Enabling One-To-One Marketing With Hyper-Personalization

Webinar Transcript

Avantika: Hello and welcome to this webinar of WebEngage. My name is Avantika Pandey. I’m the senior content marketer at WebEngage. I’m your host for today’s webinar. Before we begin, here are a few housekeeping topics. First, you’ll be on the listen-only mode. Second, we’ll be recording this webinar, and the recording will be made available to you in a couple of days. Third, we’ll have a Q&A session at the end of the webinar. So, if you have any query during the session, you can put your questions in the comment box below and we’ll address them later. For those who don’t know, WebEngage is a Marketing Cloud that helps businesses engage and retain their existing users. More than 45,000 companies trust us with the cross-channel marketing automation. If you want to know more about us, you can check out www.webengage.com.

Now, let’s get back to the webinar. A couple of weeks back I was having a conversation with Priyam, who’s also the presenter for today’s webinar. And we were discussing our experience with personalized communication, how it influences our buying behavior. As the conversation progressed, we realized that there’s a big gap between how brands like Netflix, Amazon, Starbucks, and Spotify are communicating and how as compared to other brands. So, that’s how we can see this idea for today’s webinar. We wanted to discuss how these brands have shown sort of personalization maturity and whether we can learn from them. So, now I’ll let Priyam take over this webinar. Over to you.

Priyam: Hello, everyone, and welcome to this webinar on hyper-personalization. I am going to be your presenter for today, and I will be talking about how the top brands in the world are enabling one-on-one marketing with hyper-personalization. This is going to be the agenda for the day. First, we will be discussing the difference between personalization and hyper-personalization. Of course, it’s a prerequisite. Then we will be talking about the need for hyper-personalization today and what is happening in the personalization scenario with brands right now. The way ahead with predictive personalization and we will get to my favorite bit, which is industry leaders and how they are using hyper-personalization. Let’s look at few of the brands. There is Amazon, there’s Starbucks, there’s Netflix, there’s Spotify. We will be talking about all of these brands and how they have approached hyper-personalization today. Finally, we will end with a small note on the future of marketing, which is going to be all about context.

Let us begin. Right. So, the difference between personalization and hyper-personalization. I’m sure a lot of you must be wondering, what is the difference or is there any difference? What is personalization? Well, the definition says that it is creating a tailored experience where users’ needs understood and met faster by studying previous interactions. Now, I would like to explain this with a simple example. Consider a coffee shop in your neighborhood which you frequent. You go there and you ask the barista to get the regular order for you. Now, your barista is aware of your regular order. It is a cappuccino. It arrives on your table with your name written on it. That is personalization in its base form.

Your experience with the coffee shop was personnel based on your previous purchase history. This led to the barista identifying the order correctly because he had enough data about you. This guy always comes in at X, Y, Z time and always orders a cappuccino. So, there was minimal interaction, but you still manage to get your favorite cup of coffee. Let’s look at hyper-personalization now. The definition says that it is boosting contextual relevance by customizing a product or service offering according to a user’s personal preferences, buying, and browsing behavior. So, you can see there are a lot more interesting elements involved here. I would like to explain this with this same example.

You are back at your favorite neighborhood coffee shop. You ask for your regular order. Now, however, your barista checks the purchase data from your last 20 visits. The data shows the user, which is you, ordering cappuccino in all of those visits. It also reveals that 18 out of 20 times the coffee was served with a double shot of cream, chocolate syrup. This analysis makes your barista acutely aware of your preferences, therefore, he modifies your order accordingly. Instead of a regular Cup of Joe, you get a completely customized cup of coffee according to your tastes and preferences, which is your double shot of cream, your chocolate syrup, and your name on it. That is hyper-personalization. It is all about being acutely customer-centric so that it remains the only goal in mind. Here, the coffee shop had enough data and it was integrated with technology, like machine learning and marketing automation, to create a deep profile of the user.

Now, let’s move on to personalization and marketing. The idea behind personalization is to create a unique, engaging, and meaningful experience for each individual user, right? With a personalized campaign, the consumer feels more connected, you feel more drawn to a product or a service because the communication is more direct, more personal, addressed directly to you. Brands leverage the power of the first name. In marketing campaigns, it is very, very common to see the first name being used very liberally. It compels users to pay attention because it is your first name, you identify with it.

So, now that we have talked about personalization and marketing and why it is important, I would like to talk about the psychology of marketing…the psychology of personalization. I’m sorry. Why do people prefer personalized experiences? Why are brands so crazy about personalization? I’ll tell you why. There are two main reasons. The University of Texas recently did a study on this very subject and zeroed down on two factors. The number one factor is the need for control. When a user sees something that is different, which is not the same as something which is available for the masses, has been customized specifically for their tastes and preferences, they get a feeling of control. Why? Because this has been tailor-made for them. Only them in that moment of time. It gives them a semblance of control.

