Search the website

Learn how to use to data science to understand your customers and embrace data-driven dynamic and personalized marketing to maximize customer loyalty.

Video Transcript

– [Amit] Hello, everyone. Thank you very much for joining us today. Quick apologies for misunderstanding we had with time of this webinar. I’m happy to have you all here today. My name is Amit, I’m the Head of Marketing for Optimove and I’ll be your host for this morning. Today’s feature topic, we’ll be talking about Speed and Precision in Customer Loyalty. We’ll be featuring this webinar two distinguished speakers, two domain experts. Rusty Warner of Forrester, Principal Analyst, which excels at enterprise marketing technology expert, and Pini Yakuel, Founder and CEO of Optimove. With no further ado, Pini, the stage is yours.

– [Pini] Thanks a lot, Amit. I’m very excited to be here today with Rusty. How are you doing this morning?

– [Rusty Warner] I’m doing great. Pleased to join you.

– Wonderful, so let’s get started. We have some really thought-provoking content for you guys today. I’m excited to have Rusty here with me to kind of like support us in terms of showcasing all the knowledge they have at Forrester. Kind of like serving the entire market and understanding the trends and what’s happening in organizations, large and small. I’m going to start by introducing Optimove. So I’m just, as Amit said, I’m the CEO and Founder of Optimove.

About Optimove

My name is s Pini Yakuel. And we are the science-first customer marketing hub. This means that we help brands to focus on their existing customers, leveraging data science and machine learning to help them better understand their customers. And then create emotionally intelligence communications to talk to each customer in the most relevant way that creates intimacy with the customer. As a business, we have a offices on three continents. We have the Tel Aviv office, the London office, and the New York office. I’m based right now in New York, talking to you from New York. We work with 250 brands, very innovative brands like Stitch Fix, which are leading in terms of fashion and styling, 1-800-Flowers, some gaming companies like Caesars, and many, many, many more. And we are a fast-growing company. It’s kind of like really doing well in this space. Specifically, I have gained a lot of experience to knowledge in terms of loyalty just by working with tons of different customers, by being immersed in the space for over eight years now. And I have some things to share, so hopefully you guys can find this interesting. So let’s get started.

CRM marketing and customer loyalty in the digital age

We start from this notion of customer loyalty, which is the Holy Grail in the digital world. And I’m saying that because loyalty is becoming a more and more elusive concept in today’s world, right? So in the past there was a lot of barriers for customers to easily switch between services. But these barriers do not exist today. So the fight over loyalty as marketers are, you know, ability to attract customers and win their loyalty is really, really more challenging, but it’s still even more important than ever. And the way we do this, the way we get customer loyalty is by utilizing customer marketing. So customer marketing or retention marketing, or CRM marketing, you can give it tons of names, but essentially it’s this art and science of talking to your customers and engaging with them to foster loyalty. This will be the thing that will help us with loyalty. Obviously, the quality of our service or product and the quality of the core thing, of what we do, is one of the biggest things in driving loyalty. But we can impact that as well as marketers. And eventually if we do have loyalty, we’re going to have an increased customer lifetime value, which is the goal of any business. Rusty, over to you.

– [Rusty] Sure. Well, as Pini just said, because of how customers have evolved, and because of being connected to lots of information that we have at our fingertips, it makes it more difficult for marketers to drive that long-term loyalty that they might be looking for with individual customers. And if you think about it, we can all relate to this because we are all consumers as well as being marketing or technology professionals. And we know that because we have access to the Internet at any moment of the day, we have higher buying power. We’re exposed to more choices and more information. What we found in our research, though, is as we get access to more information and we get that buying power, we’re not just motivated by price. We become more value driven and we’re looking for an experience. Of course, we all like a good deal, but if we have the money to spend, we will often go to the supplier that is going to offer that better perceived value or that better overall customer experience.

We know that because we use our devices every day, that we are becoming more willing to adopt new things. So we refer to that as hyper adoption. We’ll very quickly jump on a new piece of technology, but we also have to be aware, as brands or as technology providers, that people will also abandon just as quickly. So as soon as something doesn’t perform, we’ll move on to the next thing that does. And finally, to make things even harder for the marketers in this new empowered world, customers don’t always believe what you say in your advertising. I’m dialing in from London. And I know if I travel on the London public transport system for just 45 minutes, I’ll be exposed to 130 to 150 ads that are representing between 80 and 100 products. So with all of that information coming at me, it’s very easy to just turn off and tune out. So all of this, which is good for the customer, makes things harder for the marketer. But, Pini, I know you have some ideas on how to approach customer marketing.

How does Optimove help you interact with customers?

– Yeah. So thanks, Rusty, excellent points. And in terms of customer marketing, so we asked ourselves the question, “What does it take to achieve the ultimate in customer marketing?” So if that’s the craft I need to master in order to achieve customer loyalty, what does it take? So one thing I realized is that actually what brands are looking for, they’re looking to start a conversation with their customers, and this is the first point. And to me it made a lot of sense in my head.

