Personalization: What You Need to Know Right Now (Part 1)

How can you measure the success of personalization? How many user segments should be used? Is real-time personalization a current reality—or a far-off dream? In this two-part Q&A series, we ask a senior member of our data science team to explore the trends, challenges, and possibilities of personalization.

How are brands measuring the effectiveness of their personalization efforts? Are they using attribution?

When it comes to site optimization, experimentation provides the most direct measure of lift of the metric of interest over a holdout control during the same time period. This is a great place to start because when you look at this channel only, attribution isn’t necessary and causality can be inferred from the experimental results.

Moving forward, when combining the digital experience with print and other forms of advertising, attribution modeling would be necessary to understand the relative ROI of personalization efforts within each channel.

Is it important to demonstrate a quick win when ramping up a personalization program?

It’s important to provide a proof of concept that aligns the business on what is technically possible using the available toolsets and the strategic direction of the personalization campaigns, coupled with the long-term ROI potential.

Buy-in from the business stakeholders will provide the initial funding for the personalization campaign and will sustain the program even in the face of potential short-term setbacks. In taking this approach, there will be less pressure for quick wins.

That said, you should always start small, with a data-driven focus on easily executed strategies that have the potential for a large impact.

What channels and devices are brands prioritizing?

Generally, most brands have higher conversions on desktop channels than mobile, so that’s a great place to start.

However, most analytical approaches to personalization (and personalization technologies) make it relatively simple to contemplate device type as an applicable dimension in the personalized experience. When the device type is a contributing factor across all dimensions of the data, it could be valuable to vary the experience across devices.

How are brands managing experiences across channels?

If you view device as a channel, the newest technology allows for almost seamless visitor stitching across devices, as long as you can eventually authenticate your users with a common identifier.

So if a visitor has authenticated on a desktop and then opens an email on a mobile device, the visitor—and all past mobile actions—will be combined with the desktop profile, allowing for an enriched profile that joins the actions taken across those two devices.

With that holistic view of the customer, a broader array of actions can feed the personalization algorithms, which will allow for more intelligent and complete personalization campaigns.

Do you see a movement towards personalizing based on real-time data, or are brands not there yet?

There are still some technology hurdles in front of true real-time personalization. In most approaches available right now, any action has to be registered as an event, added to the user profile, and finally be made available for use in targeted experiences.

Depending on the tool and how quickly the user is navigating the website, real time (meaning current or next page load) can be difficult. Simpler, and therefore more commonly executed, personalized campaigns are directed to the first page load of a new session for any returning visitor.

How many user segments tend to be used for personalization?

Ah, trick question. The data will tell you!

Basic segmentation approaches fall short of providing a reasonable solution, because these methods are dependent on an analyst making an (inherently arbitrary) decision as to which and how many dimensions of the dataset are most relevant.

A better approach is more algorithmic: an analyst uses unsupervised learning techniques, such as clustering or latent class analysis, to determine the number and definitions of each group. The algorithm will only partially answer the question—the business will need to determine if it’s profitable to actually direct distinct campaigns to each group.

Do brands tend to treat segments substantially different for personalization? Please give examples.

Advanced personalization efforts—like clustering or latent class analysis—will leverage unsupervised learning techniques to identify each group. They’ll also apply some business sanity checks to estimate the potential ROI of creating distinct campaigns to personalize to them.

An algorithm, for example, may unearth a very unique and well-defined segment, but if it is small—in terms of either total number or revenue impact—or if it’s composed of visitors who are inelastic to marketing efforts, that group should be discarded at the analysis phase.

The best programs use analytics to identify existing groups and then predict how profitable each group could be if it received a personalized experience.

How are segments managed across mobile, email, the web, and other channels?

Ideally, the tools used for personalization should be omnichannel and support automatic and near real-time stitching of visitors across devices and channels.

Visitor stitching across digital devices will become more easily obtainable as strategies and incentives are developed to facilitate interaction across devices—and ideally, these strategies will lead to user authentication.

When looking at joining offline data, you must provide a compelling reason for the customer to provide his or her information at the point of sale. This information can then be uploaded and combined with the online profile, using a common identifier like an email address.

Each segment or group should be inclusive of the channel and device type via a master visitor profile. When this occurs, segments or groups won’t have to be managed separately, because device and channel become dimensions in the master profile of each user.

Are brands leveraging real-time segmentation?

It depends on what you mean by real-time segmentation.

Often, “real-time segmentation” is used to refer to the practice of creating the rules that define the inclusion or exclusion of a particular user into a particular group. I don’t recommend this type of real-time, automatic rule creation, because it removes the benefit of layering business strategy on top of the algorithmic results.

Ideally, real-time segmentation describes the practice of auto-populating groups of users based on near real-time actions. The approach is beneficial because it acknowledges the reality that users can quickly move through different levels of engagement. The technology necessary to accomplish this is available and is being used.

For example, you typically don’t want to personalize an experience aimed at increasing initial brand awareness for someone who already has an item in his or her cart and is entering to the checkout stage of the purchase funnel.

Are you seeing brands personalize at a 1:1 level or at a segment level? Is moving to a 1:1 level worth the effort?

Brands are still focusing on personalizing at a group or segment level, because it produces the most profitable results. I disagree with the notion that 1:1 personalization is actually an attainable strategy.

1:1 personalization implies that for each of n infinite visitors, the brand needs to approach an infinite number of distinct combinations of products, messaging, and user experience to meet each user on that individual level.

In the process of doing so, each group will become smaller and smaller, approaching an infinite number of groups with each “group” consisting of only one user. With that comes the ever-decreasing marginal utility of this subsetting exercise due to the costs of unique products, the expense of unique creative design, and the digital burden of modified user experiences.

There is a sweet spot associated with personalization. How far down the path toward individualization you proceed centers on questions such as:

• Can you reach each group in a contextually appropriate manner with a relevant product?
• Are the groups truly unique?
• Will the groups respond differently to different campaigns?
• Are the groups large enough or profitable enough to personalize to?
• Can you measure the impact of personalizing to all groups?

Read part two of our Q&A with Reid.

Reid Bryant is a data scientist at Brooks Bell. He uses advanced analytics and applied statistics to create data models, refine methodology, and generate deep insights from test results. Reid holds a Master of Science in analytics from the Institute for Advanced Analytics at North Carolina State University.

—

Brooks Bell helps top brands profit from A/B testing, through end-to-end testing, personalization, and optimization services. We work with clients to effectively leverage data, creating a better understanding of customer segments and leading to more relevant digital customer experiences while maximizing ROI for optimization programs. Find out more about our services.