Now, this might not be true, actually, because you may be seeing a personalized homepage somewhere because the service provider chose to show you that. But still, you get a perception of control. You feel that this has been made especially for me, I am in control of the situation. And that is the main factor behind the psychology of personalization. The second factor is information overload. Users are aware of the fact that there’s too much of information out there. How do you sift through all of this data and actually find something that is relevant to you?

So, if you actually personalize something and the user feels that the information presented is completely tailored to their suites and preferences…their needs and preference sorry, the actual process of sifting through tons of information to find something relevant and useful has already been taken care of. Now, please keep in mind that all of these factors only come into play when the user is provided with certain visual cues. You have to let your users know that the content or the experience has been personalized.

Let us move on now to the Coke campaign which was The Share a Coke campaign. Now, Coke played a lot on the psychological triggers with The Share a coke campaign. Back when it was launched in 2011 in Australia, Coke did a research and found out the most popular Australian names and they put it on their products. As you can see in the image over here, it has names like Amy, Bobby, Zach. The idea behind this was to capture the excitement around the first name. This campaign was so popular for Coke that it helped the brand climb out of a 10 years sales slump. In a week, the brand increased sales by 30%. In 2013, the brand managed to get 28% more consumers compared to the same time period in 2012. So, a personalized or rather a hyper-personalized campaign was actually helping a brand like Coke acquire more users. So, people got intrigued when they saw a can with their name on it. Even none Coke drinkers grabbed a can. What do you think about this Avantika?

Avantika: You know what, Priyam, my friend Bella found a bottle with her name on it and without even thinking twice, she just bought it and posted everything on social media. What’s interesting to note over here is that before this campaign, or before buying this bottle, she never preferred Coke.

Avantika: Exactly. So, this was something that I read about. And a lot of non-Coke drinkers also ended up picking up a bottle or a can because the presence of the first name was something that compelled them to grab the product. So, moving on to need for hyper-personalization in marketing. Why do we need to personalization? Generic personalization tactics are being blatantly overused. I’m talking about those generic emails that you get with your first name in the subject line. It probably worked the first few times, but what next? How are you going to actually hold the user’s attention? Attention spans are getting shorter, try eight seconds for size. This is an actual true statistic. Eight seconds is all it takes for us to actually hold the attention on one particular thing and then we move on.

How are you going to ensure that your user is going to be compelled to take action in eight seconds? Today’s consumer is hyper-evolved, hyper-connected. Their purchase decisions are heavily reliant on user reviews and ratings. The average individual in the UK uses up to four devices to access the internet. They are hyper-connected. There is so much of information out there that users actually don’t feel like trusting ads and this is a true Google statistic. People actually prefer opinions, reviews, and recommendations from actual users, friends, and family. You know why? Because there is a trust factor involved. They feel like a actual user will be able to give them a better recommendation. Their friends and family, who know them very well, would be able to suggest something, which is tailored to their needs and preferences. Users want brands to give useful recommendations for them.

We expect brands that we interact with to know everything about us. We want brands to actually predict our behavior. Give us recommendations that we actually going to need in the future. We don’t want them to be stuck in the past. We want them to actually think about us. Value, I’ve mentioned value a few times. And what is value actually? But value is nothing but experience right now for the new user. Band experience is held in extremely high regard. And if you have a possible product or service, then you are going to end up losing users because you will not have any value. Your users will immediately jump to your competition that provides a better overall experience.

What is happening today in the personalization scenario? Let’s check it out visually. Here are current scenario of personalization in marketing. As you can see from this graph, on the left-hand side, we have revenue and on the other axis, we have personalization maturity. Now, most bands today are hovering in the area of maximum basic segmentation practices. You ask any brand, “What is personalization for you? And they’re gonna tell you, “Oh, we create segments, and on the basis of the segments we send campaigns.” Really? How well is that working out for you? Not very well. Exactly. You have to move to behavioral personalization. You need to understand your users deeply by studying their online activity and their offline activity. You need to standardize the brand experience for all of your brand touch points, and that is omnichannel optimization.

Now, as you keep climbing the graph, you see there is a point called predictive personalization over there. You see the brands that are hovering in that area? Yep, these are the actual top heavy hitters that we are going to talk about, Amazon, Starbucks, Spotify. They know…they have hedged their bets on predictive personalization because they know it is the future of marketing. We’ll touch upon all of those things in the next few slides.

Yep, my favorite part begins. We’ll be talking about the industry leaders who are using hyper-personalization. We’ll start with…yeah, these are the four brands, Netflix, Starbucks, Amazon, and Spotify that we will be covering in this webinar. And we will be starting with Spotify. The world’s largest music streaming platform. Yes, ladies and gentlemen, they have 71 million paid subscribers, 159 million active users. They’re in 60 plus countries right now. Its closest competitor is Apple Music and they are at 30 million subscribers, less than half of what these guys have right now. Spotify has personalization at its very core, right from music discovery, to playlists, to suggested artists. Spotify has been created, keeping personalization in mind.