So essentially every campaign I receive from one of the brands I interact with, that campaign is always them trying to start some kind of a conversation with me. So, “Hey, you have a birthday. Here is something,” Or, “Hey, guess what? We just launched something new.” Or, “Hey…” So everything they…whenever they come at me with customer marketing, what they’re doing is trying to start a conversation. And that understanding. I think, is key. It helps us to well frame our craft. So often understanding we want to start a conversation, we do want to start an emotionally intelligent one. And what I mean about that is I mean that it’s not just okay to start a conversation, you need a very good reason to start a conversation. And your customers are going to give you various triggers and various points in their life cycle, like queues and hints, that could amount to a very good conversation starter, right?

So a good conversation starter is one that will resonate with that person if I do start a conversation. So an emotionally intelligent conversation needs to come from the data. We need to know the customer really well. We need to listen to changes in customer behavior. So that when we do engage in a conversation, that conversation becomes that much more powerful. And the whole thing together, we like to call it moving at the speed of your customers. So moving at the speed of your customers mean whenever there’s an opportunity to start a conversation, I do want to start a conversation. I want to say something. So I don’t want to leave untapped opportunities on the table. I want to leverage them. But when I do so, I want to be intelligent about it. I want to be emotionally intelligent about it so that I can resonate the message in the best possible way with the customer.

Keeping marketing campaigns dynamic and personalized

– And what Pini just described is key to this conversation that we’re having today. One of the questions I get often from marketers is, “Well, Rusty, you’re telling me that my campaigns are dead. So if I can’t send any campaigns, what should I do?” And I would say, “Your campaigns are not dead. They’re simply evolving.” And when we are sending campaigns on outbound channels, we have to take into consideration what Pini just described. We have to think of those campaigns as conversation starters. We need to make sure that when we’re sending something out that it’s consistent with what the customer sees when the customer initiates the interaction with the brand. We have to make sure that what we’re doing is relevant to where that customer is in his or her purchase cycle, or life cycle with the brand. And, of course, back to that point I made earlier about all of the messages that are competing for the customer’s time and interest, we have to be more engaging.

And, Pini, if you’ll click to the next slide, we’ll talk about the next point in this story, which is all about customer context. So if you want your campaigns to evolve and be more relevant to the customer, and be those engaging conversation starters, you really have to put yourself into your customer’s shoes. And that sounds really straightforward but of course it’s not, it’s incredibly difficult to do this, but you have to think differently. You have to think about creating marketing campaigns and programs, that not only are aligned with your brand strategy, but that are focused on creating a difference in the life of your customers. So that you are delivering a value exchange. You have to make sure that you’re not just being clever and sending out offers that have a gimmick or a hook. That might work in the short term and it might drive some uplift and might optimize a campaign where you’re trying to sell a product, but ultimately it doesn’t work in the long term.

Instead, you need to make sure that what you’re doing is perceived as useful to the customer, and you need to ensure that utility is something you deliver over time. Not just one-off but across all of your marketing programs. And if you get that right, you can create a pull effect with the customers, where they’re coming to you and initiating the interactions, initiating the conversation, or keeping that conversation going instead of you constantly having to push something at them to start the conversation. And that leads me to the next point about the contextual marketing engine. So if you embrace customer context, you’ll need some technology that will help you take on board what’s going on with the customer at any given moment, so that you can deliver those experiences which will be relevant to the customer’s current context, where you are delivering value and something of use to the customer, instead of just content or an offer.

How Optimove helps marketers build and maintain customer trust and loyalty

And as I mentioned a minute ago, you’ll want to do that at the right moment, but you’ll want to do it at the right moment in the customer life cycle, which means you’ll repeat that over time, so that you build up that consistency with the customer. So that there’s a level of trust, and they begin to expect a certain level of service from you, that you can then maintain and optimize. And as you’re building that contextual marketing engine, we’re going to get into a lot of details on some of the technologies that you can use to ensure that your contextual marketing engine is effective. But at a high level, we think it needs to accomplish five things. One, you need to be able to recognize the customer. And that means that you will need technologies that will help you with cross device and cross channel identity resolution at an individual level.

In some cases, that will mean recognizing a customer explicitly when they authenticate with a digital session. In other cases, it might mean understanding that an anonymous visitor has been with you before and you have data that gives you an understanding of who that customer is and what they expect, even if you don’t have an explicit identification in terms of a name or an address or an e-mail address. You’ll then need to take that recognition and put it into the context of what’s going on with the customer right now. What can you understand or interpret based on his or her behavior that will tell you what that person is looking for right now. And in addition to that, you need to take into account the history with the customer as well as external factors.

What time is it, what’s the weather like, what are the other events that are going on in the world that might impact what the customer is thinking right now. Then you’ll have to apply the analytics that will help you determine the appropriate action that you should take. Sometimes that will be a message, sometimes it’ll be an offer, in some cases it might mean that right now the right thing is to do nothing, which is often very difficult for marketers to accept. Of course, when you’ve made that determination of what you should do, you then need to orchestrate that. And to our point earlier about the conversation, this needs to be a two-way dialog. It’s not just putting some content in front of the customer and hoping for the best. It is the beginning of a conversation. And then you’ll capture all of the data around those interactions so that you can optimize.

