Implementing Personalization

1.  Everybody likes to feel special and if you feel special, if you think that a message is carved out for you, without even knowing it, you’ll have more of a chance for that recipient to open an email and look at it because they think, “Hmm this is interesting. It’s really about me.”
2. Another definition of personalization is that acting upon the things that you learned about somebody.  Reacting to the little bit of information that you collected and just harping on it, well people call that personalization and it is actually.
3.  The ultimate personalization is personalizing for everybody, every chance, through every channel.
4. Don’t just use data just to react to certain things that you know. You’ve got to learn how to bring out the potential of that.
5. Marketers need to think about what the limits are of personalization and have some conscious decisions to make in terms of what would you want to do with it to avoid getting creepy or to personal.

Below is a lightly edited transcript of Episode 21 of the Inevitable Success Podcast.

Transcript:

Damian: In today’s episode I’m going to get personal with Stephen. It’s funny because while Stephen and I both like puns, we’re going to be talking about personalization and what it means to you. So what does personalization mean to you? Stephen: Well I know what it means to me. It’s very simple, I’m dealing with this merchant xyz and I’ve been dealing with him for a long time and if they know about me and actually analyze me…not about my whole personal life…I mean if they start asking me “how old is your daughter?” or creepy questions, I’ll turn that off. But I really believe that as a buyer of services or products they really understand what my passion is. For example, say you go to the deli in the morning every day. If the barista knows how you take your coffee it’s easier for her. It’s easier for me. So you save time on both ends and I get the service that I normally want to get without saying much. If you do that for millions of people at the same time using the power of machine and modeling and all the other things and that’s personalization to me. Well, that’s one way to say it. Damian: You tell stories that always remind me of stories. So there actually is a deli that I used to get breakfast from every single morning and they got so good at knowing who I was that I didn’t have to say anything. They would just make my sandwich. But it went to the point where at some point I decided “I think I want to do this keto diet thing” but they anticipated what I wanted so they just made me a bagel every time. It was almost like, you guys are too good right now and they make me think that there may be some situations where anticipating personalization can backfire. Stephen: Yes. If you imagine that you don’t have any human interaction and you know that the machine is doing it, sometimes you don’t want to be too good because it becomes creepy. You have to draw the line. But let’s talk about the most rudimentary form of personalization and then start from the bottom. Instead of starting with the most elaborate kind and dialing it back. I’ll get to both ways. So this is a question that I ask when I talk about personalization. I have been speaking about this in conferences for a number of years already because I thought that the natural next step of a big data movement is about personalization. What are you going to do with all this data? The premise is very simple. Everybody likes to be special and if you feel special, if you think that a message is carved out for you, without even knowing it, you’ll have more of a chance to open the email and look at it because you think, “Hmm this is interesting. It’s really about me.” So you have more time to open it because if you give a group picture to a bunch of people you know they look at their faces first. How do I look in this picture? Everybody is about themselves. This oldest trick is that you personalize and yeah, they will respond better. So that’s the premise of this whole exercise. Damian: So they say that’s why Facebook even works. It has people wanting to look at pictures of themselves, hence Facebook right? Stephen: So that’s why we do it. And if that is the case, and by the way they all do it, big data companies like Google and Facebook, those guys they don’t use words like “big data” or “personalization”. It is just part of their DNA. But if you break it down, it is really doing the personalization using big data. That’s what they do. So what is personalization? It ranges from as rudimentary as sending an email and it starts with “Dear Stephen”, for a lot of people that’s personalization. That’s one way to put it. The next level is studying at least what he bought in the past and then put some collaborative filtering on it or people who bought this item also looked at that. Now that’s very rudimentary product recommendation but it works, a lot of times it’s wrong or a lot of times it’s overbearing. For example, when I bought a garden hose nozzle for my wife because she took to gardening as a new hobby, I think this reputable retailer sent me a set of garden hose nozzle offers for about three weeks. That’s just stupid. Now wait a second here, you’d think I’m in the business of collecting these things. So this is what happens when you do the product kind of a personalization but is that personalization? It is actually better than not doing anything at all. Another definition of personalization