Revolutionized music discovery with the UI. The UI used to be, like, a blog-style interface, but it has now become a regularly updated personalized newsfeed type interface. Something like social media platforms. We’re going to see that in a bit. Yep, this was back in 2006, 2007. As you can see, the layout is pretty old. Slowly, they changed. They had a big visual UI update. Everything went black for them. They wanted to create a movie theater experience where once the screen goes black, the focus is entirely on the screen. That is what they wanted to do. They wanted people to actually put their focus on the music and the artists and the album art.

This is the most current screenshot. This is the current iteration of the Spotify UI. And as you can see, they are all playlists, right from Discover Weekly to Haitian Heat to Teen Party, everything is connected and all of them are playlists. This shows that playlists that are personalized are absolutely huge for the plan. We’ll check them out now. Personalized playlists. Spotify has Discover Weekly. Now, how many of you have used Discover Weekly or Spotify because us Indians haven’t had the chance to do that yet? What do you think? My colleague Avantika feels very strongly about this.

Avantika: Of course, I would love to be one of the playlist junkie of Spotify, but we have to do with Apple Music as of now.

Priyam: So, yeah if Apple, you guys are listening to this, you guys have a very loyal user in Avantika. Yes, the success of Spotify is Discover Weekly. Now, what is this Discover Weekly? It is a 30 song playlist that is launched every Monday, and it consists of songs that are appealing to each unique user’s tastes and preferences in music. Now, Discover Weekly was a game changer for Spotify. The company actually took stock of the situation after the heavy success of Discover Weekly, and they started investing in algorithmic playlists. Now, these playlists are personalized at scale for users. We will be talking about how Discover Weekly actually personalizes. But we have to understand that there are diehard fans of this feature and personalization has been really important as a brand building factor for the brand.

The next one is Spotify’s Release Radar. Now, this is another playlist which is powered by a data-driven algorithm that spins up to our playlist. It consists of completely new tracks from artists that users already listened to. It’s updated every Friday. What is the only difference? This is brand new music. The songs in these playlists are latest releases, and they have been hyper-personalized into playlists for each unique user. How do they do this? More of Spotify’s technical wizardry. It takes into account the entire listening history of the user to come up with suggestions.

Finally, my favorite Spotify is Your Time Capsule playlist. It consists of a 30-song playlist made up of songs from the past unique to each user’s individual tastes. Spotify calls it “The soundtrack for a trip down memory lane.” How poetic. Right. Now, we move on to how Spotify is actually personalizing. We will be talking about the things that Spotify has done to actually personalize music. They acquired this company called “The Echo Nest” in 2014. Echo Nest is a music personalization data provider and they acquired them for 50-million users.

It was instrumental in helping them get 5 billion song streams in the first 10 months of Discover Weekly launch. Echo Nest is basically using algorithms that analyze text plus audio data. It uses machine learning and a web scraping technique in a combination to build up music databases. This has been instrumental in Spotify’s personalized playlist building strategy. Yeah, how does discover weekly work now? Now, Spotify analyzes listeners browsing and listening history and combines it with popular hot music to create a playlist of 30 songs every Monday. Discover Weekly looks at data that is around six months old to come up with a playlist. The two most important aspects that affect the personalization engine are music stream data and playlist activity data.

The algorithm has been so accurate that it probably knows a user’s taste better than their own friends. It has been the major factor for its stickiness. Now, how has hyper-personalization been a factor? In less than a year since launch, Discover Weekly had 40 million unique users. To put it into perspective, at that time, Spotify itself at about 75 million active users. So, about 50% of total active users were trying out Discover Weekly. So, this was absolutely the game changer for Spotify because, in the first 10 months, 5 billion songs were streamed by users, hardcore personalization, hyper-personalization.

Now, we’ll be talking about the three approaches to personalization that Spotify uses, collaborative filtering. Now, this is a little technical, so please bear with me. This approach analyzes the browsing and listening behavior of users to generate data about popular music and curate recommendations for different people. So, Spotify user’s implicit data, like, the number of songs streams, visiting the artist’s page on Spotify after listening to a song, adding the track to a playlist, etc., to understand the like and dislike factor.

Let’s say there are two similar users in terms of behavioral and personal attributes. User one likes track A, B, C, and D, and user two likes track B, C, D, and E. Now, both users like track B, C, and B, right? So, based on this collaborative filtering logic, Spotify will now suggest track A to user two and track E to user one, which were the only unique songs that both of them hadn’t heard. So, there is a high probability that both these guys are going to like these songs. Why? Because their tastes in music match. Now, this is user to user collaborative filtering.

