinix’s channel partners are uniquely positioned to address customer demand for unbiased guidance, integrated solutions, and advanced services. Case in point: the federal sector, where it is critical to identify partners with necessary clearance and previously established relationships.

“Unlocking the immense potential of AI to deliver a tangible impact to our business was a big priority for our IT organization,” says Milind Wagle, the company’s CIO. “Building an innovative, intelligent AI-based prospecting engine for our channel program was the perfect use case that enabled us to combine the power of AI technology innovation, build competitive market differentiation for the company, and help improve the experience for our customers and our channel partners.”

Dubbed “AI-driven partner prospecting,” the initiative  pinpoints which prospective customers are best served via Equinix direct sales versus indirect partner or channel sales. The initiative, which earned Equinix a 2023 CIO 100 Award in IT Excellence, has two goals: To identify partners with the highest potential to drive new logo acquisition, and to prioritize those partners that are predicted to generate the highest bookings value.

“This allows Equinix to focus its investments and resources on the partners best suited for joint sales and resell activities,” says Ted Dangson, senior director of applied AI strategy and analytics at Equinix.

With the full rollout scheduled for mid September, reviews of model outcomes and dashboards continue to show how data science can enable IT to help sales better target and drive improved revenue using AI and ML.

The power of prediction

Ram Bala, senior principal data scientist at Equinix, provides scope for the task at hand.

“Equinix has unique needs when it comes to opportunity identification and partner prioritization,” Bala says. “More than 1,300 technology vendors and service providers worldwide have gone through a rigorous vetting process to be recognized as Equinix partners, and they’ve registered more than 9,000 deals with Equinix over the last three years. From a federal corpus perspective, with access to many opportunities and numerous requests for proposals [RFPs] in the US alone, it’s imperative to identify Equinix-relevant RFPs and partners for joint sales.”

By applying the right data management, propensity-based analytics, ML, and business intelligence tooling, Dangson says his team realized in 2021 that Equinix would be able to analyze data from channel partners and end customers to pinpoint which customers were best served directly via Equinix sales versus indirectly via partners and resellers. It would also be able to map end-user needs to partner services affinity and surface insights to help all parties accelerate revenue growth.

Dangson’s team, which worked closely with Equinix’s partner and federal sales and marketing teams to identify the opportunity, first looked for vendors that might have out-of-the-box solutions that could solve its use case, but ultimately decided to build a custom AI model framework in-house in collaboration the Equinix’s IT, data science, and engineering teams.

As part of that work, Bala and his team of data scientists undertook extensive analysis of internal and third-party data to identify data sets critical for developing an effective partner prioritization data science strategy.

“We utilize prospect and partner-relevant firmographic and technographic data attributes and rely on past government contracts and awards data from open-source federal databases,” Bala says. “In addition, we solicit consolidated access to text documents and PDFs, which provide extensive information on upcoming opportunities and RFPs. We also identify pertinent past relationships between similarly profiled customers and partners from Equinix internal datasets.”

The team then set about building ML models that used those data to:

• develop global- and country-level scoring and recommendations for partner prioritization for corporations
• identify government-initiated digital transformation projects relevant to Equinix, and develop country- and agency-level scoring and recommendations for partner prioritization for federal agencies
• scientifically validate existing partners and identify new partners to prioritize
• pinpoint end-prospects and customers best served via direct or indirect sales
• reorient the pursuit of channel sales to enable partners to prospect for customer activation and data-driven selling.

Equinix’s partner prospecting platform leverages natural language processing (NLP) algorithms to extract relevant excerpts from RFP documents, accompanied by a relevance score for each opportunity, says Dangson, who notes that the algorithms also provide support reasoning behind their recommendations.

“These additional details have revolutionized the way end-users interpret and utilize the model predictions, leading to a gradual rise in adoption and overall success,” he says.

Data annotation and inadequately labeled samples for training ML models proved to be the project’s biggest challenge, Bala says. The lack of annotated data made it difficult to build high-accuracy and computationally efficient models to identify Equinix-relevant RFPs from government agencies, and inaccurately labeled samples made it hard to train ML models to prioritize partners for corporate sales.

“To solve these, we leveraged technologies that stemmed from diverse academic and corporate research facilities,” Bala says. “It took us nearly four months to develop the minimum viable product and another five months to develop a scalable, integrable end-to-end solution.”

Delivery vs. innovation

In the wake of deploying the solution, Equinix says end-users have come to see it as a key tool to make their jobs easier, faster, and more accurate. In its second-quarter 2023 earnings report, the company said its channel program accounted for 40% of books and nearly 60% of new logos.

Bala believes the key to success in driving digital transformation through projects like AI-driven partner prospecting is to strike a balance between delivery and innovation.

“Our goal is to foster an environment where innovation thrives, ensuring it aligns with the objectives of delivering measurable business value and maximizing return on investment,” he says. “As we spread this culture of innovation across the organization, we’re starting to see the gradual rise of transformative initiatives gaining traction. This journey not only fosters creativity and positively influences team morale, but also creates an environment where failure is accepted as a valuable learning experience.”