What else can you do when you’re already crushing it?

You and your team are likely already familiar with your website engagement figures – whether through your eCommerce platform or Google Analytics. 

What are the daily traffic numbers? How many customers are repeat buyers? How many site visitors are returning visitors versus first time visitors? You have dashboards that report summarized metrics, allowing you to keep a gauge on what’s going on. 

As a result, you are probably also quite aware that the vast majority of website visitors are anonymous to you – never filled out a lead form or purchased from you – and have no phone number, email, or CRM record.

This is at least 90% of your traffic, but for many eCommerce sites it can be 98%+ anonymous!

Beyond the tracking of your traffic – I’d also wager that your team is running A/B tests on your website regularly to determine the highest-converting messaging, recommendations, content, (color scheme!), etc to optimize the throughput of your web store. 

Additionally you’re likely deploying an array of other tactical optimization techniques and higher level strategies… which might be the topic of fun (or maybe stressful) team meetings!

Our goal here is to share another, more granular method to optimize your eCommerce store conversions for those anonymous site visitors. 

The not-so-secret secret is this: use web analytics data to create micro-segments for each individual prospective customer to match with, and then tailor your website experience to each shopper based on predicted affinities. 

Anonymous visitors are still data-rich

There are several data sources that can be used to identify, segment, or optimize for anonymous traffic like this.

The first, and most obvious, is through some kind of web form or questionnaire. If a shopper has filled this out – with or without an email address – you have usable data to make experience alterations thereafter. Obviously at this point they may no longer be anonymous, but any input besides contact information can be put to use.

The second – most prolific, easy, and immediate – is the contextual and browsing data that is generated and left-behind by every person on the internet. Most of this is captured by Google Analytics (GA). Hundreds of attributes are captured automatically and immediately, while others can be collected by creating custom tracking events in GA or Google Tag Manager.

Finally, there are third-party data sources and technologies that will do cookie matching or provide an ocean of widely-sourced data that can be used to create profiles and run matching algorithms on. Products like Oracle BlueKai have thousands of segments defined where you can create custom targeting to match to lookalike anonymous visitors.

Our focus here will be using the second category: data collected upon website entry and throughout website engagement by a web analytics tool. This is your first party data, generated on your site, and free to use immediately.

Web Analytics Data

Google’s support website provides a list of the data points that can be extracted from Google Analytics (in this case through Google’s Big Query product, if you’re interested).

There are a lot of items in the list!

If you’re thinking what I’m thinking, then I think that you and I are both thinking that this is awesome. But sincerely, we have a lot of gold to work with here.

To summarize some of that data into categories, we have: 

• Date, day, and time of visit
• Referring site URL
• Geography and network location
• Web browser type
• Device type and operating system
• Campaign and other source information (UTM parameters)
• Dozens of other technical attributes, such as Flash version, JavaScript support, screen resolution, etc.

In reality, the article above confirms Google Analytics captures literally hundreds of individual data points.

So while you don’t have the shopper’s contact info, product preferences, past purchases, or feedback responses… you will know all of that above information before your homepage even loads for the shop visitor.

Of course once that person actually interacts with your site – scrolling, clicking, visiting specific pages or content, applying filters and preferences, selecting features, etc – all of that behavioral information is also there to be collected by GA. 

In short, you have large buckets of information about anonymous shoppers to work with!

Next up: how to use that anonymous shopper data to improve conversion rates on your store. 

Getting to work

All of your known customers who have purchased from you, plus all of your leads, have data profiles. This is built from their buying journey, contextual info, survey or other submissions, and behavioral data. 

Through analysis, you can determine correlations between those known customers’ profiles and the preferences or attributes that lead to purchases.

In essence, this exercise defines the characteristics about a customer from earlier in their journey that are statistically more likely to lead to certain types of actions down the funnel. 

For example, shoppers that came from certain ad content and engaged early with multiple product categories are more likely to buy just once – versus different initial behavior from a shopper may correlate with many longer term repeat purchases from your brand.

After that exercise, your next step is to match the data set for anonymous website visitors (from Google Analytics) to your historical data. You can create a similarity matrix between individual anonymous persons and your previous customers. 

So, what does that mean? 

After mapping and conducting statistical analysis, you’ll be able to determine the “lookalike” anonymous shopper to some sub-segment of customers. 

Getting from data & analysis to action

You wield great power when you know the relationships between real purchase trajectories and the web behavior upon which they were predicated. 

That power allows you to serve up the messaging, product recommendations, or experiences that you’ve seen work successfully in the past. Whether mimicking the old, or testing for new successful pathways.

Further, as you tailor experiences for shoppers you may notice some content versions or product recommendations aren’t converting as expected. The new, real-time success/failure data can be factored back into the analysis to improve the next time. 

Thus you aren’t creating an optimization model from only past data, but also from in-the-moment lessons to keep pace with how consumer trends adapt.

As you may already be seeing, there are challenges in all this: running that complex statistical analysis; having accurate and valuable output from that analysis; and then doing real-time matching of the current anonymous site visitor to the right set of personalization elements. Plus doing it all at scale.

To solve that challenge, you’ll need to deploy an intelligence engine.

Using machine learning models, real-time data ingestion, and API integrations, personalization engines like Jarvis ML can do the heavy data processing and update website pages dynamically. 

With this in place, your team can display landing pages, hero text, banners, product recommendations, and other content tailored to each specific anonymous user. And all based on the data that’s already being captured.

Applying machine learning to increase conversions

Machine learning (ML) thrives in an environment of large data sets and predicting uncertain behavior. Why? Because it is not explicitly programmed to follow rules-based logic.

Instead, ML figures out the best path to optimize for your goal: conversions. The software is equipped with data (as much as possible) and statistical models. It’s situated within scalable infrastructure so it can calculate decisions, push that intelligence live onto your site, and then improve from each success or failure. 

For instance: a shopper may arrive on your store landing page via a specific paid social ad: which has a certain copy, creative, source, etc. Plus they are associated with a device, operating system, time of day, geo location, and so forth. (All of that GA data we talked about earlier fits in here.)

ML will determine probabilities of different behaviors based on tons of input – impossible for humans to do with any scale. It will run calculations in real time and adjust on the fly to create a continuously evolving decision engine. 

Based on the intelligence generated from your data, this example shopper will get a web experience that matches their predicted affinities and purchase trajectory. The next shopper one second later may get very different content and recommendations!

The premise of this whole strategy is predicated on the lesson that customers convert at a higher rate from personalized experiences. That piece has been widely understood and utilized for years.

So what’s the big picture?

The key takeaway is that you can use data about anonymous website visitors to hyper-personalize their experience on your online store.

It’s a narrative we hear often in different ways, but we wanted to provide a tactical lens to break down how it’s really working in order to help to paint a clearer picture of what’s possible.

Needless to say, this real-time, dynamic personalization based on predicted affinities and recursive learning is not something you can throw together one afternoon. That’s why tools like Jarvis ML are designed: to give you an automated, data-driven personalization brain that will engage and delight your customers. 

With the right technology and all your 1st party data, your team can capture in-the-moment signals to instantly match a prospect with their purchase trajectory – then use that to adapt web page components to that specific anonymous user: product listing pages, hero banners, promotions and recommendations, calls-to-actions, and more.

Unlock that 90% of web traffic: from anonymous shoppers into high-converting, highly engaged customers. Your data is the key. And an engine like Jarvis ML can help you find the right door.