LinkedIn’s entry into predictive analytics has sparked an important conversation, both regarding the state of the emerging predictive industry and LinkedIn’s place in the enterprise software world. Given many of the company’s moves – most notably its Bizo and Fliptop acquisitions – it is becoming increasingly clear that LinkedIn intends to be much more than just an online “professional network.” There’s little doubt that it wants its place in the B2B sales and marketing software stack.
The question is, what does this mean for the other big players? LinkedIn’s latest announcement was very likely the first of many moves that we’ll see in the predictive market from folks like Salesforce, Microsoft, Oracle, Marketo, HubSpot and others. In fact, Marc Benioff recently spoke with Fortune Magazine about the ‘AI (artificial intelligence) spring’ saying, “When I look at the next set of technologies that we have to build in Salesforce, it’s all data-science-based technology. We don’t need more cloud. We don’t need more mobile. We don’t need more social. We need more data science.”
If you look at how the AI spring is likely to play out, there are a few logical approaches the big software players will take as they look to bring predictive capabilities into their product portfolios.
Adding Predictive Features
The first of these is to extend their existing apps with basic predictive functions – essentially playing it safe. These predictive features will probably be based on data the vendor already controls and should work with minimal customization. For example, rather than requiring manually assigned point values to arrive at basic lead scores, marketing automation vendors might enhance their lead scoring capabilities by using a handful of variables that are consistent across their customers to start calculating predictive scores.
These improvements would help with product adoption because they can reduce configuration hurdles for large swaths of their customers. However this approach would only scratch the surface of what predictive can do because these vendors are unlikely to support custom-fit approaches. For a model to be highly predictive, you need to really understand the customer’s data, their process, their objectives, and you need to be aware of overfitting. If a vendor is not on top of all those elements, they’d run the risk of making bad recommendations. It is equivalent to Waze going awry and sending you an hour out of your way, only in this case your customer’s real revenue is at stake. That can create major issues for these automation companies that might even impact their renewals and recurring revenue.
While the big software players could well build a deeper predictive service, it is more likely that they would move into it via an acquisition. There are a handful of predictive vendors that have been working on this for years, and acquiring one of them would give the incumbents a jumpstart on the talent and technology front. The challenge is that building custom-fit predictive models is a very different business than selling an empty database or per-seat application like CRM. It requires a distinct technology architecture, new sales and support models, and a fundamentally different perspective on whether you should be able to actively see your customer’s data. When a company buys a database, they don’t want their vendor peeking in, but with predictive modeling that is a prerequisite (otherwise they can’t monitor and tune the predictions).
Another hurdle that will likely keep big providers on the sideline is that we’re not yet at a point where the revenue is material. Even Salesforce’s smallest product lines probably generate more revenue than all the vendors in the predictive space combined. That’s not to say that the predictive market won’t one day be substantial and strategic, but we’re still in early days. And right now the pace of innovation is so fast, that pure play predictive vendors have somewhat of an advantage. Rather than be pulled by competing initiatives they can focus all their energy on building the deepest, most competitive product.
Cultivating a Predictive Ecosystem
The third strategy the big players will inevitably pursue is to build their ecosystems. If a vendor has dozens of certified partners with different flavors of predictive, they can give their customers more choice; all the while reaping the adoption benefits of AI. Predictive intelligence makes workflows, analytics and advertising that much more effective. By playing Switzerland, large software providers can invite in more innovation. And when it feels like the market is too big to ignore, they can make their move. This is very similar to the strategies we saw with the mobile CRM or marketing automation industries.
In the coming months, I expect that some of the marketing automation vendors will start building their own basic predictive capabilities. I think we’ll also see smaller talent acquisitions as some of the predictive vendors get squeezed out. And there will surely be pure play vendors who stay focused and separate themselves from the pack.
At the end of the day, most businesses will eventually want to adopt custom models that meet advanced predictive requirements. Many will bring in specialized solutions and expect their software vendors to plug in whatever predictive scores they want. By supporting this ecosystem, the Salesforces and Microsofts of the world can unlock more value for customers and add stickiness to their own products.
Whether vendors buy or build their own predictive features, I predict that they’ll also continue to partner with several different best-of-breed vendors in the predictive marketplace. This approach not only minimizes their risk, it offers their customers’ choice and drives innovation across the industry.