Next is natural language processing. This system analyzes textual data from tracks metadata, articles, blogs from the internet to find out what words people are using to describe a particular song or artists. That’s crazy, right? Now, there are several thousand associated keywords reaching every day for artists and songs. I mean, we use these words to actually describe music that we listen to, like, “Amazing”, “Awesome Killer”, right? So, Spotify uses all of these words, and they call them Cultural Vectors. Now, all of these keywords, these Cultural Vectors, carry a score, which is correlated to the probability of someone actually using this word to describe that song. So, this vector representation of songs of this data is used to find out similar tracks. Like, for example, you call track A as amazing on a social media channel. And another track, track X, is also called awesome on let’s say a blog.

Now, Spotify’s natural language processing will assume that these songs are related because the same keyword has been used to describe them. The third is audio track analysis. This technique uses neural networks and complex algorithms to understand what an audio track sounds like. It generates data for it so that a computer can understand and use it to match and classify it against other songs. For example, a Korean pop band has a song called covenant, and a Swedish dance metal band also has a song called “Covenant.” Now, that’s a huge problem because they’re completely different types of music. Also, they are relatively new tracks.

So, the neural network will actually carry out an audio track pattern analysis to find out what these songs sound like. Of course, both the songs are different genres, so the audio patterns are different. Therefore, Spotify clubs these songs to the related genres. Easy peasy, right? Not really, there are absolutely billions of data points, and Spotify has been able to do it at scale. By combining all three of these approaches to music personalization, they have created an absolute powerhouse of music recommendation engine.

Yes. Now, we will be moving on to Netflix. My favorite. So, I have just started binge watching on Netflix. Avantika has been a regular user of Netflix. Avantika, what shows are you watching on Netflix?

Avantika: Right now I’m watching “13 Reason Why” season two.

Priyam: Oh, so you are on season two, wow.

Avantika: Yes, I am. What about you?

Priyam: I am a diehard “Stranger Things” fan. I bought an actual subscription because I love the show so much. So yes, let’s see what Netflix is doing. A quick run through. It’s the world’s largest OTT media streaming platform with 118 million subscribers. About a billion hours of video content is streamed on a weekly basis, guys, so pretty amazing. They have a really advanced recommendation engine, which we will be going to talk about in the next slide. Personalization is their main USP. Ever since they were a mail-order DVD business, they have been personalizing movie titles for people. Now that they have about a few billion data points more, they are doing it at scale digitally online.

More than 75% of site activity is driven by personalization. I mean, wow. That is absolutely ridiculous numbers and that just goes to show the power of hyper-personalization. Retention master class. Netflix has solved the rabbit hole problem. They have so much of data on their platform. And still, users are able to find out what they want to watch very easily. How do they do that? It is very easy to get overwhelmed on Netflix, believe me, but still, their hyper-personalization techniques are one of the best out there. Let’s check them out. AI-assisted recommendation engine has cut costs for them.

Personalization has actually helped Netflix bring down their costs and they have got about $1 billion worth of savings. You know how? So, people in the company believe that because of personalization, people are actually hanging on to the platform for a lot longer. They keep finding new shows to watch with their personalization engine, and they never cancel their subscription. So, without this personalization engine, because of customer churn, they would have probably lost $ 1 billion every year. That is pretty amazing.

Right recommendations is equal to more time spent on the platform. That’s what I was talking about. Now, there is a very simple logic over here that they have found out. On an average, a user spends about 60 to 90 seconds on the platform to try and find a good show. They might review about 20 titles, three of them in detail, or a couple of screens. Beyond this point, however, they will start to lose interest, and they will move out of the platform. So, they have actually solved this problem where they actually…where users end up finding their next show in under 90 seconds. Amazing. And only 20% of video plays, by the way, come from search. Seventy-five to 80% of videos are from recommendations.

Yes, Netflix relies on an absolutely massive amount of user data to understand taste and preferences. Say that you belong to a tiny island nation somewhere, and you enjoy watching anime. Now, Netflix’s algorithm will be able to match you with the entire anime-watching community in all of the 190 plus countries in the world. What then happens is you get awesome custom anime recommendations suggested by people who are just like you, anime lovers.

Their content rating system has changed. The earlier had a five-star rating system, but now they have so much more data that they have simplified this. It’s a simple thumbs up or thumbs down system. You like something you give it a thumbs up. If something sucked you just give it a thumbs down. Yes, this is what I was talking about. This is a screenshot from my actual dashboard. And after I am done watching a show it says, “Did you like it or did you not?” On the basis of that, they create a suggestion for you.

Right. Let’s look at a few personalization cues. This is a screenshot from my Netflix dashboard. I have highlighted a few personalization cues over here. Bang, bang, and bang. You see, because I’ve watched “Peaky Blinders,” because I’ve watched the movie “Bride,” they are going to suggest a list of movies which are going to be very similar to the ones I watched. They also find out the top picks for me right over here. So, they have so many visual cues that actually make my job easier. What do I watch next? “Not a problem, we’ve got it covered.”