And ideally, you’ll want to optimize what you’re doing with a customer in the moment, you’ll want to optimize what you’re doing with the customer over time, and then collectively you’ll use that information to optimize your strategy for all of your customers. And with that high-level view, I’m going to hand it back off to Pini, and he’s going to take you into some of the details of the technology that you can use to build this contextual marketing engine.

Four speed and precision optimization techniques for marketing organizations

– Thanks, Rusty. So yeah, I’ll take more of a complex practitioner view of things. Like I’ll try to provide some value with some main useful concepts and tips. So I think, first of all, I’d like to use this analogy of us as marketers being like martial artists. And I think even today with Conor McGregor versus Floyd Mayweather fight coming up, it’s actually even more inspiring than ever. So the two main traits that we need to develop both as marketers and as marketing organizations are precision and speed. So what does it take to reach the ultimate in customer marketing? You need to have those two things very well pinned down, precision and speed. And I’ll show you that you, as marketers, and your organizations, you need to acquire a few different techniques. And each one of these techniques I like to describe in the sense of precision and speed.

So you see it by the two logos at the top. So the four techniques that we thought about are org built for speed, data science, experiment design, and catering to the customer journey. So these are four like made up techniques, big techniques that you need to master. And if you do that, eventually you’re going to move at the speed of the customer, you’re going to master customer marketing, drive loyalty, drive customer lifetime value, and you’ll just be better at your jobs. So let’s go at them one by one.

Technique 1: Build your marketing organization for speed

So org built for speed, right? So org built for speed is essentially you need to ask yourself the question, “What does it take to run one campaign? How much time does it take?” So in many companies you still see the structure of an assembly line. So one of the marketers will draw your brief. They’re going to send that brief to the data team. The data team they’re going to write a query to get some kind of a list. Then you’re going to go to the marketing team to get some creative. Then you’re going…then you push the campaign out via some or one of the channels or a few channels. Then you go to the analysis team and ask them to analyze the campaign. That process is very lengthy, it’s cumbersome, and it doesn’t allow you to move at the speed of your customers.

In order to move at the speed of your customers, your org needs to move away from the siloed approach of many, many different departments, and each one of them may be taking care of one channel into the studio approach. Well, we have a united team. We call them kind of like a CRM A-team. So instead of having four different departments doing that, we’re going to narrow it down to two different departments. The CRM A-team can both run the segment, build the campaign. You’ll still need the creative department. But with two departments you can run a campaign, and that gives you a quick turnover from ideation to execution. That fosters a lot of creativity because imagine if any one of your employees gets up in the morning and they have five different ideas and they can execute one of them within a few days, you’d have this creativity and ideas like popping up and coming up every day, every day, and you’re going to execute them. If they don’t work, you’re going to stop them. If they do work, you’re going to continue. So you have a fast pace of iteration.

And that’s something that we’re starting to see. We help a lot of our customers achieve that with our products, and this is a shift. And I predict that like in five-years’ time, we’re going to see most of the orgs built like that. So that brings me to the next point there. Rusty, can you share some thoughts on this?

– Sure. Because even though we’re talking a lot about technology today, I’m really happy that we started with the changes that are required within the organization, and for the processes. What you just described is very much aligned with what Forrester describes as a customer-obsessed operating model. Where you are using customer data so that you can become insights driven, but as an organization you become a lot more agile. You become fast, you become connected, you become empowered in terms of how the employees work together to focus on the customer requirement. And we think there are six important aspects of this operating model. And technology is critical. It is one of the six, but it’s only one of the six.

How you structure your teams to adhere to the corporate strategy and your culture as a brand, how you develop people and their skill sets, how you define the metrics that those people would use to measure their success. And how you find those processes on a day-to-day basis to ensure that agility, to ensure that they’re fast and connected and collaborative, and that everything they do is driven by their understanding of the customer and the customer’s requirement. That’s what we mean when we talk about this customer- obsessed operating model. And as you can see, there’s a lot of dependency here on the organization getting things right as well as deploying technology to help solve the problem.

Technique 2: Use data science to get to know your customer

– Cool. Thanks, Rusty. So that covers basically the first technique, org built for speed. It starts with a good organizational structure that can help you move the speed of a customer. Next technique is data science. And when I say data science here, I just want to be clear, I don’t mean data science in the hardcore sense. I do not expect each and every one of you to become a data scientist. It’s a profession, it takes a lot of experience and education. But what I do mean is to be data science literate. Meaning know how to use it and how to get value from it. And let’s talk about what is it. The way I see it, and this is my definition, so when I talk about data science as a whole, I talk about the fact that… I talk about kind of like organizing, manipulating, modeling, and analyzing customer data in the ultimate way to really get to know your customers. So I just want to get to know my customers, this is what I’m using the data from. Because this is how they tell me who they are, right?

