So, what exactly did Netflix get right?
Lazy Saturday evening. Looking for a Dramedy (that’s oddly specific). Burnt through Google’s generic Top 50 already. What next? Head on over to Netflix. ‘Recommendations for you based on your Viewing Preferences.’ Perfect! Good Will Hunting. Perfect again! Oh, look at that, Robin William is on the cover. Well done, Netflix. You got yourself a happy consumer.
- Made the consumer feel special with those recommendations – who doesn’t enjoy a good movie suggestion?
- Their suggestions were perfectly in line with consumer’s expectations – the one-time expectation vs reality wasn’t disappointing
- Positioned those recommendations perfectly – top of the page is the best way to capture one’s (maybe) toddler-like attention
- Captured attention with the right cover picture – no thank you to gut-hungry zombies, yeah?
A+ in terms of getting the ingredients right.
Simply put, Netflix hyper-personalized the experience using customer analytics data. What they understood from a marketing standpoint was that consumers enjoy special treatment, possibly because an overload of information is overwhelming. ‘Custom made is always better than mass produced,’ seems to be their MO.
Having said all of that, we wanted to test the waters ourselves and we thought a good way to answer the thriving question of whether people enjoy being told what they need and don’t need would be to ask some of them!
Survey for the employees of LatentView Analytics:
We conducted a survey and asked about 60 LatentView employees in the age range of 21 – 50 years, analytics and otherwise, a couple of questions to gauge their perception of marketing advertisements. Following is a sample of the same:
1 Question: On a scale of 0% to 100%, how close are the ads to your requirements at the time? For example, a person who is looking to take a vacation might say an ad for luggage bags is 75% relevant for his requirement at the time.
- A little over 90% of them thought that the ads shown to them are at-least 25% relevant to their requirements at the time.
- Nearly 46% of them gravitated towards the ads being 50% relevant to their needs.
2. Question: On a scale of 0% to 100%, how often do you enjoy these product recommendations?
- About 80% of the population enjoy the product recommendations at-least 25% of the time.
The internal survey indicates that hyper-personalized product recommendations are definitely catching speed; let’s delve a little deeper to understand the math and science behind it.
What is hyper-personalization?
Hyper-personalized marketing strategies are based out of an extremely deep understanding of customer analytics data. Thisthat comes from analyzing every consumer’s search history, their preferences, proactivity, engagement etc. They aim to tailor their suggestions to the consumer’s requirement.
Consider this:
- An online retail store with a consumer base of 30000
- They have a website to bridge consumers and vendors
- Their homepage is quite exciting with a recommendations section for their users
Quite simply put, hyper-personalization in communication would mean 30000 different versions of the homepage section, one for each of its consumers as opposed to maybe, 2 or 3 variants in total in a different scenario. It might sound like a stretch but how it is achieved is quite incredible and something we will take a more discerning look at.
Why is it called hyper-personalization and not just personalization?
Personalization aims to make the user feel included by using the consumer’s name, location or purchase history. However, it may not be engaging enough for the user to follow through on. On the other hand, hyper-personalization in communication also takes into account browsing behaviour and changes recommendations to adapt to changing consumer needs.
Take a look at these two examples:
Example 1:
Hey Rachel! This end-of-season sale is bigger than ever! 30% off sitewide.
Rush before we are all out of stocks
Example 2
Hey Rachel! Those XYZ brand snow boots you wanted last week are finally on sale!
30% off sitewide! Hurry before stocks run out so that you’re all set for the upcoming winter! Also, here are a few other things we thought you might be interested in to get ready for winter next month:
- ABC’s Cherry Burst lip balm
- XYZ’s suede winter gloves
- MNO’s leopard print scarves
Example 1 is personalized whereas Example 2 is hyper-personalized. Clearly, Example 2 is more likely to get Rachel excited because an item she wanted but had probably had forgotten about, is on sale. In Example 1, it is quite possible that she didn’t quickly recollect what she had been looking for and ergo, probably thought the ad was irrelevant and didn’t make an effort to visit the website.
Why is hyper-personalization as coveted as it is?
- Good way to capture attention and improve conversion:
Every message has about 8 seconds to capture the attention of the consumer. For it to mean anything, it must be useful to the consumer and this happens only when it is in line with their needs.
According to a recent report, nearly 79% of the consumers only like getting coupons, offers or promotions based on their earlier engagement with a brand. They do not appreciate an overload of information and get frustrated when the brand is unable to connect with their needs, interests and dislikes. It’s also been said that Call-to-actions (CTAs) convert 42% better when they are personalized as opposed to generic CTAs. Further, with the help of personalization, customer acquisition costs are slashed by 50%, consumer spending is improved by 500% and revenue is lifted by nearly 5-15%.