Netflix combines the personal and behavioral attributes of every user. This is the beginning of creating a deep user profile. Basic general statistics like name, age, location, device, and more complicated attributes like online browsing activity are combined. Multiple events are considered to generate recommendations. What are considered customers’ viewing? What has he viewed? search, scroll, ratings data, as well as time, the date, the device being used, all of this is analyzed, and then it is layered on top of tagged content to create the content recommendations. Now, what is tagged content? Content is classified into separate groups. Everything that you see on Netflix has been tagged with metadata. Metadata is nothing but a brief description that is used to describe the content. What is this movie about? What is this show about?

Now, this is used to match content with other tags. So, this helps Netflix generate similar kind of content that a user might be interested in. Now, this user behavior is analyzed and cross-referenced with content tags. In 2014, Netflix had about 76,897 all genres. What is an all genre? It’s an alternative genre. Like, this was actually used to classify content into different categories. Let’s say Avantika watches a movie which is a comedy and right after that she watches a movie which is an action. So, now Avantika can be classified into an all genre, which is a mixture of comedy and action. It can probably be calm action. So, Avantika will now get recommendations that both have metadata containing comedy and action in it. So, this data is used with the users past viewing behavior on the platform to create highly relevant recommendations and predictions. It is as simple as that.

More user data generated is equal to better recommendations. Over 100 million hours of content is being consumed every day on Netflix. It generates a lot of data and the more data that they generate helps them understand users better, their content better, what works for who, all of these things. Now, I’ll give you an example. When Netflix launched their first original show called “House of Cards,” they wanted to do a really kickass job of promoting it. So, what they did is they segmented the lead actor Kevin Spacey’s movies, and they highlighted all of the people who had watched Kevin Spacey movies in the past on Netflix. Then they found out the people who had enjoyed watching the first few episodes of “House of Cards.” They combined both of these users and created a super segment. They kept pushing relevant messages to these people in the super segment and they were highly personalized.

Something like, “Hi, Avantika, ‘The President Will See You Now,” season one, episode five is now streaming on Netflix.” Now Avantika, who is a very big fan of “House of Cards,” is absolutely going to go crazy because it’s so personalized. It’s directly related to her viewing habits. And it ended up becoming an absolute hit for them because people were sharing everything on social media and it became really popular. So, that is one way by which Netflix use data to promote a show and made it absolutely viral.

Now, we move on to the next slide, Starbucks. Everybody loves Starbucks. I am pretty sure the whole world of Starbucks, including me and Avantika. How is Starbucks doing what it is doing? Now, I technically fee that Starbucks is more like a tech company than a retail food and beverage. Now, I was talking to Avantika about this and she absolutely used the same phrase. She’s like, “They’re like a tech company from Silicon Valley. They’re not a food and beverage company.” Now, they have been at the forefront of technology. They have stuff like connected coffee machines. The number one mobile payment system in the world. They are actually using technology to further the brand experience.

Now, they’re focusing on creating a unique, engaging digital experience for customers. They’re bridging the digital world, which is the app ecosystem with the physical equals system of Starbucks coffee shops, their digital flywheel strategy. We will be giving you a visual outlook of that. It’s nothing but four factors which Starbucks feels is going to be the game changer for the brand, and is going to be responsible for growth in the next 10 years. They are the rewards, the personalization, the ordering, and the payment system. We’ll talk about that in the next slide when you guys can see it visually.

Lastly, the growth of the rewards program. Starbucks probably has the most successful loyalty program of the past 10 years. Starbucks really stuck the nail on the head with this one. They grew their loyalty program from 5 million to almost 15 million in no time. The whole focus is on AI personalization and digital focus. Repeat usage is actually rewarded with things that people actually value, you know. So, they actually understand what people want and a lot of loyalty programs had that missing. Now, let’s take a look at the digital flywheel strategy. This is taken directly from a presentation that the Starbucks CEO had made at an investor conference. These are the four aspects that Starbucks is going to be focusing on in the coming 10 years. The first one is their rewards program. They are going to be basically incentivizing every single purchase that a person makes.

Personalization, their offers, their communications, their messaging is going to be customized for each individual user and this is going to happen at scale. The payment system. They are absolutely going to simplify the payment system and they have done that. Starbucks now has the number one mobile payment app, the mobile payment system in the world and ordering. They have decided to revolutionize ordering, and they have managed to cut down lines in their physical stores with their mobile app system. All of that in the next slide.

So, this is the basic strategy that Starbucks follows. For their regular customers, the intent is to upsell or cross-sell. Basically, they want these people to either buy something which is of a higher value or buy a complementary product. So, they do that via personalized food and beverage recommendations based on past purchases and their preferences. For occasional customers, the intent is different. They just want people to visit them more, buy more stuff from them. So, they do that via their loyalty program. They give star rewards. They give a lot of other discounts and freebies to actually hook these people in.

