Machine Learning
7 Steps to Get Ready for Machine Learning
May 14, 2018 Mike Thompson

Per The CMO Survey from February of 2018 , 61.4% of companies aren’t yet using Machine Learning, but about one-third of companies have begun investigating or have initiatives in-progress.  Fewer than one-third of companies surveyed have no near-term plan to explore the use of machine learning.  We know firsthand that organizations which have successfully implemented ML report improved conversions and greater insight into their data.  

That said, there is a good reason to be skeptical.  While machine learning is consistently touted as the next big thing in marketing, results on the ground have been varied.  The vast majority of marketing technology vendors simply don’t have subject matter experts on staff. Much of the resulting work is haphazard, relying on third-party ML services to get it right.  

If we’re to believe the hype, all you need to do is switch on some ML services and in an instant high-value interactions will become obvious, and clusters of engaged visitors will appear. You’ll be able to automatically tailor personalized omnichannel communications and deploy them at the optimum time for an open or a click.  If only it were so...

“There’s a mass influx of newcomers to our field and we’re equipping them with little more than folklore and pre-trained deep nets, then asking them to innovate.”

- Ali Rahimi, Google


 

He’s right.  

So the lay of the land is this: Roughly one-third of companies are using ML with varying degrees of success.  Another 40% have initiatives kicking off.  Of all the companies that are wading into machine learning for marketing, only some will find success and competitive advantage.  What differentiates the organizations that are able to successfully use ML?  

Since we’re data nerds here at Coria, we thought we’d share some thoughts on how best to prepare for your ML initiative and help ensure success.  

 

 

1) Collecting Web Analytics

Web analytics is the first step of collecting data for a website. The data can be used to understand the interactions a site visitor has across your digital channels. Analytic tools collect data from your site through a small piece of code called a code snippet. This is usually added to your site through the development process, depending on the content management solution you are using. There are some solutions like Sitecore that have the ability to track interactions without the need to deploy a code snippet.

The basic information collected by a code snippet is the number of pages viewed, time on site/page, pages viewed, number of visits (cookie enabled), and referring channel. This information allows your marketers to create reports that provide insights into how your visitors are interacting with your website. 

Channel traffic example:



By implementing the following steps, you’re marketing team will have access to additional data points to make decisions from. 

 

 

2) Augmenting Web Data

Augmenting the data collected on your website can foster a better understanding of your clients and prospects. By adding additional information to your site visitors’ profiles, you can enhance this understanding. This can yield valuable data to reveal possible intent and allow you to engage them with relevant content. 

GeoIP data is one service providing this information. GeoIP is a form of geolocation that can provide you with information of the location of the computer an individual is using in a database that ties an IP address to a location. 

If you are a company selling products at a retail location, this would allow the site to show the nearest store to a site visitor’s location, saving the site visitor from having to search or input information into the site. This increases the likelihood that they will covert on a sale.

In addition to the location of a computer, there are solutions that provide enhanced details that can assist in building a profile of the site visitor.

A service that can assist in augmenting your data is Maxmind GeoIP. The services provide additional data to further populate a site visitor’s profile with a location, company name, and domain name.  If your company is B2B you can gather company names on anonymous visitors to provide your sales team with potential clients to target.

There are additional solutions that provide SIC Codes, Industry/Vertical, company size and other valuable information. Later in this article, I will talk about personalization and explain how this information can be used to engage a first-time site visitor. 

 

 

3) Deploying campaign tracking

Accurately attributing the influence of a site visitor to a campaign helps a marketer determine the benefits and ROI of that campaign. Analytic tools like Google, Sitecore, Omniture, and others allow marketers to create a token to be added to a URL. 

Below is an example of a Campaign link created in Sitecore, the marketing analytics solution we use.



By adding this token to the end of any URL within our site, we can attribute the traffic back to the campaign. The token can be used in an email, or in social channels, shared links, hyperlinks, etc.

Campaign Tracking functionality allows you to examine granular analytics on your various campaigns. This allows you to understand the channels visitors are coming from, the content that is driving engagement, and the areas to focus on based on success or failure. 

 

 

4) Adding metadata to content

Metadata can provide detail to the content created on your site.  Two examples of metadata could be the author and the category. Categories can help define the area of focus for a content item.  The example below is from this blog post; as you can see, it can include multiple areas of focus.

 

This information can be used to filter user based content through the site search and to gather analytics on the most commonly read categories.

 

 

5) Creating custom facets 

Custom Facets provide the ability to further categorize your campaigns, goals, and assets so that logical attributes can be assigned to your items for reporting purposes. Tagging your goals with specific facets allows you to track how frequently contacts are converting on goals which are segmented based on the areas that make sense for your business.

Assets can be broken down further into the following custom facets:

  • ebook
  • white paper
  • worksheet
  • video

 

 

6) Personalizing the web experience

The average time people spend on a website is less than three minutes, with most people leaving after around 10-15 seconds. Site visitors want to find what they want quickly, or they will be gone.

Utilizing personalization on your website can lead to increased year over year revenue by up to  30% (Aberdeen Research). Personalization allows your site to listen to the queues the site visitor is giving and bring the right content forward to them. 

You may be asking “What queues?” 
1) Paid advertising from google
-Provides the keyword searched to get to your ad, and personalize the content to associate it with the keyword
2) GeoIP 
-Provides a testimonial of a client that may be well known in the area
3) Industry
-Showcases studies of the industry relevant to the visitor

If you are interested in learning more about personalization, check out our eBook here.

 

 

7) Utilizing Testing/Multivariate

Testing content can be as simple as an A/B test. You can have two variants of a component on your site, one could be with red writing and the other blue. Solutions providing this capability will track which one leads to the highest rate of conversion. More advanced multivariate testing could include multiple components on a page, or the entire page. 

Example of a form test: 

Short form Long Form

The longer form is less likely to convert and have less trailing engagement.

These 7 steps allow your company to collect additional data points. Using machine learning to enhance these seven steps will allow the algorithms to learn from the data they are receiving and predict future outcomes. Without this foundation, machine learning will not provide the value you are looking for.

If you would like to discuss where your company is at today, we would love to grab a coffee and discuss how a detailed digital roadmap can drive you to Machine Learning. Contact us!

 
Mike Thompson
Sales