is that acting upon the things that you learned about somebody. Let’s say your cable company or a satellite company or some video on demand company or whatever and then you notice that somebody shows up and he started clicking away about Game of Thrones. So you know that you have that piece of information and then they are compelled to show you a Game of Thrones picture every time you land. It’s not a bad thing, I mean it’s ok  but only for like first three times. If you keep showing the same thing again and again they will say well “hey it’s not just that, I have interest in other things” right? So reacting to the little bit of information that you collected and just harping on it, well people call that personalization and it is actually. The problem with that is, and this is where the ultimate personalization happens is, that instead of just acting on the data that you have for real. So my question is what is the coverage of that field or data point? A lot of times data as specific as this person likes Game of Thrones or this person likes the football game on Sunday. It’s really hard to come by, by the way. So what happens is that you end up bothering the people that you have data for and ignore the rest. There are ways you could be as big as like 10 percent to 90 percent so you end up really annoying 10 percent of people because you know something about these guys and then 90 percent of your time you just don’t do anything based show generic messages. So there are levels of personalization obviously. Now the ultimate personalization is this, you should be able to personalize for everybody, every chance, through every channel. Now that sounds like a tall goal but how do we do those things? Well one is that you got to really start thinking about smart use of data. You don’t just use data just to react to certain things that you know. You’ve got to learn how to bring out the potential of that. For example, if you have a set of people that you know they like a science fiction show whatever. I’m just making up examples. You could do this for fashion industry or the gift industry or whatever it is. If you know that certain people like certain things or somebody is an early adopter or somebody is a gift-giver or a bargain-seeker. If it is the travel industry, somebody is mostly traveling for leisure versus business versus luxury travel versus budget travel, I’m just listing all the things that you collect about people. Now if you react to certain things and things you do tell us do that. I may know that you are here mostly on business and this is how I’m going to treat you and the treatment that you get from that hotel a bit different because they recorded everything you did. Again, you’re reacting to certain data points but unfortunately there’s a coverage issue you don’t know anything about everybody. Damian: And some of that data could be wrong too. Stephen:That’s another thing. People get scared about the word “modeling” and you shouldn’t because I’m not asking you to build it. Either a professional mathematician or statistician will build it for you or down the line maybe the machine will build it for you. We’re not asking you to do it. But people should open up their minds about what a model means to you. What it means is simply, I do not know 100 percent that this person is an early adopter of a certain type of a technical product but if some person or machine will tell you that “I think it’s 80 percent likely that this person will be an early adopter”. Now is there something you can use to personalize your messages. You will even know that “you know, the answer is only like 50 percent or so, I’m not going to do anything”. Damian: I think it depends to what the risk of being wrong is too. That’s exactly right. Yes. You know, like efforts trying to figure out if you like dogs more than cats maybe it’s lower risk. But if I’m trying to figure out what your political affinity is that is very polarizing. Stephen: Actually I have done that. Funny story, we try to build a lot of personas or models, affinity models for a general population and some of the things that we did was for political organization and even with an ample amount of data, it’s really hard to predict. So you’ve got to minimize cost of being wrong but it didn’t stop us from building things like…I’ll give an example, leaning conservative or leaning liberal or hardcore liberal or hardcore conservative. What we found out is that people are so complex that if you just look at my demographic data you will not know what I am. I live in a mostly conservative town with a certain level of income. The cars that I drive all that kind of stuff. What is this guy? The prediction is not easy. So what you do is that you create an innocuous target like leaning conservative or leaning liberal. And by the way, when you do that so you can sell the model to both parties because if somebody is hardcore you don’t want to talk to them anyway. You want to talk to them for fundraising but if you want to sway somebody then you’ve got to talk to somebody in the middle. So the cost of being wrong can be mediated by selecting targets in a smarter way and in a way that you don’t offend anybody. And also, this is why I never got into the business of predicting who’s more likely to be diabetic or have erectile dysfunction. I know how much money is out there to sell this product in the nutraceutical or pharmaceutical businesses but the cost of being wrong in offending somebody is