How do Starbucks personalize their marketing? They revamp their personalization engine. Back in 2016, personalization at Starbucks was limited to sending 30 variants of email per week. It was based on data that was two weeks old. It doesn’t sound very impressive, right? In under a year, Starbucks revamped their personalization engine making it capable of sending over four lakh hyper-personalized email variants per week. One-to-one marketing at immense scale. Earlier, the email communication that was sent was mostly based on historical data of users. Not anymore. In under a year, the company managed to tweak its AI-based personalization engine to create completely unique truly one-on-one offers for their users based on their preferences and behaviors.

So, the email that I get from Starbucks, and the email that you and Avantika and everyone else gets, is going to be completely different. No two offers are going to be the same, no two beverage recommendations are going to be the same. That is the beauty of it. How are they doing this? They use data to analyze and create deep user profiles. Not very surprising as we have learned. The brand records data at multiple touchpoints, in-store and mobile. Right from basic information about the user, which is the name, the age, location, to online plus offline behavior, items browsed, items purchased, maybe even the weather data. All of this is combined and leveraged for the brand’s personalization engine. They send personalized targeted recommendations on the basis of that.

Predictive personalization. This is my favorite because I’ve actually experienced this. Starbucks sends a lot of personalized food and beverage recommendations, both on the app and even at their offline stores when you go to visit them. Now, the rewards program that they have has moved from number of visits to amount spent. You get an incentive on the basis of the amount that you spend. So, the more you spend, the more number of incentives and rewards you’re going to get. Now, this has led to the creation of so much of data because users have ended up buying a lot more from Starbucks. So, they know that the user has ended up buying a lot more products, that’s a lot more data, that’s a lot more behavioral activity. So, the potential of offering more personalized offerings, because the user spending has increased, has actually been a boon for Starbucks.

You’ll be surprised that they don’t have a single algorithm for their entire user base. It’s a very complex system that resembles a neural network. It connects with several data points to understand each user’s behavior and then it comes up with a relevant suggestion. In short, it is a data-driven, AI-powered algorithm tailored for each individual user’s behavior and preference. Like I mentioned, it is unique for each user. It analyzes each person’s individual taste, preferences, behavior, what they have had in the past, and even weather patterns and third-party app data as one of the people in Starbucks have mentioned.

If a brand wants you to get a food item with your favorite drink, then they will do a deep dive on the data. They will find out something which you had offered before and which you had purchased before and they will add it to your offer. So, let’s say you had a donut about six months ago and you keep having a cappuccino, and six months down the line, you order a cappuccino. Because Starbucks knows that you had this doughnut at one point in time, they will try to pair it, and they will do it at the exact same time that you had done six months ago. So, this enhances the possibility of you actually purchasing this.

Wow, we’re now on the results page. Three X increase in marketing campaign effectiveness. As a result of their entire hyper-personalization overall since 2016, Starbucks has managed to have 3X increase, three times increase in marketing campaign effectiveness. Their email redemptions haveincreased twice. They’ve actually doubled. So, every time Starbucks sends an offer over email, twice the number of people are now actually redeeming them because they are so much more valuable for these users. There has been a 3X increase in incremental spends via offer redemptions. Every time you send an offer to an individual, the individual, the user likes it so much that he ends up spending right at that point of time.

Twenty-four percent of the total company transactions happening via mobile app. Starbucks has absolutely revolutionized the mobile app ordering game. Look at this chart. This is a very recent data. People are using the app at least once every six months, and these are projections for 2018. Starbucks actually has the leading mobile app payment system. Almost 24 million people are expected to be using this app to make a payment on Starbucks. Apple and Google are actually lagging behind. So, Starbucks has been very successful with their mobile order and pay app.

Now, the thing is using the app, you can make an offer, you can actually purchase something right at your house, and then you can pick it up at the nearest store. This has actually ended up becoming a boom for Starbucks. They have managed to reduce the lines which are there in all of these Starbucks stores. The focus on personalization has made the app experience very delightful. People are actually enjoying using these apps because all the recommendations are absolutely lovely. You guys actually want to buy that stuff, and that is what they are sending you. So, that is how hyper-personalization has changed the fortunes of Starbucks.

Wow, now we can do the big daddy of hyper-personalization. This is the last brand that we will be talking about. It’s none other than Amazon, the king of hyper-personalization. Every shopper on Amazon is aware of the fact that their shopping experience is custom built for them. It is highly personalized. There are several visual cues that let the users know that everything that you see is specifically geared for your tastes and preferences. The usage data is combined with the personal data attributes. This has been a common recurring theme in all of the brands that we have seen so far. User data and behavioral data are actually combined and analyzed to create highly relevant experiences.

A buyer persona is critical to Amazon’s retention strategy. By combining personal and behavioral attributes, Amazon has the ability to look deep into the mind of the user. Amazon understand your motivations much better than competition, which is why they’re sending you those highly personalized recommendations. Now, have you wondered why those recommendations are coming from Amazon? They are not arbitrary, but they’re actually very, very logical. Amazon’s product suggestions are highly targeted and relevant for each individual user.

