Personalization is at the core of many modern internet services and has been a topic of heated discussion among researchers and academics over the last ten years. You may remember the Netflix Prize of $1M for teams that could boost the accuracy of the Netflix movie recommendation engine by 10%, or Facebook’s introduction of a personalized News Feed that drastically increased engagement on the site, both making headlines a decade ago. This article reviews eight common implementations of personalization technology, and makes the case that all companies should be evaluating this technology.
While major tech firms have been investing in personalization technology alongside Netflix and Facebook, only recently have a significant share of large enterprises started incorporating personalization and matching technology into their operations and products. Barriers to adoption have typically included difficulty in gathering the necessary structured data, difficulty hiring the right talent, and difficulty forecasting the expected return of these expensive, long term, investments.
The good news is that much progress has been made: the technology risk has minimized; the academic community has generated a lot of techniques and algorithms to support various kinds of personalization products; and most importantly, the business case is clear that personalization and matching technology can help organizations achieve high ROIs.
* According to the McKinsey & Co’s study, 35% of Amazon’s consumer purchases come from their personalized recommendations and 75% of what Netflix viewers watch comes from their algorithm recommendations
* According to the Aberdeen Group, personalized emails increase click-through rates by 14% and conversion rates by 10%
* According to Forrester Research, 54% of retailers reported recommendation engines as the key driver of average order value in customer purchases.
Read on to see examples of personalization in action, __connect with our AI team at Gigster__ to discuss a specific project where Personalization technology could drive value for your business.
## E-Commerce
### 1. Product Recommendations for Up-selling and Cross-selling
E-commerce companies deliver tremendous value to their users via personalization. A notable example is the __related product recommendations__, popularized in 2003 by Amazon. In this case, a user viewing a detailed product page additionally sees the products similar (cross-sell) or complementary (up-sell) to this product.
### 2. Personalized E-commerce Search
In this case, a user types a query and gets a ranked list of products relevant for the query and also personalized for the user. For example, if during the same session a user “touched” (visited) a brand elsewhere on the site, the search engine can re-rank the results and feature products coming from the user-preferred brand, or factor this information into deciding which sponsored result is most likely to be useful to the user.
Since search is a composite complex product, there are many places, where it can benefit from personalization. For example, on the image below, Staples presented __personalized query suggestions__ based on the initial keywords inputted into the query bar.
Interestingly, the same personalization techniques used for results ranking and re-ranking can be applied to provide more relevant query suggestions. For example, a user, who visited a specific brand page, can see this brand higher among the suggested keywords if it also matches the prefix of the query (e.g. “Ca…” could be completed as “Canon” and not as “Calvin Klein”).
### 3. Targeted Emails and Notifications
Apart from presenting relevant search results and recommendations on the website and in a mobile app, companies benefit tremendously from sending __personalized emails and notifications__. According to multiple studies, it __helps increase revenue, conversion rates, and engagement__ with the company offerings and products. In a nutshell, the problem always boils down to three questions: *__who__ should see __what__ product and __when__?*
## Media & Entertainment
### 4. Content Recommendations
Netflix is perhaps the most famous personalization and recommendation engine. However, almost any other media company could benefit from such a personalization and matching technology. For the users, the experience is very similar to the aforementioned __“Related Products” recommendations__. The only difference is that here instead of products, the service will __feature related content__ — — movies, images, news articles.
## Travel
### 5. Search Results
Now that we’ve seen various implementations of personalization technlogy, you should be able to spot a few familiar elements that are powered by personalization, all on the same page. First, we can see __personalized search suggestions__ based on location (“San Francisco”). Second, since we searched for “hotels for executives”, the *Promoted Result* shown at the top of the search engine result page does feature a hotel that is supposed to be for executives, rather than just the highest bidding ad.
It is worth noticing that the results are personalized even though the user isn’t identified (we didn’t login). However, the results would have been much more relevant had we done this. Very likely, the website would have shown the __hotels based on the past travel history, rating, and other inferred preferences__.
## Social
### 6. Feed
After Facebook introduced its feed in 2006, the concept quickly became the most popular design pattern for social services. Initially, when the social web wasn’t as crowded, the need in personalization wasn’t clear. However, over time, the amount of content has significantly outgrown the capabilities of humans to process it as more and more people have joined Facebook and other social services and shared their stories online. Unsurprisingly, personalization and matching technology came to the rescue.
For example, Pinterest asks its users to select several relevant content categories at the beginning of the journey to personalize the feed and __show the most relevant posts__.
Additionally, Pinterest personalizes its search, search suggestions, navigation, and emails with the featured content to add stickiness to the website.
Likewise, Twitter started __re-ranking tweets by relevance__ for each user based on the __past user engagement and feed content__ when it realized that the amount of tweets in each customer feed had exploded beyond a manageable scale.
### 7. Friend Recommendations
The power of a social network comes from the network effect and the connectivity of its members. Therefore, the intelligent social networks (e.g. Facebook, Pinterest, Twitter, LinkedIn, etc.) invest serious R&D efforts to __increase the connectivity of the social graph__. *“People you may know”* (LinkedIn) or *“Who to follow”* (Twitter) are great examples of this functionality as they facilitate more and more relevant connections between their use
## Recruiting, Staffing, and Human Resources
### 8. List ordering
Recruiters get a lot of incoming resumes from candidates, however, only a few incoming resumes are relevant since candidates often apply for jobs indiscriminately. Recruiters waste a lot of time and can miss great talent. Personalization and Matching technology helps __reduce the time recruiters spend processing incoming resumes__. Likewise, candidates are exposed to more __relevant jobs__ and save their __job search time__. For example, the Gigster Network delivered projects to help recruiters __prioritize incoming applications in the recruiter inboxes__.
## Email Inbox.. (just kidding!)
Now it’s time for an example of what a lack of personalization feels like. Think about your email inbox, and what it would look like __emails personalized and ranked by priority__. Personalization and matching technology (along with text mining, deep learning, and other AI technology) could be applied to solve this problem. Depending on the __email behavior usage patterns__, we could __re-rank emails__ for __every user and sender__ at a __specific point__ in time and __location__.
Recruiters can benefit from it by focusing on __more serious candidates__. Sales people can benefit from it by focusing on __higher-probability leads__. Executives can benefit from it by allocating their attention to __the most pressing company initiatives__. You name it!
## It’s all about adding value to the experience
While all of these examples leverage complex algorithm design, massive data sets, and immense compute power, they’re nearly invisible to people who use these services, and that’s the point. Personalized services add value to the user experience and when users have better experiences, businesses benefit too. Businesses of all sizes should be thinking about how to personalize their services to make user experiences more efficient, relevant, and useful.