2. Increased customer loyalty and fewer returns:
In a recent Forbes survey, it was found that nearly 49% of the people indulged in impulse purchases. They bought items they did not really require but only 5% returned those items and 85% of them were happy with their purchases. Further, 44% of all consumers said they’ll likely come back after the personalized shopping experience which speaks to the loyalty they possess towards the brand.
3. Personalization influences what people buy:
Research shows that 59% of the consumers are influenced by the recommendations made to them. They tend to discover new product through research and advertising as well but 96% of the consumers expect the retailers to tell them what’s best for them.
The five components of hyper-personalization in communication:
There’s no doubt that the above numbers are impressive — but they require a well-constructed plan to achieve them. There are five operational components to build a winning hyper-personalization marketing strategy. They are:
- Data Collection
- Personalized messages
- Customized offers
- Timing
- Testing
Use case: The Netflix hyper-personalization model
1. Data collection:
Netflix gathers data about the user preferences using customer analytics data. In the digital world, ‘clickplay,’ gives us much more detailed information about a user than just their gender and age.
2. Personalized messages:
The biggest personalization in Netflix is the rows of shows a user is shown. On average, people check 40 to 50 options before they choose what to watch. This is majorly based on the user’s watch history.
3. Customized offers
Netflix creates multiple different landing cards – the images that are shown as people scroll through shows – for each of its titles. It shows someone the art cover for a show they are most likely to click on.
4. Timing
Netflix customizes its recommendations based on when a user is watching to determine the optimal time to deliver personalized messages and increase desired response during the user’s next login.
5. A/B Testing
Netflix runs 250 A/B tests each year. This A/B testing strategy These tests presents users with two slightly different experiences to understand how they respond. If one version gets more people watching a show, it may be rolled out across the whole service.
Skeleton of how Netflix chooses the content to present to its users:
For artwork personalization, the online learning framework used by Netflix is ‘Contextual Bandits.’ What does this mean? Simply put, Contextual bandits are a class of online learning algorithms that trade off the cost of gathering training data required for learning an unbiased model on an ongoing basis with the benefits of applying the learned model to each member context.
The goal of Netflix’s personalized recommendation system is to show the right titles at the right time to each member. That being said, the bigger question is, how can they convince the user that a title is worth watching? The solution is to provide the visual “evidence” for why the title might be good. If the perfect image is presented on the homepage (as the wise once said, – a picture is worth a thousand words), then maybe, just maybe, the user will give it a try.
For example,
- Best artwork for each of the members to highlight the aspects of a title that are specifically relevant to them.
Different images for different themes in the show rather than a single image portray.
1.Personalize the image based on genre preferences of the user
Someone who has watched many romantic movies may be interested in Good Will Hunting if the artwork shown contains Matt Damon and Minnie Driver, whereas, a member who has watched many comedies might choose the title if the artwork contains Robin Williams, a well-known comedian.
2. Personalize the image based on casting preferences of the user
A member who watches many movies featuring Uma Thurman would likely respond positively to the artwork for Pulp Fiction that contains Uma. Meanwhile, a fan of John Travolta may be more interested in watching Pulp Fiction if the artwork features John.
Challenges involved in hyper-personalizing content:
- Understanding whether the particular art cover presented ultimately influenced a member to click (or not click) on a title or if the click would have happened regardless.
- Identifying if changing artwork between sessions has an impact on the user and his/her motivation to click.
- Understand how artwork performs in relation to other artwork shown in the same page or session.
Just to complete the picture, here are few other big names that depend on personalization and customization in marketing:
- Spotify: It relies on the users listening habits to deliver the right music recommendations. It studies an individual’s music choices and cross-references it against the music choices of other individuals with similar taste. Finally, it makes suggestions music from one’s playlist to the other.
- Amazon: Amazon has access to past search history, time spent on the page, items in the shopping cart, items reviewed and rated etc. It exploits this capability to send hyper-personalized emails with specific references to brands, product types etc.
- Starbucks: Starbucks uses app interaction data to customize the menus for its consumers. Their customization in marketing strategy takes it one step further and sends emails with offers in line with what they would be interested in.
In conclusion, it is rather fair to say that mass segmentation is a thing of the past and it is now up to companies to embrace the beauty of data-driven retailing. It does not have to be an expensive and daunting undertaking and it could start off with something smart, small and effective and evolve into something fruitful to the organization that will keep consumers coming back for more.
Hyper-personalization in the age of the “attention economy”
For brands that are vying for attention with each other in a digitally evolved economy, hyper-personalization can make all the different from ‘maybe’ to ‘yes.’ Leveraging the power for data and analytics, LatentView Analytics helps you gain an advantage when it comes to hyper-personalization to help your brand stand out and increase engagement and conversions with your target audience. To know more about LatentView’s end-to-end custom analytics solutions, please write into: marketing@latentview.com
