Myth Busting: Understanding Profile-Based Segmentation Analytics vs. Intent-Based Predictive Intelligence

 In Predictive Intelligence

A fairly regular question I get when speaking with B2B marketers interested in predictive intelligence is how 6sense is different from everything else out there. Inevitably, the conversation leads to a discussion about prediction theory and what data is truly indicative of a prospect or account being in an active buying cycle. It goes a little something like:

“Okay, I get how you are different, but I’ve spoken to other predictive vendors that provide a score using thousands of individual data points about a company. These vendors claim that they have ‘intent’ data, and that ‘intent’-data is NOT as useful because its difficult to gain useful signal from so much noisy data.”

This is a discussion I love having because the answers aren’t always obvious. And every time I see this discussion play out to eureka moments for marketers, it makes me that much more passionate about everything we’re doing as a company at 6sense. To provide my perspective here, I’m going to break down the claims made into smaller parts so we can address each specifically.

  1. “Don’t all predictive vendors provide the same predictive score based on thousands of data points about a company?”

So let’s say you are a company that sells cloud-based security software. Using basic rules of thumb that you’ve come to understand about your business, you sell to companies that are:

(a) based in the United States because your product caters to US requirements;

(b) your sales team is setup to focus on East, West, and Central regions of the country;

(c) since you are a premium provider, you only sell to companies that typically have greater than 2000 employees because they are the right size where a need for your product becomes evident, and compliance becomes a challenge they need to solve; and

(d) you know that only companies that are “cloud-friendly” – i.e. use Amazon Web Services, Google Apps, and such – are the ones that will typically consider buying your cloud based security product.

Now Acme Inc. is a company in your database but isn’t a current customer. Acme turns out to be a San Francisco-based technology company, with 3500 employees that uses Amazon Cloud Services. Bingo! Or is it?

Every marketer at this point will tell you that you have to ‘nurture’ the account that shows the best firmographic characteristics via marketing campaigns to test for need and interest. Sales will tell you they hope the timing is right and that the account is in an active buying cycle when they reach out and try to connect with a decision-maker. So you’ve identified that the company fits the criteria to be a target at this point, even though you haven’t established need or timing.

But didn’t we start out with defining the ideal customer above using some basic rules of thumb? So in effect the approach of look-alike modeling via a solution that can understand a ton of firmographics of an account and automate the math would able to accomplish segmentation analytics at scale. This type of modeling, called ‘look-alike’ modeling, basically looks at firmographic patterns and answers the question: “How similar is a company to those I’ve sold to before?”

Wait, but what if you have thousands of company firmographic data points for the company – would this be predictive or more useful? Well, it depends. Many of the data points would be highly correlated with each other, adding very little incremental ‘predictive value’ as such and often cause collinearity issues in certain types of predictive models, which lead to model instability. Such data is more interesting in the context of micro-segmentation for things like competitive take-out plays and broad-base targeting. This is definitely insightful segmentation data for sales and marketing, but not really predictive and actionable from a timing perspective as it doesn’t allow you to predict that an event will happen at a specific time.

Takeaway: Segmentation analytics are NOT predictions.

  1. “Doesn’t every predictive vendor have intent data?”

All predictive vendors have some kind of intent data, but not all intent data is created equal, and not every vendor’s approach to using it is the same.

There are some sources of commodity data available for sale by certain data providers. The challenge with these datasets is that the data is typically aggregated, thus losing it’s fidelity and granular view into topics, keywords, and companies. It’s one of the reasons we’ve seen multiple predictive vendors talk about their struggles with deriving value from intent data they’ve tried to leverage. Time-based intent modeling is a hard problem, and many enterprises that choose to attempt a DIY project eventually throw in the towel and opt for buying a solution.

At 6sense, we’ve spent more than 5 years in technology exploration and building out the 6sense Exclusive Data Network. This includes a variety of third-party data sources that span exclusive publisher relationships, blogs, online communities, buying-guide websites, social interactions, ad targeting networks and search, that feed our models with a level of data that is not available for purchase from any data provider. It’s the reason we can offer unique products such as SearchSense, and why we feel real predictions should go far beyond the lookalike modeling and profile-based segmentation that others focus on.

Takeaway: Intent data comes in many flavors that determine usefulness; it’s not all created equal. 

  1. “I’ve heard that intent-data isn’t useful in predictive intelligence, because the buyer’s journey is too complicated.”

The changing buyer’s journey is a topic every B2B marketer has had to grapple with as more content has moved online, more web destinations became available for buyers to do their research, and more prospects come well-informed with the knowledge garnered from digital efforts. The buyer’s journey has morphed into being completely non-linear, and not to mention done by many people within a company Vs. just one individual buyer. Forrester represented this new buyer’s journey in an interesting graphic back in 2012:


This non-linear and multi-user journey is exactly why account-based marketing (ABM) has seen such a tremendous increase in interest, and why trying to draw out the most signal from this buyer’s journey becomes critical. Said a different way, company intent becomes far more important than company firmographic and contact demographic data. The changing buyer’s journey has compelled us to launch a new product, 2sense, which attempts to offer enterprise companies the benefits of this timing insight based on their web log data.

Further more, a company has limited visibility into a prospect who comes to their website and fills out a form because what is missed includes:

(a) all the activity the prospect did before getting to your website;

(b) all the activity done by 90% of your site visitors who remain anonymous and will never raise their hand and fill out a form;

(c) the fact that there are many people in a buying committee performing multiple research activities anonymously in multiple places.

To make predictive intelligence work and detect prospects that are in market earlier in their research cycle, it’s necessary to consider these scenarios within data mapping and data science modeling.

Takeaway: Non-linear buying journeys and anonymous activity make intent-data and time-based modeling critical. 

  1. “But, I’ve been told that intent data is all noise and no signal. True, right?”

Imagine if you could take advantage of a buyer’s digital footprint as described in the section above on their non-linear journey, even slightly more than you believe you do today, would that be valuable? Almost every person I speak with would shake their head vigorously, because the fact is, ‘known’ visibility today is a drop in the ocean when comparing it to volume of anonymous B2B research events that take place across the web.

So while it’s hard, and intent data does have a lot of noise, there are ways to extract usefulness for detecting active buying cycles.

6sense chose to work on this problem from inception, and our patented technology is all about building time-based intent models that actually ‘predict’ that an event will happen at some point in the future. This is where we truly excel, and ultimately our goal by focusing on time-based intent modeling is to answer a different question that segmentation analytics cannot answer: “Which companies are in-market to buy my products and when?”

Takeaway: Intent data carries a tremendous amount of valuable signal; you just need the right technology and the right data relationships to make it useful. 

As a technologist and product manager, it’s exciting to work with our customers and the entire 6sense team to keep pushing the envelope of what’s possible in marketing and sales technology. I personally enjoy explaining these concepts to a growing segment of marketers that believe predictive technology and data-driven approaches are the next wave of their internal marketing technical stack innovation. And it doesn’t hurt that our approach to intent-based predictive intelligence is driving some amazing results for customers.

So how are you thinking about your own marketing and sales technical stack and its upcoming evolution? Please reach out as I would love to hear your perspective on the differences between segmentation analytics and predictive intelligence and on your own journey with data and predictive technology.

Recent Posts

Start typing and press Enter to search