So the first thing about the data science is this notion of having a single customer view. So having a single customer view, it means let’s make all the customer data accessible to marketers. Let’s see the entire customer story in one row of data. So usually we see the single customer view as just kind of like flat table. Every customer has one row. And there’s many, many, many columns that tell me who the customer is. And I can just read one row and get to know that customer really well. Of course this table is always correct for a single point in time . And we feel that we work with, as I said, 250 brands, and we always see it’s an infrastructure, it’s a foundation. If you don’t have this foundation, and usually what we do, we build this thing for our customers, but we need them to have a good row data. And companies that have good row data are companies that have organized their data and they tend to their data, and they cater to the data, and everything they do they make sure that the business is recording data in an appropriate fashion. This becomes a huge asset.

From raw data to segmented data

So the very beginning is something pretty simple. Just have the raw data. Think about us as cooks. I just want to have a good pantry. I want to have a lot of ingredients, really fresh. That’s it. Now the art of cooking with segmentation. So the next notion that I need to being data literate. And so after I have my skill segmentations, they are going to give me the first level of acquaintance with my cost savers. So my average is you looked at the few metrics, that’s the very first thing you’re going to do, and it’s pretty powerful. So let’s go forward to segmentation, which I just want to explain really briefly the differences. It’s going to help in the three device types I operate in. So I have U.S., Canada, and the UK, and my customers are using my product via iPhone, Android, and the Web. So this becomes nine segments. So I have customers who do U.S. in iPhone. It becomes the multiplication three times three, so I’m going have nine segments based on just simple slice and dice and filtering. However, in many cases, I could have many, many segments. What if I’m working in 50 countries and I have 10 devices? So all of a sudden, it’s 500 segments. It’s really hard to manage this simple slice and dice.

Clustering segmented data: difference between cohorts and segments

So then we use something called Clusters. And with Clusters what I want…Oh, I’m sorry about that. With Clusters what I want is I want a more succinct yet meaningful view of customer personas. So I don’t want to work with 50 or 100 or 500 different segments. I want to look at five personas. And cluster analysis is a technique that allows me to cluster the data, cluster my customers based on density. So how similar they are to one another, and how different they are from one another these are. In this case you can see five clusters, and this gives me a well organized and insightful view of my data. So I’m going to have a few personas. And the last one is cohorts, which is essentially the difference between a cohort and a cluster, is that, or segment, a cluster and a segment, this is a group that’s based on criteria. So today my bank sees me as a VIP Customer because I have a lot of money there. But tomorrow morning I take my money out, I’m no longer a VIP Customer in my bank. That’s a segment. A cohort is a group which you are going to … So you’ll forever be a part of the cohort. So a cohorts is a closed group.

As an example, I’m a graduate of the class of 1996. All of my life, I’m always going to be in class of ’96. I’m never going to leave that segment. And usually what people use cohorts for, so you can see each row in the cohort is like triangular analysis. Usually they look at the number of new customers they have on a monthly basis. And then they follow the new customers month by month to see kind of how the cohort evolves. And then you compare different cohorts to see trends in customer quality. Are we getting the same quality cohorts today as we got a year ago? Maybe a year ago we got customers of highest value, of higher value, or we got less of them. Today we’re getting more customers of lower value, but still the overall pie of money is increasing. So that’s what we use cohort analysis for.

Problems with segmentation

So these are like three forms. But more importantly what you want to understand about segmentation is you want to understand the fact that migration is actually the biggest part of segmentation. So the problem with segmentation is your customers don’t stay in their segments. It’s so annoying, right? You want them to stay in their segments so you can analyze this and predict and bank on the next action you’re going to take. But the problem is that if you look at, for example, active customers, and this is a very useful tool I like to use, it’s the from/to metrics. So if I’m looking at just basic life cycle segmentation, I can see that my active customers every month, 11% of my active customers, are actually migrating to the churn segment. So analyzing using this from/to metrics to analyze migration of customers between segment to segment is a very, very powerful tool. And you’re not going to understand your database and your customers if you don’t look at migration and only look at segmentation.

Again, the biggest, one of the biggest, parts of segmentation is the fact that they don’t stay there. And last but not least, it’s this notion of data science. Sorry, this motion of predictive models in general. So predictive models, it’s something more serious that sometimes you want to use. And just a few points on that for you guys to understand what is a predictive model. As humans, in our day to day, we always do predictive. So, for example, when I hire new people to new positions, I have a data set in my brain. I’ve interviewed many people and I now have some kind of a model. I know that I’m looking for people with a certain type of experience, with a certain type of education, because I predict they are going to be more successful at that job based on my past experiences.

I’ve hired some people with various degrees of education or various degrees of experience, and I’ve designed this model in my head in terms of who are the kinds of people I’ll want to hire specifically for this position. And don’t get scared of a predictive model. Just know that as humans, you guys do it all the time, all the time. If I’m predicting, if I meet somebody and I want to be their friend or not, again it’s based on past experience, experiences with various types of people.

