Collective[i]: How the FAANG companies inspired a B2B sales solution
It's a sign of the times that it's practically taken for granted when we hear about applications incorporating AI into their solutions. But when the company was founded back in 2012, Collective[i], short for Collective Intelligence, it was one of the first to tread that path. At the time, as the world was getting infatuated with big data, Internet giants like Amazon, Google, and Netflix were perfecting their approaches to applying neural networks and other machine learning approaches to propel their core businesses to dominance, and that's where the company got its inspiration.
Collective[i] picks up where CRM, sales force automation, and customer journey solutions leave off to optimize the last mile: providing a cloud-based SaaS solution helping B2B businesses with predictive and prescriptive analytics to close the sale. CRM tracks customer interactions, as captured by internal transaction systems, while customer journey solutions approach the problem from the opposite perspective: what are the paths that companies open to their customers, and how are their customers navigating them. Meanwhile, sales force automation, traditionally part of CRM, manages the workflows for handling sales leads, forecasts, and tracking team performance.
At first glance, the questions that Collective[i] addresses aren't all that unique: how to best forecast and close sales. It focuses on the workflows from forecasting to prospecting. Collective[i] is not alone here as People.ai and Salesforce Inbox also capture client activity, and like Collective[i], collect email, appointment calendars, invoices, video conferences, recorded phone conversations, external market data, and so on.
But there is one huge difference that sets Collective[i] apart: While most sales optimization solutions reflect only the client's internal data that may be augmented in some cases with third-party data, Collective(i)'s data set is far more massive. Collective(i) clients agree to make their customer interaction data available so that it can be analyzed anonymously as part of a master data set. A form of enterprise crowdsourcing, Collective[i] characterizes this as a network. Its data set is quite wide; the company estimates that in 2019 alone, it captured 2% of the world's B2B data traffic. That's the data set that is used for generating models supporting forecasting and various forms of customer buying proclivity analyses.
Their solution starts with a use of Robotic Process Automation (RPA), automating capturing of activity that can be fed back to CRM systems; this eliminates or at least greatly reduces the legwork that users of CRM have come to hate. It automates forecasting by using machine learning to generate probabilistic predictions based on aggregating buying and market data, rather than relying on subjective opinion.
Stealing a page from the next-best offer or product buying recommendations that consumers get from e-commerce sites, Collective[i] then generates recommendations of next-best actions for sales people to do. And with it come collaboration capabilities through "virtual deal rooms" where sales teams can coordinate prospecting and then identify and provide targeted guidance for stalled opportunities. The capability, currently compatible with all major CRM providers, logs email and calendar activities and mines communications for email signatures that provide current titles or positions. It is now being enhanced with a capability to decipher digital signatures, videoconferencing, and VoIP calls, and is adding opt-out and other capabilities for complying with data protection mandates such as GDPR or CCPA.
AI is key to Collective[i]'s solution. That in itself is hardly unique. For instance, Salesforce has embedded Einstein into its Sales Cloud, Service Cloud, Marketing Cloud, App Cloud, Analytics Cloud, and Community Cloud. Meanwhile, Oracle and SAP have sprinkled in AI to answer questions such as the probability of closing a sale, what actions should be taken to close a sale, what deals are likely to be closed the quickest, and what prospects should be prioritized.
But, as noted, Collective[i]'s difference is that it goes beyond the confines of an individual company's sales history, and by doing so, captures a fuller picture of the buyers who are being targeted, such as what other suppliers they work with, and what other products they have bought or are about to buy. Collective[i] has built graph neural networks for resolving the identities of individual buyers and business entities and has patents on how they are able to unify all the identities under a common data model.
It also applies recurrent neural networks (RNNs) and graph neural networks to sales pipeline analysis, forecasting the likelihood of deals closing, and when, and for what value. Models are generated based on the corpus of data that Collective[i] has aggregated from across its client base, showing what deals did and did not close, and all the extenuating factors around them. The underlying graph represents the strength and nature of interpersonal relationships, while the RNN model is built with time series events, with historical reference, that can decipher the strength of the opportunity.
So, when you apply AI on a much broader data set, how does that improve the results? The answer is getting a more nuanced picture that ultimately can shed more light on buying proclivity. As mentioned above, it starts with the broader base of data, from multiple companies, on which to train and develop AI models. Then within a single company, it could train the gaze on seemingly unrelated transactions that could yield signals on the likelihood of an individual sale closing. For instance, a company that has just changed its office leasing and and reducing the square footage is more likely to be in the market for new office equipment, and potentially, cloud computing services. Or a company using mainframe software is less likely to be buying clusters equipped with advanced GPU processors.
The insights can also capture human dynamics by drilling down on the behavior and interrelationships of different buying groups within an enterprise. If one group requires the other's support to greenlight a purchase order, their history of cooperation or lack thereof should provide an indicator to the sales team on how much to prioritize pursuing the lead.
Collective[i] is not the first company to build analysis based on networked data. As noted above, it drew its inspiration from the FAANG companies that leveraged their wide reach with consumers to build powerful advertising and entertainment businesses. The difference here is that Collective(I) applies that approach to the B2B world.