The term “data-driven marketer” is rapidly becoming a redundancy: to be a marketer these days is to be data-driven, whether you work in demand gen, marketing ops, content, or social.
And yet, being an evidence-based marketer is not without challenges.
- Data - We are simultaneously drowning in data while also missing some critical pieces of it.
- Time - Marketers still have programs to run and don’t have all day to spend looking at reports.
- Skills - Data analysis is a skill. Not all marketers have developed it. And that can lead to very bad decisions (a little analysis is a dangerous thing).
Predictive analytics software is so appealing precisely because it solves all these problems. Predictive tools tools tap into our existing datasets, combine them with thousands of external signals, and use advanced statistical methods and machine learning to produce statistically valid insights in very little time.
If we are serious about being data-driven, it’s only logical that more and more of our marketing decision-making will become powered by this sort of technology. What will this future look like?
With this question in mind, I spoke to six predictive analytics vendors about how predictive can be applied to empower marketing. I distilled their feedback into five top use cases for integrating predictive into marketing automation that are all possible today.
How Predictive Analytics Tools Work
Before we explore these use cases, let’s consider how exactly predictive tools work and how they should fit into your existing technology stack.
There are many predictive marketing vendors, each with their own unique features, but at a high level they all do the following:
- Read Internal Data - Evaluate the demographic/firmographic data in your CRM, behavioural data in your MAP, or potentially even information from a product database or internal data warehouse.
- Add External Signals - Enrich those records with external data points on your contacts and accounts (e.g., technologies used, job posting activity, intent data, and so on); most vendors have their own web scrapers to collect this information and also purchase it from third-party data vendors.
- Build a Model - Evaluate your historical success metrics and use analytical techniques to build a predictive model. This model identifies variables in your data that correlate with successful outcomes, and it can now be applied in various ways to make data-driven marketing decisions.
How Predictive Marketing Fits into your MarTech Stack
So how does a predictive tool fit in with your existing marketing automation platform? The key thing is to understand the roles of each system in your stack.
The System of Execution
Automation tools like Marketo excel at enabling marketing execution. Marketo is a workflow machine, allowing you to define rules-based actions with enormous flexibility and complexity.
Marketing automation also tends to be the central integration point for all marketing activities - the hub of the wheel that drives outreach across multiple channels through more specialized integrated apps. You can apply that incredible workflow power to achieve outcomes directly or through a variety of integrated systems.
For all these reasons, it makes sense to think about marketing automation as your primary “System of Execution”.
The System of Insight
While marketing automation gives you the tools to act, they don’t necessarily tell you what you should do. That’s where predictive comes in. A predictive tool can ingest all your data, combine with external signals, and provide decision support and guidance. As Jim Walker (VP of Marketing at EverString) put it to me, this is your “System of Insight”.
From this perspective, predictive analytics can be the “brain” that powers the execution of your sales and marketing programs via your MAP, CRM, and through a stack of specialized execution tools.
(Side note: there are clearly some areas where Marketo is both system of execution and insight -- e.g., RTP provides predictive content recommendations already. And Marketo is investing heavily in building up its own predictive capabilities, so we may see more convergence between execution and insight going forward.)
Now that we’ve got some context, let’s look at how this interaction between insights and execution can work in practice with some concrete examples.
How to Combine Predictive Insights with Marketing Automation
This is the most well-known marketing application of predictive analytics, as most vendors started out as predictive lead scoring tools before expanding to other applications.
Predictive lead scoring is a lot like the traditional lead scoring that we build in Marketo, only smarter. Instead of using anecdotal insights or limited data analysis to decide that a job title containing “CMO” should be +30 points and someone attending your webinar should be +15 points, you feed all your data into your predictive model and let the algorithms determine how leads should be scored.
Don’t get me wrong: traditional lead scoring can still be tremendously valuable in prioritizing your leads. But there are limits to how far you can take it, even if you are a data geek and spend a lot of time working in a spreadsheet. A computer can consider thousands of variables and the correlations between them to deliver a lot more predictive power.