Optimove’s solution: predictive models

So the second point is to understand that in many cases, I can predict next month’s revenue by just looking at last month’s revenue. So sometimes a simple heuristic, a very simple mathematical operation, will be a very good predictive model. Not always do we need a hyper sophisticated machine learning, seven PhDs working on something to build a predictive model. So when do you use… So when do you need something more serious is when you are super sensitive to performance. For example, if you’re spending a ton of money on PPC campaigns and you’re buying customers for your business, and you’re spending millions of dollars every month, perhaps if you work on a lifetime value model versus optimize on first purchase, maybe because the stakes are so big and performance is so crucial, you do want to use the best-of-breed predictive model. It’s not enough to use a heuristic of just looking at first purchase.

And when you do build the model, there’s a few things you need to remember. So first of all, they are as good as the raw data they sit upon. They require customization. As an example, a very popular model is a product recommendation model, and that these models do not work if they are not customized. If you take a generic one, an out-of-the-box predictive model, most cases they are not going to generate good results. Because life is just like that. It requires some kind of a minute understanding of the situation, the business case, the business story. And lastly, it’s not that easy when you want to build a very high performing predictive model. You need a very talented data scientist with tons of experience and tons of talent to make it done, and to really move the needle in comparison to the simple heuristics. So that’s on predictive models. Rusty, over to you.

Why predictive analytics is critical

– Sure. Well, I think Pini is just giving you a really good description of the data science that’s required. And if you are an organization that’s going down this path of being customer led and insight driven, then I would definitely recommend that you build a solid data foundation so that you can begin to understand your customers on a very detailed level. That means you’ll want to prioritize your data management and the analytics and segmentation that you do. And if you want to become more sophisticated to Pini’s point, then you would adopt predictive analytics and you would constantly be connected to a flow of digital intelligence data that would help you begin to finely tune those models, optimize those models, and you may want to look then at some machine learning algorithms. So that the models can tune themselves as you get better at understanding the key levers for performance and what the customer behavior indicates in terms of your performance metrics. So, again, just to summarize what we’ve been talking about here with data science, it’s absolutely critical as your technology foundation when you start down this path of becoming a customer led organization.

Technique 3: Experiment design (combining data with art)

– Cool. So this brings me to the next technique. The next technique is good both for precision and speed. I call it a general experiment design. And I think that before we cover the three experiments, the three types of experiments, I think that this is where you guys can showcase your art. So experiment design, it’s definitely something human. So in the future, when machines take over and all of our jobs will be replaced by machine learning and stuff like that, these are one of the few things that people will still do. That’s my prediction and I think it’s very much a human trait, and it’s what humans excel at. So if we think about every one of our campaigns as a marketing experiment, and we design it like that, we can then learn from it, improve, and this would give us a better campaign the next time. A better conversation starter which will resonate more with our customers, which will generate more lifetime value. So this is both good for speed and precision because it’s basically the main thing that goes on. Sorry. And we identify three types of experiments.

So there’s two types of experiments with just one interaction. I’m sending this campaign on Friday, and I’m going to talk to my customers. In those senses, in this sense, we have one interaction with just one message. So I have one thing I’m saying. I have one interaction with a few messages, let’s say an A/B variation or something like that. And then I have two completely different communication streams. Which is not one interaction, it’s a series of interactions happening to achieve a specific goal or to utilize a specific tactic. So let’s go over them one by one. So in terms of one interaction with one message, it’s the simplest experiment. I’m doing something on Friday. I’m using maybe one channel, just email as an example, and I’m offering my customers a 10% discount.

Step 1: Isolate a control group

Now, the very basic thing to do an experiment design in this sense is to make sure you isolate a control group. Test versus control, the basic test versus control. And you will actually be surprised. This may seem basic but we still see a ton of companies not doing that. And it’s just a shame, it’s one of the key levers of understanding, constant improvement, getting to know true measurable impact. And just as a basic example here, you can see that I’m sending this campaign to 673,000 customers, and I have 35,000 customers as my control book. I see that in terms of the average response rate, the test did beat the control. Not by a lot, but this basically accounts for why we made money. And we use a statistical test to validate, beyond any reasonable doubt, that the impact, the cause for this increase, is the campaign that we ran for our customer. So this is very basic but it’s powerful. The first thing you do when designing an experiment is make sure you isolate the control group. The next one is one interaction with a few message alternatives.

Step 2: Test several marketing message alternatives per interaction

So again, I’m running this campaign on Friday, maybe I’m still using only e-mail, but now I have two different offers. So I’m doing 10% discount versus 20% discount, as an example, and I want to see what’s better. And a basic A/B test is going to tell me, “Hey, B is the winner, you got better results of B. Customers bought far more and stuff like that.” So then I’m going to say, “Okay, next time I’m just going to run this thing only for a… I’m going to run this campaign and I’m going to offer only B, because now I know who’s the winner.” But the problem is you’re running this campaign on this segment and to many, many sub-segments of this campaign.