so high that we don’t do it. But let’s just set that aside. It’s a good topic for next week by the way, which is just because you can, you shouldn’t be doing everything that you can. What’s the limit of predictability and ethical ways? We have all these like ethical regulations being imposed on us. How do we react to that? We have to talk about that at some point. But today, let’s talk about an innocuous target like, likely to be an early adopter, that is not going to offend anybody if you’re wrong about it. The worst thing that could happen is that while you send that cutting-edge product offered to certain people who are not really cutting edge people and fine. This is a game of probability. Even if you do know that somebody likes certain things can you guarantee hundred percent response out of that person? The answer is no. Now this game is not about knowing absolutely everything, it is not about having a guaranteed response. We’re not doing that. All of this is about increasing the probability that the person will open the email and actually buy something. So that means having a score like, “I don’t know for sure but there’s an 80 percent chance that the person is a fashionista or early adopter”. What I want to say out loud here is this a beneficial thing to do. When you do these things, the coverage of the personalization increases infinitely because you don’t know for sure if somebody is really an early adopter but if you build a model to predict if somebody is or is not and put it on a scale like 70 percent chance, 80 percent chance there’s no missing value and you could build up a rule based on a hundred attributes for everybody. Right, this is a hypothetical situation. And by these hypothetical people do these things, I don’t know for sure but if you notice that hey this guy’s score is really high as a frequent traveler but his early adopter score is only like six out of 10. Then you’re going to offer a traveling offer first. If you started doing it this way, the dream of personalizing all the time everybody through every channel is that much closer. So I’m not talking about how we actually do these things although I’ve seen a lot of this kind of activities in real life. Also I’ve helped a lot of companies to build personas for the purpose of this type of personalization. Today I just wanted to define the realm of what we call personalization because again, last week we talked about posers right? My God. Personalization is another hot topic like “yeah you should just buy the software we will personalize every message for you”. Well really? With what data? Do you have that for everybody for every occasion for every product? I don’t think so. What is your plan to fill in the gaps so that it is really for the general public and not just some selected few who you have some data points on and begin to just harp on it. Damian: I like how we actually started off. It’s got to help both parties. I thought that was really a great story. Stephen: Actually the political parties are really good at it. In fact, unfortunately, these organizations still keep them around because they only happen once in four years. At some point, I’m not going to say who, but can you imagine what candidate? They were so good at it when they sent out emails they had the ideal amount of donation for each person. So certain people will get the twenty-five dollars as the default, some other people get a hundred dollars or some other person will get ten dollars as the default. They would actually predict that! This is the power of the modeling. I don’t know for sure but what is the most likely chance for that person? And then if you react to it and actually go all the way to send that the right message for that person at a proper time, that’s personalization. So there are many levels of these things, starting with “Dear Stephen” going to “Oh you bought this I’m going to just keep sending you the same offer”. The same thing happens all the time because I bought a wireless mouse and I get wireless mouse offers for like three weeks. I’m like, “wow that’s the last thing I want right now”. That happened. But you called that personalization all the way to you know, “oh I have a real piece of data so I’m going to react to it”. That’s a good start because if you do not know how to show different things to different people then all these modeling and persona and affinity and all these words that I’m using are totally useless because you don’t know how to show different offers to different people. So it’s a good exercise to know how to react to a certain piece of data. Yes, it is a good thing but you can go higher level where I don’t really need to know everything. I’ll make the best of what we have and increase the probability of them opening up my message. Damian: Speaking of that, and I’ll wrap up here. But actually, I think a few years ago you told me a story about you know, what’s too much data to have on somebody? And I think you made the analogy that if I’m sitting on a plane next to somebody and they ask me like “oh what’s your name? that’s OK. If they ask “what do you do for work?” it’s OK.  