Calculating a predictive score for a lead is just one part of the puzzle. Many tools can also help enable salespeople to act on the data. For example, Nipul Chokshi, Head of Product Marketing at Lattice, stresses the importance of making it easy for sales to take action on the highest-scoring leads. He suggests to accompany the predictive score with insightful context, such as the external signals that drove the score. For example, you could tell a rep that the prospect account uses complementary technologies or that contacts at the company have been demonstrating intent, and so on.
He notes it’s also important to make the predictive data actionable - for example by providing call plans with talking points and discovery questions all in one easy-to-access place.
These steps ensure marketing is building real alignment by providing full context behind the score and actionable insights for next steps.
Demand Generation/Account Targeting
Lead scoring is fundamentally an inbound application of predictive analytics. Demand generation turns this on its head, allowing you to go outbound and support an account-based marketing strategy.
Instead of telling you which of the leads already in your funnel are the good ones, predictive demand generation tells you which accounts should be in your funnel and then gives you the information for contacts on those accounts.
Account selection is sometimes called “look-alike” modelling, because the tool will take the profile of what a good customer looks like for your business (How many employees? Annual revenues? What technologies used? etc.) and then match that profile against a proprietary database of companies. Every vendor I spoke to has one, and they typically have millions of accounts.
You can then import these new contacts directly into your CRM and MAP and from there engage them in outbound sales and marketing programs.
Nurturing is all about cultivating relationships with the right people until they are ready to buy. But how do we know how ready they are? Using predictive analytics there is a variety of ways to more intelligently route leads into the correct nurture.
Architecting Your Nurture Around Fit
Sean Zinsmeister, Senior Director of Product Marketing at Infer, shared some of his approaches for architecting a lead nurture using predictive analytics with me. In Sean’s diagram below, leads are routed based on likelihood to convert (fit data) so the messaging sequence can be calibrated accordingly. “A” leads are routed directly to sales while other leads are given increasingly softer messaging the less likely they are to be sales-ready.
Architecting Your Nurture Around Buying Stage
When it comes to nurturing, Amanda Kahlow, CEO of 6Sense notes that “a key factor missing in this is timing: whether the prospect is even in an active buying cycle to care about your products and services.”
According to Kahlow, the key to resolving this is better activity data that allows marketers to connect the behavioural signals collected on their own website with “buyer intent signals from third-party activity data across the B2B web.”
For example, she notes that if multiple employees from Acme Inc. are clicking on ads for server virtualization technologies, browsing content related to virtualization on major publisher sites and buying guide portals, or downloading related resources across the B2B web, Acme Inc. might be considering a purchase in this area.
“This intent data from Acme’s buying committee is aggregated at the account level and fed into our predictive engine, where it can be used to determine not just whether Acme is in-market for server virtualization, but even the specific buying stage they are in.”
I think this intent data has powerful implications for your nurturing. For example, if you were equipped with detailed buying stage insights about your prospects, you would no longer need to start everyone on an early-stage or “awareness” nurturing track just because they are new to your database. If you have intent data to suggest they are actually well-along into research or even consideration of your specific product, you can route them accordingly and give them the most relevant content.
A marketing rule of thumb is that the more relevant your content is to a prospect, the greater impact it will have.
One big challenge to relevant marketing is a lack of knowledge about your prospect and their interests. Are they after product A or product B? Which of 5 different pain points/value propositions should you emphasize? Should you talk about integrations with complimentary products they are already using? Many predictive tools can help you solve these problems.
Nipul Chokshi suggests creating multiple scoring models for different products. This way you can evaluate which product has the highest score and provide the lead with content they’re most likely to be interested in.
Other vendors expose rich demographic/firmographic/technographic data about your prospects and allow you to build marketing campaigns around those data points. I’ve seen Leadspace at work during a client engagement and was impressed with how it provides full-scale data enrichment in addition to building predictive models on top of that data.