And this is happening because of data. We will take a look at a few visual cues that Amazon offers to the user. This is actually the Amazon website and I have logged in from my account. “Recommendations for you in Computers & Accessories.” So, it knows that I have logged in from my account, first of all, and it sends me recommendations in computers and accessories. “Recommendations for you in Electronics.” Now, all of these are based on my past browsing behavior. “Your recently viewed items and featured recommendations.” Inspired by my browsing history, Amazon has suggested a stream of products.

Now, I was trying to look for a OnePlus 5 phone and see the behavioral recommendations that they have done. They’re actually sending me a recommendation for the OnePlus 5 Insurance program. They’re sending me recommendation for OnePlus 5 screen guides and mobile cases. So, these are actually useful things for me. They’re completely related to my past browsing activity. Once again, it’s giving me recommendations for books that I have browsed and books that I have purchased. Amazon’s recommendation engine powers 35% of their conversions. Compared to other e-commerce brands, Amazon’s recommendation engine brings 60% more conversions. How is it doing it? Well, the algorithm, which powers the recommendation engine, is called item-to-item collaborative filtering. What is this technology exactly?

This type of filtering matches each of the users purchased and related items to similar items. It then combines the similar items into a recommendation list for the user. Now, each product on Amazon is analyzed and grouped into a neighborhood of similar products. Every time you buy a look at an item, Amazon will start to recommend you another item from its related neighborhood of products. This approach differs from a user to user approach, something that we saw in Netflix and Spotify. Here, user-to-user approach, users are analyzed them grouped together.

How does item to item collaborative filtering work? These are the four basic things that Amazon considers, your previous purchase history, the items that you have in your shopping cart, the items that you have rated and like, and items that have been liked and purchased by other customers. All of these things are actually combined, analyzed to create suggestions that are highly targeted to that particular user. Now, Amazon does a lot of analytical work to power its recommendation engine. There are absolutely thousands, hundreds of thousands of data points at Amazon looks at. Here are a few of them.

Purchase shopping carts, what are people actually buying? Items from abandoned carts, stuff that you put in your cart, but never bought. Wishlists, what are the things that you want to buy, but are not buying for some reason? Maybe you are poor like me. Referral sites, did the user come from YouTube or did the user come from other sites? Dwell times, how much time is each user spending on a particular product page, a category page. Online pricing A/B tests. Now, this happens a lot of times. Probably I bought a T-shirt for 500, and Avantika got the same T-shirt for 600.

This is also another factor that helps Amazon determine what price works for which kind of user. Ratings and reviews, as we all know, these are highly critical on Amazon. Everyone reads Amazon reviews to buy a product. Demographic information like area-wise interests. Mumbai people are more interested in mobile phones, Delhi people are more interested in, I don’t know, cricket bats. Product page views, marketing campaigns, CTR, how much has your ad…how many clicks did you get on your AdWords campaign? Stuff like that.

Now, this is a real example that I’m going to explain. This is an email that I actually got from Amazon. Let’s see. “Puma Men’s Mega NRGY Knit Olive,” was the subject line. Now, I’ll give you some feedback on this. I searched for green running shoes on Amazon, but I did not transact for them. Amazon sent me this email, which is full of recommendations that were very close to my preferences. Amazon accessed behavioral data from my user profile and it merged with my personal attributes to create a user persona. In my case, I have purchased Puma footwear in the past several times. As you can see, they have sent me a personalized email with my full name, it says, “Hello, Priyam Jha,” and they have actually mentioned whether I was looking for something in the running essential store.

Now Puma, I have bought Puma shoes in the past, which is why the product recommendations that you see are from Puma. They know me so well that they actually suggested me shoes for size 11, which is actually my shoe size. Now, how did they do this? In my example, Amazon had access to my full name, my search query, my average time spent on search, my past purchase history, my brand affinity, category browsing habits, time of past purchases, and average spend amount. On the basis of all of this information, they were able to create that highly personalized email that they sent to me. And guess what? I actually ended up buying the product.

Yes. With that, we have now come to the end of the webinar. We will be talking about the future of marketing, which is all about context. I’d like to give you guys a small statistic. Gartner predicts that by 2020 businesses will see a 15% increase in profits from personalization technology that recognizes customer intent. What lies in the future? Now, Forrester recently did a study. They had data that suggested that matching user behavior to products and crafting marketing communication on that basis gave brands a 15% uplift in conversions.

Today, users are empowered, educated, and evolved. They have already started showing signs of hyper-evolution in terms of their buying behavior. There is multi-screen behavior happening, users are not just using one device to make an online purchase, they are traversing multiple devices, may be a mobile phone, their computers, their tablets. So, brands are looking to create seamlessly connected experiences that are harmonious. Attention is very scarce right now, and hyper-personalization is the key to customer retention. You need to ensure that the user’s attention remains with your brand, and you can only do that if you provide them customized, personalized targeted value.