So if I’m choosing a winner based on the aggregate, maybe I’m leaving money on the table. And this is where you can utilize something we call a self-optimized campaign. Which means that there’s no just one winner. I can look at a lot of sub-segments and a lot of micro-segments or clusters, and see that maybe the overall action, the overall winner, is A. But in this micro-segment over here, B is the winner. And in that micro-segment it’s C is the winner. And just do a smart mix and match and offer every cluster the message that resonated the best with them. And this campaign keeps on learning and tweaking as time goes by and being smarter and smarter all the time. It requires a good foundation of subsegments or micro segments as we like to call them. It requires this multi-armed bandit approach to constantly iterate and learn what works better. And just as this understanding there’s no loser, there’s just like the best action for every micro segment and not fill the entire aggregate. So that’s that and this brings us to the next notion of communication streams.

Step 3: Avoid cannibalization – compare communication streams

So let’s think about it. When I’m running an individual campaign, I actually have a problem. What if I have hidden cannibalization? And what I want to say when I say that is, I mean, what happens if you are seeing an uplift on a campaign, but actually all you did is you got the revenue sooner? So I’m a customer. I want to buy now. I’m expecting a baby soon and I want to buy a bassinet. I’m planning to buy that bassinet for sure. I’m going to do it, I’m expecting the baby in a month’s time. So I was planning probably to buy it a week before the due date, and I just got an amazing coupon from this company, and I’m going to buy it now. So you’re going to see uplift against the control and you are going to say, “Yeah, we just made a lot of money.” But essentially all you did was you got the same money from me as a consumer but you got it sooner. But it’s not real uplift, right? So we call this hidden cannibalization.

How do I know that I’m actually pushing lifetime value up? How can I prove it? And we have this concept of forever control, it’s the only way to measure it. So think about that. If I have this segment of customers who came back from churn, and I see that 62% of them go back to churn, and they churn again, they don’t become active, let’s imagine I want to run this strategy to change that. I want to run a different strategy, two different strategies on customers who came back from churn, and I want to try it for three months. And it’s here, over here, and I want to use different marketing programs.

So I want to do the test stream. It’s going to be 80% of my customers. I want to try a program with many different offers and segments of the control stream. I’m just going to do what I’m doing today, and I’m going to run it for three months. And the main thing is in this case, the test and control, they’re not being separated for individual campaign, they’re being separated to two completely different streams. People in the control stream do not see the messages of the test stream and vice versa. And what this gives me after three months, I can see, “You know what? Look at that?” People in the test stream, I got 50% of them to become active. That’s amazing, right? Instead of 38. This means that I did impact lifetime value, I did create a great difference by trying out this different strategy. So that’s kind of like the top level, the highest level of experiment design. Rusty, over to you.

Why good experiment design is crucial for successful marketing

– What you just articulated there demonstrates the need that we see to balance what we would call systems of insight with systems of engagement. If you think of all of that data science that we talked about in the previous section as a system of insight, it’s going to deliver you a lot of intelligence that you can use, but it’s pretty worthless if it’s not actionable and if you don’t put it to use to engage with the customers. So what we mean when we say align the systems of insight is to leverage that data, that deep detailed understanding of the customer to begin to develop the appropriate content to communicate with the customers on the channels that they prefer.

And then you’ll want to not only personalize, but to Pini’s point, you’ll want to optimize what you’re doing across those various channels. And that will mean becoming the type of organization that would continuously leverage that data to continuously optimize. It’s not something that you apply on a one-off basis. It’s something that you have to do continually so that you can begin to understand where the value on the uplift is coming from, what is driving those changes in customer behavior that you’re seeing, and how to continue to invest in the strategies that are going to bring you the best return.

Technique 4: Cater to the customer journey – adding precision to speed

– Great. Thanks a lot for that, Rusty. And this takes me to the last point of catering to the customer journey, like the last technique. So this technique will give you precision. Actually, it’s also going to give you speed, it’s debatable but it’s very important. So everybody talks about the customer journey, how do we cater for the customer journey, and that’s crucial in terms of moving at the speed of your customer. So I want to present two different approaches.

So the first approach is the more popular approach. The one that you see in most marketing automation systems, you see the static journey. So the static journey is essentially, “I’m a marketer and I’m going to blueprint exactly what I’m going to offer a group of customers that I’m going to take them through this journey.” And I decide as a marketer, “I’m going to force them into this journey and I’m going to build the decision points and I’m going to build the flexibility of the journey based on everything I blueprint.”

So the advantages of this, it’s highly clear when you show it to your executives, this is what we’re doing with our customers. You can show a flowchart, everybody understands a flow chart. It’s really cool. The big disadvantage and I…with all honestly, I think that in most … I’ve never seen a very sophisticated organization use this because what happens is it breaks. At some point it’s not scalable. And the reason for that is that, and I have this nice quote here from John Lennon, “Life is what happens while you’re busy making other plans.” So we cannot plan for everything. The marketer cannot blueprint every possible scenario on earth because it’s really, really difficult. First of all, I don’t know everything that happens in the data. I’m not sure how to design these journeys. And even if I ever create it, and and I do know how to design a journey, it’s still static and there’s many things I’m not going to plan for.

As an example, let’s say, and you’ve seen this in illustrations, I say, “Do this and then wait two days or wait four days, and then do that.” But what if during those four days that I’m supposed to wait, the customer calls the call center and something really bad happened. The customer just had a horrible experience. Us as marketers, we want to start a conversation now. We have a reason to start a conversation. We need to apologize, we need to rectify what happened, and we need to do something. But we’re still stuck in our blueprint because we didn’t plan for that specific thing. And that’s not moving at the speed of your customers.