As an example of what this enables, you can detect if a prospect is using a complementary technology and then personalize emails and landing pages to highlight your compatibility with that technology or have a completely different series of touchpoints based on this knowledge. It also allows you to potentially cover both data enrichment and predictive needs with a single app.
Lastly, it is also possible for a predictive tool to do the heavy lifting in building rich profiles for segmentation - profiles that go far beyond just a handful of data points. Infer has a really interesting tool for this purpose, which they call their Profile Management solution. It aids users in combining many data signals together to form what Zinsmeister calls a “hyper-segment”, a profile of one of your ideal customers. Infer will use its predictive model to assess how desirable the segment is in real time and gauge the business impact all the way through opportunity win-rate.
Now you can work to develop highly relevant content that is completely designed around the needs of that profile. And as you identify or acquire new contacts that fit this desirable profile, you can feed them the extremely targeted messaging that resonates with pain points you know they have.
Marketing Program Evaluation
Jessica Cross, Director, Demand Acceleration and Customer Marketing at EverString, had some great tips on using predictive analytics to evaluate your marketing programs. By evaluating the quality of leads each program produces in real time, you can get a leading indicator about where to invest your marketing dollars. Without predictive, two programs producing the same volume of leads might seem equally appealing, but with the lens of a predictive model you could see immediately that one program is producing leads of much higher quality and should be prioritized.
During his keynote at Marketo’s Summit in 2015, CEO Phil Fernandez presented the metaphor of the self-driving car as a vision for the future of marketing. In this vision, the marketer will input the goal, and the technology will figure out how to get leads to that destination.
This vision would seem to be a logical conclusion to the development of predictive analytics. But does this mean that the marketer’s creative inputs will soon be obsolete? Far from it, for the marketer still needs to create the destinations on the journey (by designing relevant content and touchpoints), while predictive tools provide the capacity for determining the best route for each customer to take.
Mintigo’s new Predictive Campaigns functionality is designed to do exactly that. Predictive Campaigns can evaluate the relationships between marketing programs, contact and account profiles (defined by technographic and intent data), and successful outcomes to develop predictive models for campaign orchestration.
These models would provide the ability to determine, for any particular person, that they should receive the following touches in the following order through the following marketing tactics to have the best chance of reaching a successful outcome. In other words, this helps marketers offer the right message through the right channel to the right person at the right time – essentially 1-to-1 personalized marketing that’s dynamic to each individual buyer’s journey.
It should be said that it is still early days for this capability and the feature is only available for Eloqua right now , but nonetheless it is a powerful example of where these technologies are headed. And hopefully they will roll out a Marketo version soon.
What struck me most when chatting with all these smart and passionate vendors was the extent to which the market is converging in terms of core functionality. All the tools work in more or less the same way and have a shared set of basic applications. The ability to build a predictive model in itself is rapidly becoming commoditized.
Where vendors are differentiating is around the use of that data for enabling sales and marketing to make better decisions. We are beginning to see some new and novel applications emerging, and this is where I would expect to see the most innovation in the months and years ahead. The practice of leveraging predictive data in the system of execution (marketing automation) is still in its early days - I believe the applications above only scratch the surface. This is exciting, because there’s lots of room to grow.
Share Your Use Cases
If you have figured out some advanced ways of using predictive analytics in your marketing automation or sales enablement efforts, I’d love to hear about it! Please share in the comments.
I’m grateful to the following people for spending some time to help inform this post.
Amanda Kahlow, CEO at 6Sense
Jessica Cross, Director, Demand Acceleration and Customer Marketing, and Jim Walker, VP of Marketing at EverString
Sean Zinsmeister, Senior Director of Product Marketing at Infer
Nipul Chokshi, Head of Product Marketing at Lattice
Kylee Hall, Senior Director of Marketing at Leadspace
Tony Yang, VP of Marketing at Mintigo
Disclosure: I work for Perkuto and EverString is a partner of ours.