Personalization and Why It Is Different from Optimization, Predictions and Recommendations
Since the dawn of the browser, people believed that the internet would deliver a digital experience smart enough to know our interests and wants. However improving the relevancy of the browsing experience is still in its infancy and we are still only at the beginning of understanding how to create and deliver personal experiences online.
Since the dawn of the browser, people believed that the internet would deliver a digital experience smart enough to know our interests and wants. However improving the relevancy of the browsing experience is still in its infancy and we are still only at the beginning of understanding how to create and deliver personal experiences online.
But the solution space generally referred to as “personalization” is crowded and there is a widespread state of confusion between the different types technologies available. That makes perfectly sense, as there is a sea of providers that more or less pitch the same story of the “personal user experience”.
When you dig below the surface, you’ll find that “personalization” actually falls into four categories of solutions, each with specific use cases:
Optimization
Optimization, also known as A/B and multivariate testing, determines the best possible presentation of something to a broad set of users. Typically optimization rotates a series of variations in front of an audience of users over a period of time until a statistical significant winner is declared.
An example would be trying to determine which copy pulls the most clicks for an advertisement. Marketers can typically run these tests over and over with little complex coding or back-end work required.
The key takeaway is that optimization doesn’t seek the best presentation for each individual user, but rather, for the audience as a whole. Optimization is great when making decisions about website or app layout and design, shopping cart conversion workflow and pathing, and advertising design and layout.
Recommendation
Recommendations mean selecting items an individual user might be interested in. Most recommendation platforms are based on collaborative filtering technology. Recommendations algorithms build profiles for each item in terms of the most frequently viewed or purchased items when compared to other items.
Approaches for making recommendations tend to fall into one of three categories:
- Popularity: All things being equal, recommend the most popular (or most likely to be engaged with, or most profitable) items.
- Item-based: Given an item, identify similar items (e.g., users who purchased X also purchased Y).
- User-based: Given a user’s interests or past behavior, find similar users and recommend items those users have expressed interest in.
Recommendations are used to personalize the content of communications, they engage users and make them feel like the brand is tailored to their unique interests and desires. This is best-achieved with algorithms that are user-centric and leverage a rich history of user engagement behavior collected across devices over time.
For example, the algorithms may find that a large percentage of shoppers who buy a certain set of shoes also frequently buy a particular handbag and therefore will recommend those items together. Same technique can be applied to content for publishers in order to generate the most clicks possible from the total traffic on the site.
As with optimization solutions, recommendation solutions form suggestions based on behaviors across a large group as opposed to tailoring results for the individual user.
Recommendation technology is well applied to the state where users are researching their options and discovering relevant combinations of offerings. But in many cases recommendations in the later state of a purchase or conversion process may actually hurt conversions by confusing the user with alternative options that increase drop rates out of check out flows.
Predictions
Predictions are about anticipating what an individual user is likely to do in the future. The most valuable and widely used marketing predictions fall into these three categories:
- Value: How much value will an individual user create in the future for the brand? This is a combination of purchase propensity, order value, retention and profitability.
- Transaction: If a user makes a transaction, when will it be and how much will they transact for?
- Engagement: How engaged will a user be in communications in the future, and how will this vary by channel?
Predictions make a quantitative statement about a user’s likelihood to take a future action. For example, one user might have a 34% chance of purchasing in the next 30 days at an expected value of $35, whereas another user might have an 11% chance of purchasing in the next 30 days at an expected value of $103.
Personalization
There is a big leap from optimization and recommendation solutions to personalization solutions. The reason is the collection of data from the browsing behavior of each individual user. Generally this is collected through the use of tracking scripts. Personalization solutions build a comprehensive profile of each user over a period of time, and in some cases, create detailed profiles of all the content a publisher is producing, as well.
These content profiles serve as additional inputs in user profiles. For instance, understanding the context of a set of content can inform the profile of a particular user who likes to consume content about sports, stock prices or politics. The profile also includes information about the time of day, browser, location and device of an individual user. Some also merge the browsing profile of an individual across multiple devices - a technique called profile stitching.
This profile information is the baseline for marketers to create a personal experiences for the online user. Such experiences may include content, promotions, recommendations, advertisements or cross channel interactions.
Marketers can integrate this profile data to customer relationship management (CRM), email and advertising platforms to further expand the potential use cases. Continually pushing updated profile information back to the email marketing platform enable most organizations to benefit from improving the relevancy of their email. Basically by transitioning from generic high volume scheduled sends to individually triggered emails with highly personalized content. It also enable much improved timing by being individually triggered emails when the user is known to be online.
Personalization requires a high level of sophistication on the backend to support tracking, assembling and maintaining individualized profiles as well as injecting conditional content back into the user's browser - all in real time.
Conclusion
While optimization, predictions and recommendation technology will enable you to incrementally improve your general performance on your site - only personalization enable a leap in performance and an radical new box of opportunities. Most are still to be explored, but the ability to treat your visitors as individuals require that you track and build deep profiles of your users and are able to trigger content based on this profile in real time.
Building your personalization strategy it is paramount not to confuse what technology delivers what, as most organizations eventually will require a mix of the technology described above.
Kresten Bergsøe
Monoloop