Marketing optimized for customers journeys is dynamic and personalized

The way to truly move at the speed of your customers is to use something we call the dynamic journey, which creates a much simpler exercise for the marketer. We’re saying instead of thinking of the entire journey and blueprinting the entire thing, just think about individual intervention points. Just think about individual things, individual points in the customer life cycle where you want to start a conversation, where you have something to say. For example, one of them would be customer just had a very bad experience and called the call center. So this would be one of these rows. And this blue box would be one of these campaigns. But in this way, whenever a customer is changing the data, they move to the next intervention point. And you essentially let the customers plot their own path. Each customer has their own individual journey. There’s no pre-defined journey, there’s just different like points and, sorry, different steps of the journey, but the customers themselves, they jump between the steps as their data changes.

So this is the key difference and you see it happens across all the channels. And it’s actually easier for the marketer and it’s a lot more scalable because it’s just one map. It’s just one place that I am going to insert various types of conversation starters, various types of intervention plans. And when the customers want that change in the data, they are just going to jump between them and each customer is going to have a different journey based on the way they change in the data. And great. Over to you, Rusty.

Optimizing for the future and Optimove: Embracing the data revolution

– Sure. Well, Pini, I think that what you just described there leads us to this statement at the top of the page here. And I try not to deal in scare tactics or hyperbole, but I do believe in this case that the evolution that we’ve been talking about here is necessary to the very survival of brands. I think that you need to understand your customers, you need to do that with data science. You then need to make sure how you interact with the customers and make sure that what you’re doing is meaningful to them in the moment. So that you can cater to that entire customer driven journey that you just described. And of course, beyond all of those technology components is the structure of the organization so that it has the speed and precision to be able to succeed. So, again, I think, without turning up the contrast too much here, it is necessary to embrace this evolution and start to look at some of these organizational and technological changes that will let you become more relevant and more contextual to your customers and their journeys with you as a brand. And with that, I’ll hand it back to you, Pini, and to Amit.

– Yeah. Thanks a lot, Rusty. I think we’re basically done with the presentation. I had a lot of fun and I hope you did, too, Rusty. And I think now is a good time to take questions from our audience. Amit, back to you.

How does Optimove get feedback from every customer?

– Thank you very much, Pini. Thank you very much, Rusty. We’re receiving some questions. So for all of our audience, please continue on to get your questions in. First question we have is for you, Rusty. This question came in from Jonathan. He was thinking about a two-way dialog with a customer. “How do we get feedback from every customer we interact with?”

– Well, that will depend a lot on the different channels that you’ll be using to interact with those customers. But there are technologies available to capture especially the digital interactions. But also there are ways to begin to capture those offline or face-to-face interactions as well. And I think all of those data points are necessary if you’re going to build up that detailed accurate understanding of the customer behavior. Remember what Pini said there at the end with regard to the customer journey. The journey for each individual is going to be different and ultimately that journey will be driven by the customer, not you as a brand or you as a marketing organization. So being able to capture that interaction data will be absolutely critical to you being able to understand the different paths that the customer might take in the way they interact with you.

– Yeah. That makes perfect sense. Thank you for that, Rusty. Pini, is there anything you want to add on top of that?

– No, I’m good.

How does Optimove manage several communication streams?

– So another question that came in from a Jonathan. Now, this one is for you, Pini. “How technically is it possible to manage several streams of communications?”

– Yeah. So actually managing a few streams of communication is pretty challenging but essentially it starts from this very notion kind of like having in your customer database or in your single customer view, you need to make sure that every customer has some kind of a tag, a random number tag, associated to them. So for example, imagine that every one of your customers in your database has a number between one to ten, and this number just stays there. And when you acquire new customers, again, each one of them gets one of these numbers between one to ten, a random number. And then you can say, for example, when you start to design those streams, you’re going to say, “Okay. All the people with numbers one to two,” so you build the stream for, let’s say, for active customers.

So all of the active customers with random number one and two, they’re going to go to the control stream. All of the active customers with random numbers three to ten, they’re going to go to the test stream. And then basically you’re going to build your programs and automations on this kind of like a split. But you need basically the tech to do that. You can also do it obviously generally if you’re using your old data science departments and/or database departments. It’s not extremely difficult but it does require some care and it requires tending. So you do need to manage that. But, again, the benefit once you’ve completed, the level of clarity and insights is going to be tremendous. You’re really going to know what your serum is doing. If it’s really working and moving the needle.

Transitioning to an marketing organization built for speed

– Thank you for that, Pini. Another question directed to both of you guys. Pini you mentioned org built for speed. What does this transition look like and how long does it take and which technology are involved in this until you achieve the Holy Grail of the studio?

– Rusty, you want to start?