The second point is extreme market clutter. There is so much of competition and all of the bands are clamoring for the same users’ attention. The market is flooded with communication. So, the users have learned to filter out most of these messages. Only hyper-personalization will help your brand create a success story with a contextual, engaging, and valuable content, which is unique for each user.

Thank you, ladies and gentlemen. I hope you enjoyed this webinar. Everything that we have discussed, and my transcript, will be available for you guys. We will be creating something for you like an E-book or something, and you will be getting that. I will now transfer the mic to Avanthika.

Avantika: Thank you so much, Priyam, for this webinar. Alright, guys, we’ll start answering the questions now. Over to you, Priyam.

Priyam: Right. So, I hope you guys enjoyed the session, and I see that a few of you have shared a few questions. I’ll try to address all of them. So, Anubhav Tewari has asked a question, “How WebEngage gets the history of the users of Spotify, is Spotify sharing the user data?” You’ve also asked, “Whether Spotify itself uses NLP or WebEngage does that.” So, WebEngage does not get the history of the users of Spotify. Spotify is not sharing user data. Spotify indeed uses NLP themselves. They are not using any other service. In fact, Spotify has one of the most complicated, one of the most powerful personalized recommendation engines in the world right now. And they have been developing it over the years themselves. So, they have their own system.

Avantika: Priyam if you would, can you throw some light on what’s an NLP.

Priyam: NLP is Natural Language Processing.

Avantika: All right.

Priyam: So, Nathan Bellotti has asked us, “Do you think the example of Netflix or Spotify will be of any help to people with an insignificant budget?” So, I am not very sure about this question. I think he means about brands who don’t have a very massive budget, how they can use hyper-personalization. If that is indeed what you meant, Nathan, yes definitely. Hyper-personalization is very useful for smaller brands. Also because the engagement aspect, where a brand engages with the user, is somewhere where hyper-personalization is going to have the most amount of effect, and this can easily be done by data. The amount of data that you collect, and the amount of data that you actually use and implement in your marketing campaigns, is irrespective of your size. It all depends on the kind of tools that you use. So, there are several tools available, and it all depends on your needs.

The next question is by Rishali Gupta. she has asked us, “Which personal care brand is into personalization?” So, the one brand that comes into my head right now is Sephora. Sephora is actually at the forefront of integrating technology with their offline experience. And Sephora actually gives their app users a lot of personalized recommendations. It also takes into account your past purchases and your browsing behavior. So let’s say, you are a big fan of X, Y, Z lip care brand, then if there’s X, Y, Z lip care brand has a new product that is about to launch, you are going to get personalized push messages and notifications and emails about that product.

Nimit Shah has asked a question, “Is there any way we can leverage these techniques in mass marketing like Facebook posts?” Yes, definitely, Nimit, these techniques can be done for mass marketing like Facebook posts. But hyper-personalized and super targeted techniques are something that will be difficult to do because if you are going to do it on a social media platform, you need to actually integrate your recommendation engine to this social media platform. However, marketing is indeed possible.

Facebook and Google actually allow you to create segments, and they actually allow you to upload, lets’ say a CSV file of these users on their backend, on their ad engine backend. You can actually show them customized ads, which are targeted only to these individuals. This is actually a part of retargeting. But if you’re talking about hyper-personalized mass media campaigns, then I am not very confident about that. And there might be some people who are doing it, but this is going to be very difficult because it calls for a very deep integration of your personalization engine with social media platforms.

So, Lawrence Oakland has asked us a question, “Are the technologies you mentioned developed internally by Amazon, Spotify or do they leverage…?” So, Lawrence, a lot of times, these brands have a big mixture. Sometimes they do develop the technologies themselves, in fact, most of the times they do develop the technology themselves, but they also take help of a lot of marketing automation platforms. So, you have to understand that the scale of operations in these businesses is absolutely massive. And they cannot probably do one on one personalization for all the regions for their global operations using one system. So, sometimes it’s an amalgamation of their personal systems and systems that they source from the outside.

So, Mr. Nabodhi Trivi has asked the question, “How difficult is it to implement personalization when it comes to e-commerce website if you have a small tech team?” Mr. Nabodhi, you will be surprised to know that it is not very difficult to implement personalization. All it requires is for you to actually have a marketing automation platform that can actually help you control all of the data that you are actually getting every day. There are several users on your e-commerce platform who do a lot of transaction. It may not be monetary, but it may be a search query, it may be something that they’re browsing. Just their heat maps can tell you a lot about their intent and their motivation. So, it’s up to you to actually use this data and create personalized recommendations for these people.

Avantika: I can see they’re a couple of questions on GDPR. So, I just want to inform you guys that our next webinar is on GDPR, which is tentatively scheduled on the 12th of June. So, I will be rolling out the communication as soon as possible. If we have not answered your question, we will make sure that we personally address them through email. And thank you so much for your time. We really appreciate it and have a good day. Thank you, guys.