8 steps to achieve anonymous website personalization
If you’re a marketer who relies on your brand’s web or mobile site for your outreach strategy, you probably are aware of the immense value a personalized experience can bring to the audience. But what happens when most site visitors don’t even identify themselves? How can your brand say “we know what you want” without really “knowing you”?
The Known Anonymous Experience
We are used to seeing personalized feeds on websites, e-commerce shops, streaming platforms, and travel portals. Have you ever wondered what goes into building such an experience?
There are two possible scenarios when it comes to user identity:
- Where you know the user’s profile, i.e. the user is signed up and logged in
- When you don’t know, i.e. the user is not logged in, and we know nothing except a few identifying breadcrumbs that standard analytics tools capture.
It is not uncommon that websites have many non-authenticated (having not logged in) visitors, in fact, most users never log in, if they don’t have to. Many sites belonging to news & media publishers, guest checkout enabled storefronts and corporate pages don’t even have a login mechanism, to begin with.
However, all users leave behind breadcrumbs that you can use to stitch together their behavioral profile. In this article, we will show you how a personalization engine powered by a CDP can turn these seemingly random bits of information into a gold mine of information about your visitors.
Remember your after-school visits as a kid to the local shop in the neighborhood? You probably picked up candies, snacks, and, if you were like me, groceries for mom. What started with a simple transactional exchange with no-frills evolved into a routine over time and a relationship was established. Few visits later, the store clerk knows who you are, at least by looking at you, what our preferences are, shares various offers that are available and you could even get that exclusive ‘credit’ when short on cash!
All these experiences are unique to us and the magical thing is, no one ever needed to find out your name, address or income. We are wired to see the patterns and signals — the shopkeeper sees what the customers are wearing, what they are buying, how they are talking and simply reacts to instincts. Everything comes quite naturally.
“Personalized experiences are common in the offline world”.
Wouldn’t it be nice to replicate our offline experience into our online worlds using all available signals that we have? This could range from minor touches like greeting visitors to highly personalized feeds, know that every bit counts! After all, the personalization that we experience in both the online & offline worlds should be natural and evolved from past interactions, because that happens to be a very “human” desire.
The Ingredients
We require some tools to build such an experience for the online world. Many of these tools are widely used by marketers already.
Website Analytics: Tools like Adobe Analytics, Mixpanel, or Google Analytics where we can export the data into some object store in plain text for analysis.
Trusted first-party cookie: Many tools set a cookie value when a user visits a site. Cookie validity is varied, it lasts from a few days to a year but this can be disrupted if the user clears the history & cookies. Also, cookies are browser-specific so if the user switches browser then the ID is lost.
NOTE: If you want to understand a little more about the differences between first-party and third-party cookies — what is being deprecated by the browsers and what is not, etc. — read more on our related Medium post.
Website Personalization tool: We use Adobe Target for content personalization, but any tool can be used as long as it supports external audience upload & it can match the audience in real-time using first-party cookie id. Tools like Optimizely and Google Optimize fit the bill as well.
Quaero CDP: The backbone of this entire solution where data is collected, unified and signalized. This is the place where we discover segments, identify patterns and then create audience activations
The Cookbook
Step 1
Configure website analytics to capture first-party cookies for each visit. This is crucial as the cookie id becomes the key identifier for repeat users. Analytics feeds generally contain behavioral traits like hit origin, reference, geolocation, search keyword, source and device type all of which we will use to build our ML models.
Step 2
Aggregate all the hit level data into the object store (like S3) by staging directly into the Customer Data Platform (CDP). Quaero CDP has out of the box connectors that let you ingest data from a variety of sources quickly. Make sure that as many relevant data sources as possible are connected, both online & offline, as this helps connect anonymous web sessions with known user sessions and, consequently, identify user traits better.
The next few steps take place inside the Quaero CDP.
Step 3
Filter out unwanted hits (bots, beware!), transform and then aggregate hit level data to convert it into session-level insights. Hits are individual actions while sessions are combinations of multiple page views in one browsing session. Quaero’s flexible data Lineage feature lets you perform these operations with ease.
Step 4
Try to resolve the identity of users using cookie id that is captured. This can be achieved by matching cookies with session data. If the same user ever logged in or filled the form, some PII information is captured. You can link anonymous visitors with known users using this.
For instance, we used Hubspot CRM to store leads and PII information. Hubspot sets a cookie for each visitor, called the “hubspotutk”. When a visitor fills the “Contact us” or “Subscribe ” form on the website, PII information gets attached to the cookie within the CRM system. We capture this Hubspot cookie and pass the value alongside the website hit data. Later, the cookie value is used to resolve the identity of the user using Quaero’s Identity Resolution engine. The icing on the cake or ‘cookie’ in this case.
Step 5
Run different predictive analytics on session-level data or use unsupervised models to reveal browsing patterns & group users. The output of this step should be an attribute or a labeled segment. Browsing pattern, over time, reveals insightful information that is used to create segments for targeting, like blog readers, prospective customers, career seekers, etc.
For example, the ‘blog readers’ segment emerged from behavioral patterns where visitors navigated to blogs & spent a significant amount of time reading our case studies.
Step 6
Build a content strategy and define measurement metrics. Once, we have the audience based on browsing patterns, we can focus on building personalized experiences. Here, we require personalization tools and their capability to define business logic. We used Adobe Target, but similar features exist in other tools such as VWO or DynamicYield. Suites like Adobe Experience Cloud help us import external audiences and orchestrate the experience in real-time.
The content strategy might start with simple questions such as
- Who are the end-users?
- What are their expectations?
- How do we define success — engagement or conversion?
The answers to these questions will help you develop a strategy. An example of one such strategy could be that a user is shown a Request for Demo form after 5 visits to the case studies page and over 15 minutes spent on the website You can use Quaero’s segment builder to create many such treatment groups.
Step 7
Push audiences (segments) to Experience Cloud i.e. Adobe Target so that we can use them to build the online experience. The Quaero CDP provides an out-of-the-box connector for Adobe Target. Discovered segments can be synced with Target via SFTP which is supported by Experience Cloud for uploading external audiences.
Step 8
Go live with the experience — monitor and let users enjoy subtle but unique experiences on the site. Personalization and experience building is an ongoing process, just like our childhood shop visits. Optimized experience is built over time but with this exercise, our future personalization is reduced to just four steps. Keep creating new experiences and use an analytics framework to see what’s working and what’s not. Further iterations could involve creating multiple user experiences, customized landing pages and multivariate A/B tests.
Congratulations — you’re now an agile marketeer, iterating & experimenting your way to glory!
The Known Experience
While this playbook is most beneficial for sites that have mostly anonymous visitors, personalization for logged in users is relatively more straightforward. You’ve probably seen this already in some shape or form.
- Taking back users to complete an abandoned journey such as cart checkout
- Personalize website page to user’s preset preferences
- Recommendations based on past purchases
As you might have probably guessed, a lot of these are driven by machine learning models working on much more rich datasets. Stay tuned for our next article on how you can achieve that with Quaero.