– Sure, I can jump in first. Well, I have yet to find an organization that is at a place where they can say they’re done with this evolution. I think it’s something that you continuously work toward. But as you get better, the goalposts will continue to move and you’ll look to improve more and more on top of where you are. When we look at an organization’s readiness or maturity to tackle this kind of journey, we start by looking at where they’re good today. And that usually means that they have operational mastery in some areas where they might be really good at email or they’re really good at optimizing the e-commerce site, or perhaps their customer service is really where it needs to be but perhaps marketing haven’t caught up.

So we try to look at where they’re good and where they have those operational efficiencies that we can then extend to the rest of the organization. We then look at how would you make that work across functions and across the different technologies and people that are customer facing. Once you’ve got that alignment, then you can think about optimizing. And when you get to that point, that’s where, like I said, it becomes a continuous goal that you have as an organization to get better every day. So you’ll always find different things that you can optimize to get better at. We do define an end state as real-time unification. But, as I said, I have yet to find an organization that can claim they are fully and truly unified across every customer touch point and across every function in the organization. But I think that’s good to have as an aspiration that’s out there in front of you that you’re always working toward.

– Great. And my take on this, Amit, is that I agree with Rusty 100% that the goalpost is going to keep moving, and there’s not many organizations that do it to the fullest extent. I think it’s two thoughts. The first thought is that it has to do with size. So smaller marketing organizations, yes, do it to the fullest extent and they could be super fast. Especially if there’s one person doing everything, that’s the fastest thing. The fastest thing is that there’s no inter-department communication, nothing gets stuck. Just one person does the whole thing, and that person can push out lots and lots and lots of new ideas and experiments into the system. When you grow and then the marketing order becomes bigger, there’s more need for governance, there’s more need for procedures, and that basically slows you down.

The way to combat it is just if you have an assembly line and you have a few departments, there’s an old notion in the field of assembly lines in industry and production, it’s called cross training. So essentially if the marketer that typically only writes the brief, if that marketer now can also create the database, whatever tool they’re using, if they’re using a tool like Optimove or another tool that democratizes data analysis or segment creation for the marketers, you take one piece of another step in the assembly line and you empower another team. So all of a sudden, one team can take you through steps one to three and then you only need to complete four and five. Instead of before just doing one, then another department does two, another does three, and so on and so forth. So the more cross training you do, the more you empower different pieces of that assembly line to do more tasks, the faster you can get. That’s just it, You add another cross training, if you allow the data people to also do some, if they can do some creatives with some types of campaigns, again, it gives you speed. So that’s how I would go about it.

Insight vs. engagement: what is the ideal balance between analysis and engagement?

– Thank you, Pini and Rusty. We have time for just a couple of more questions, so I’ll jump into a question from Daphne. “Insight versus engagement. Is there an ideal balance between analysis and engagement?” Rusty, since you brought up that point, I’ll direct this question to you.

– Sure. Well, they definitely need to be in balance and it becomes very difficult to give you a prescription for that without understanding the organization and the goals that you have, and whether you’re fully online, whether you’re online and offline as well. But I’ll give you a good rule of thumb. What you would like to do is minimize the gap between uncovering an insight that you think would make a difference to the business, and your ability to leverage that insight to engage with a customer and either prove that you were right with your prediction or prove that you were wrong and move on to the next thing. So if you can get to a point where you think you have the optimal time between when an insight is raised and when you’re able to take action on it and engage with the customer, I’d say you have the balance right. And I know that’s not a precise answer, but that’s because it will differ by every organization.

– Thank you for that, Rusty. One last question that came in from Peter. “What would you say is the optimal test period time test when doing an A/B testing? Is there anything at all as ideal or minimal test period time?” Pini, you’re probably the expert for this one.

– Yeah, I would just kind of like, I would use… First of all try to understand that we need to remember as marketers that one of the biggest benefits that we have in terms of looking at the data is we’re analyzing data about people. We’re not analyzing dots and space. When we analyze dots and space, we have zero intuition, we have zero understanding. So when you design the experiment, your common sense will prove itself to be very valuable. So obviously it depends on the type of A/B test for the thing you want to do. But the type of the campaign you’re running will tell you if you should go about it in days or weeks or something like that. And you should take into account the fact that today’s world, people see lots of messages, as Rusty said before. So I would think typically it’s within days.

So then the question is whether it’s one day, two day, three days, and some testing will tell you. If you see the metrics change over a longer period of time. For example, if the audience needs more time to show you the real results, so you need more than a day, you need two, three, four or five. A little bit of testing, a few variations will tell you the magic number for your business. Just try to see if there’s a massive difference in response. If customers who saw that message are still responding after like the third, fourth, fifth day or, if I had to give like one magic number, I would go for three days. But that’s obviously very generic and not taking into account various nuances that could happen in a given campaign or even for a given company.

– Thank you very much, Pini. This gets us to the end of our time here today. I’d like to thank you, Rusty and Pini, very much for your time. I’d like to thank all of the audience for being here with us today. We have a lot of further resources for marketer and CRM practitioners at our website at We will be sharing a survey at the end of this webinar in order for you guys to be able to help us improve for our future webinars. So if you have a few seconds to fill that in, we would be very thankful. Thank you very much and have a lovely rest of the day.