When it comes to ecommerce, make it personal

September 2021

E-commerce has been on the rise in the last few years paired with an ongoing process of digital transformation and technological innovation, from data-driven marketing to headless commerce and machine learning that can predict what consumers will buy next. The growth of e-commerce was further amplified during the pandemic, where consumers faced with lockdown measures were driven online to shop. In the US alone, retailers saw an increase of 44% in 2021 versus the previous period, while in Europe it’s estimated that e-commerce sales grew two to threefold.

Today’s customers expect it fast, easy and relevant

On the consumer side, this growing trend of online shopping behavior has skyrocketed user expectations for a seamless, more convenient and fast experience as well. Consumers expect brands to be available wherever they are online, regardless of device or channel. And they want an experience that’s tailored to their needs, whether product recommendations, content or features.

Moreover, a survey that Kameleoon, an innovative A/B testing platform from France, conducted in the US, UK, France, Germany and Italy concluded that of all participants, 61% are actually disappointed with the user experience they have when they shop online. So, there’s a huge opportunity to win over customers with an enhanced experience that meets their expectations. Also, since only 2-3% of all website users convert, that leaves 97-98% of traffic that can potentially be converted into buyers if you solve the problems that may be stopping them from doing so.

What’s ecommerce personalization?

Personalization focuses on delivering customized website or app content to specific audiences in order to meet predefined goals. For instance, our goal could be to increase the conversion rate for users who come from Google shopping (we identified high bounce rates for this segment). After analyzing the user behavior and context for this scenario, we can hypothesize that these users need to easily browse through products to find what they want, which may be a similar experience they have in the Google shopping search results page. To test the hypothesis, we personalize the product detail page for this specific segment by adding the related product carousel in a prominent position above the fold.

If you’re starting with testing and experimentation, we always recommend optimizing the experience starting from macro conversions – purchase, newsletter sign ups, demo requests. Then focus on the micro conversions, which are more glandular actions that lead to the macro conversions, such as watching a video, performing a search or adding products to the cart. This is a process that starts with simple experiments that gain complexity overtime. The A/B testing roadmap should consist of three phases: macro conversions, micro conversions, and finally personalization. Essentially, this means it doesn't make sense to invest in personalization if the overall site performance is less than optimal.

Where do we start?

When it comes to ecommerce personalization, we always start with identifying segments that could potentially benefit from customized experiences and who of course have a high potential for conversion and revenue generation provided they’re served with the right content.

Tools like Kameleoon, Monetate and Google Optimize 360 offer different features and capabilities for managing segmentation, with Kameleoon delivering both manual and automated segmentations. The manual segmentation is based on pre-defined parameters that might combine demographics, geolocation, traffic source and onsite behavior, for instance. An example for manual segmentation could be male users based in Berlin coming from Facebook ads and who view the ‘yellow shirts’ PDP. For this segment, you could show an urgency label saying, ‘Guaranteed delivery within 24 hours in Berlin’.

Automated segmentation is based on machine learning. In this case the tool analyzes and correlates website user data, which gives the tool the capacity to predict outcomes based on specific actions performed in real-time using a score system. An example of this could be three types of segments: those who have a 50% probability of conversion, those who have 70% and those with 90%. You could have a different message for each segment – the first could have a discount message, the second an urgency label and the last an upsell message.

You also can provide 1-to-1 personalization through software such as Bloomreach or Kibo. These tools offer powerful advanced merchandising and search capabilities driven by algorithms that take both user and business data into account to display and organize content to each customer individually at scale. In product personalization, the engines prioritize the SKUs that are most relevant according to user search, past behavior and overall product performance.

Taking it a step further, you could implement ​​omnichannel personalization that provides customized content for users wherever they buy (device, channel) using real-time data to deliver 1-to-1 experiences in online or physical storefronts, and even customer care in call centers.

What data points are available?

You can connect different data sources like Google Analytics and others to generate audience segments. The most powerful option today is the Customer Data Platform (CDP), which is a tool that collects user data from different sources and unifies them by adding a unique ID layer and making the data accessible to other software, one of them the personalization tool. Whatever CDP solution you choose, it’s crucial that it allows for real-time personalization. Examples of CDP are: Zeotap, Tealium, SalesForce Audience Studio and Exponea (now called Bloomreach CDP).

In addition to user data, you can and should make use of contextual data such as day, time,  weather and geo-location.

Personalization tools also have their own data management systems, which most often are based on cookies. However, as you probably know, Apple has permanently banned cookies through its Intelligent Tracking Prevention (ITP), and so did Firefox some time ago. In order to solve this, in the case of Kameleoon they rely on Local Storage to identify user segments without having to make timely server calls to retrieve user information. With local storage the data stays in the browser and therefore does a better job at both privacy and user experience.

What are potential technical issues you need to address?

The most problematic issue that affects ecommerce personalization is called the “flickering effect” caused by a delay in loading the experimentation script – the script may load after the page has already fully or partially loaded in the HTML. This is bad for both UX for the obvious reasons, and SEO, since it affects your page’s core web vitals.

To solve this problem, prioritize tools that offer anti-flickering technology that will ensure the elements that appear in the HTML page load at the same time and without latency – both the original code and the A/B testing content. Installing the testing tool directly in your website’s source code, other than via a Tag Manager, is also recommended. If flickering persists, then you may need to consider different ways of conducting the experiments, for instance split URL testing.

Finally, besides avoiding flickering effects, always make sure to optimize your baselines for site speed, which means compressing images, using CDNs, ensuring your servers are located optimally in relation to users, etc.

Creating a testing culture

In order to significantly improve customer experience, organizations need to adopt a testing culture which prioritizes constant experimentation and design iterations. This process is called conversion rate optimization (CRO). Through A/B tests and experimentation, we identify roadblocks specific user segments have and test hypotheses to solve these problems. By solving those problems for users, we’re able to increase conversions, revenue and give a reason for people to come back. 

Implementing ongoing (CRO) and personalization in ecommerce requires building a dedicated team and including various stakeholders from different areas to collaborate closely together to plan, activate and measure experiments.

This testing culture translates into a shared belief across the organization that continuous improvements and interactions are a must, and that ultimately, relying on empirical data to drive decision-making across product, design, communication and development is a natural part of how things are done. It also means that the organization puts customers first in everything that it does by investing in research methodologies, tools and people.

Lengthy and difficult discussions involving stakeholders who may look at projects from entirely different points of view, mostly from where they’re standing, happen all too often. However, if you prioritize data over opinions, the discussions are much easier to contain and decisions are made more quickly. After all, data doesn't lie.

Illustration: Justyna Dybala

Related links