And then you know they say “how much money do you take home after taxes?” Stephen: Now you crossed the border. Damian: There’s a line. Stephen: We all know that boundary. Again the limit of the technology is almost limitless. Let’s state that there’s no limit. It could be really creepy if you want to be one of the examples that we see in movies is that you remember the movie by Tom Cruise Minority Report. That movie was quoted many times and I have quoted that movie in my own article too because they show the possibility of the real-time personalization in a mall. They can read and they bring out all your past history. They bring up personalized greeting “Hey Mr. Yamamoto, I know you bought this tank tops last time. How are you enjoying this? Do you want to see new lines of winter clothes?” And they actually pop up the picture of that person and change colors and they show different sweaters. By the way, I’m not impressed by all this because we can do that now. We cannot do it that fast but imagine that 10-20 years after you know we started doing these things, of course, you could do that kind of stuff. The question is what should you do it or not that’s another question right there. But why people fascinated by this example because this shows you how far it can go. Should we go there? That’s a different discussion altogether. But let’s just break it down for a second. What just happened in that movie, somebody walked in and that means you have to recognize who they are because the backbone of personalization or what we call Customer 360 or personalized CRM and all those words that we say, the starting point of all this is knowing who a person is based on the data you’ve got. It could be the retina scan, it could be the name and address, it could be anything. Second step. You should be able to collect and be able to retrieve real fast. This is what the big data movement was or still is and retrieve the data about that person almost near real time. Third step. Great, now I know who that guy is and I know all about his past behavior. What would be the next thing that he should be thinking about? That’s where this movie is a little disappointing because I think they could do a lot better than just showing different colored sweater. But let’s say that you had some real-time suggestion engine and you built it right. The fourth step is the delivery of the message to that person. That’s the complete circle of personalization. Again, that collection on data on the individual, data retrieval on the individual, predicting what that guy is all about and suggesting the best match up the product, and then delivering the message real-time or it could be emailed and just have a different picture for you know whatever, it all this dynamic content. Dynamic content should be a lot more sophisticated based on these things. If you do all these steps, fine in the movie this whole thing happened in half a second. In the past, this process could have taken months. Now. I think it’s down to a few days. So there is some gap, it doesn’t happen in 0.5 seconds but it will be there I think. So the question that I have is not about can we do these things, which we could do with now. All these elements are there. I think it’s time to think about then what does this mean for marketers? Are you going to be just settling in? I’m happy that I can personally address a person as “Dear Stephen here is our spring line” and then send that same spring line message to everybody? Is that enough for you? And then send the same spring line message to everybody? That’s not personalization. Damian: So I think we’re going to have this episode sometime in the future. I think we’re going to sit down and think about if technology  is not a limiter what are some things that we can personalize? Stephen: One guideline that I actually wrote about this is that when in doubt just don’t be creepy. If it’s something that’s creepy and if it’s something you don’t want to have happen to your daughter or your family, then maybe it’s too creepy. Damian: That sounds like it. Stephen: The point is that the marketers should really think about what you want to do with this type of technology and know what the limits are and have some conscious decision made in terms of what would you want to do with it. Because this day’s coming around the corner. If somebody is batching and blasting the same match to everybody and hope to God that that’s OK. Well, I’m here to say it’s not because you’re going to lose to people who adapt to these type of things because if I mean how many emails you have in your inbox at any time of the day? Hundreds, right? What do you normally do? Not read them. What if something stands out in the subject line? You go like “Whoa, I don’t know how they know about it. This is something that I may want to look into.” You open it. People who do these things will win out and people who don’t? Well I’m sorry. Why do you think the Facebook stock is so high? Why? Why is that? Is it because it is the largest billboard in the world? No. It’s the largest billboard where you know who’s looking. That’s the key word, isn’t it? So you know who’s looking at the billboard then why shouldn’t you be personalizing it? And they are doing it. I mean they got better, let’s just say they are a lot better actually. So that’s why. I think that’s where it’s all going. So I think what marketers should think about is “how do I not miss out?”. What is the most cost-effective way to get there? What are the steps that I have to take now to get to that level? So maybe next time we’ll talk about steps that you have to take to get out of this lull of batch and blast and what the first thing is they have to think about and the second and third step, we can talk about that next time. Damian: Sounds great. Looking forward to it. All right. Thanks Stephen.