Video Transcript

– [Tanya] Today we’re going to talk about how to make sure we’re targeting the customers, and actually really reaching them. But before we start, let’s examine a few examples. So here are the customers we’re leaving behind, let’s say I created this amazing plan that goes out to 100% of my customers. I always have a portion of the customers who are opted-out and I need to think what’s the best way to reach those customers. I’d see that most of my customers are opted-in and receiving my emails, but however, in many cases they might ignore the form of a communication I chose to use. So, they either don’t go inside their promotion folder and up in Gmail or they, you know, just swipe out the push notification they just received without even looking at it.

In many cases, my customers will only respond to campaigns that are sent during the workdays, so sending weekend campaigns to these customers won’t have any impact at all. And let’s say my customer is updating, he sees my email, I sent it in the right time. However, I chose to use free shipping rather than a discount, which is what he really appreciates. Bottom line, even though it seems like I’m, you know, reaching all my customers, I’m only left with a part of them that are actually touched. Today we’re going to talk about how to combine with the right channel, the right timing, and the right promotion in order to actually reach 100% of my customers. The hard part will be to understand what is the right, but let’s start.

– [Noa] Okay, so how can we discover the right channel? Now, there are many direct and indirect ways to connect with our customers, but you know, using all available channels comes at a price, so when we come to choose a channel, we need to consider its relative strength and weaknesses versus the other channels. Now, we usually take in consideration the cost, how expensive it is to connect with this customer, with this channel? How much attention does it drive? Is it effective in driving my customer’s attention? Its design options, how flexible can I get with the templates that I send via this channel? And how intrusive it is, how irritating it might get for each one of our customers when connecting with this channel.

However, the thing that we most need to consider, but we don’t, is the customer’s preference or better yet, where is our customers are most contactable and what channel will best increase their likelihood to respond? Now, a research data that was conducted recently, show that 51% of the U.S. consumers respond better and are more loyal to brands that interact with them via their favorite media channel. You know, at the end of the day, we don’t want to put a million dollars on a campaign with a channel that doesn’t have enough impact. So, let’s see how we’re following these three steps in order to get the preferred channel.

The first step will be to collect the relevant data. I want to know everything that my provider knows. Like, I want to be able to measure my customer’s response based on the channel that I use. So, for each one of my campaigns, I want to know each customer, to watch campaigns, who opened, clicked, or unsubscribed from. Now once I have the data, I want to design a test, you know, a credible test that will enable me to compare between the different channels. So, for that, I will have to create a similar environment, you know? For example, sending out the same promotion on the same day of the week. You know, this week with one channel, on the following week with another channel, so that way I can understand to which one of the channels my customer responded better. Now, while conducting all these tests, I want to also make sure that I’m generally analyzing each one of the channel and customer’s response to it. For example, if we take a customer that did not open four SMSs in a row, what is his likelihood to open the next one? So that will help me understand what is the minimum number of campaigns that I want to send before I can, you know, determine what is the preferred channel.

If you want to hear more about testing and optimizing, the next session is with Yochai and Tal about Test, Improve, and Repeat, so that can be very interesting for you. Let’s move on to the next step.

The next step is taking all the data I have and the data I collected from the tests and create one single customer view for each one of my customers. So, I’ll take all relevant metrics for all the channels, and I will put them in one single customer view. So that will be the email open rate, you know, by the SMS open rate and for each one of the channels that I want to, basically test. So, let’s see an example of exactly how it’s done. The first step is getting all the data. After I get all the data for my providers, I can put it in one table in this structure. That will be all campaigns sent out to all customers by date, by channel, and most importantly, if they responded or not. Now, a response can be either an open or a click of this campaign. It can also be an actual transaction, if the customer actually came and purchased or made a deposit, for example. After I have all the data, I have the data structure. I start sending out campaigns, you know, channel by channel, and based on the test that I’ve planned of course.

So, I’ll send the first campaign, second campaign, and so on. For each one, I will test if, you know, my customer responded or not to each one of the campaigns. And I will make sure to send enough emails or enough communication based on the channel, based on what I’ve determined before. We recommend sending at least four communication when sending out these campaigns. Now after I do that, I can calculate the response rate, which will be the number of response divided by the number of campaigns that I sent out. I will do the same for all those channels that I want to test, of course. Here for example, we have also email and direct mail, and then I will be able to structure it in a single customer view.

So, for this customer over here, I have the SMS response rate, email response rate, direct mail response rate, and so on. But the main purpose is to create the preferred channel after all. So here is SMS, but let’s see how we actually, you know, decide what it’s going to be. So, to determine that, we can go with open click-rate. We can choose the preferred channel as the one with the highest open or click-rate. For example, over here, we have a customer that, you know, opened more communication from Web Push than it did with, you know, other channels, so that will be his preferred channel. We can also go with a response rate in actual transaction. A customer that we know we communicated in one channel, for this example over here it’s push notification, and this channel made him, you know, come and purchase more than other channels did, but we recommend actually to use the combination of both, the response will be defined by either open or click, or either an actual transaction to make sure that we’re covering, you know, all exposure of the customer.

Now in the real world, we know that you don’t always have all the data that you need from your providers. I mean, it should be quite easy to get, but sometimes it’s just not possible in the way that you want to use it. So, we can still optimize our communication and still use a preferred channel based on, you know, data that we do have or we can collect. So, for example, we have customer activity. If I have a customer that is more active on mobile, for example has, you know, 70% of his activity on mobile, then it makes more sense to make his preferred channel as a mobile channel push notification or SMS. We can also conduct a survey. Now, a survey’s… You know, actually asking the customers, “What is your choice? What do you prefer?” But we do need to keep in mind that not always the customer choice is, you know, the channel that they best respond to. So, we’ll have that as a note. Now, if I have a customer that responds to all channels in the same rate, well for this customer that will be a piece of cake. I will optimize my communication based on the campaign objective or the factor that we’ve discussed before, the attention, the cost, and so on. And if I have a customer that does not respond at all, you know, if you have a lot of customers of those, that will be a great opportunity to, you know, consider adding an additional way of communication, adding another channel.

– So now that we understood what channel we’re going to use for each one of my customers, we now need to ask ourselves, “When? When is the right timing to send that communication?” So what you see here is a heat map. You see the different hours and the different days, and we’re going to use this to categorize my customer’s activity, and basically the activity distribution between the different times of day. Now, an activity can be either opening an email or a log-in, but it can also be actual activity. So, in order to have a simpler approach I can, you know, divide my weekdays to my, yeah, my weekdays to workdays and weekend, and it can also split my times of day into three, morning, afternoon, evening, and night just to have a simpler approach. Now with a heat map I can see that you probably don’t see the different numbers inside, but you can see that most of the customer’s activity is concentrated in the weekdays during the morning. I’m going to do so on a customer level, right? So, I can do the same, understand each one of my customer’s distribution between the different weekdays and between the different hours. And by that, understand what is each one of the customer’s preference.

So, in order to understand the favorite day, for example, what I did here is I just summed up the columns here and got to this bar chart. This is the activity distribution by weekday for a specific customer. I see that Thursday is his day with his most activity. I see that also, Tuesday is a highly active day, but I also see that the weekends are fairly weaker. So, for this customer, my preferred day will be Thursday. I’m going to do the same for the activity distribution by the times of day. And as I said before, I’m just going to split it into three, so morning, afternoon, and night. I see that for this customer, over 45% of his activity was done in the morning. Okay, so for this customer, the preferred time, it will be morning. Now, we need to keep in mind that in some cases I have preferred timing based on different channels, so for each channel, a different preferred time. However, let’s keep it simple and just use the customer’s activity for now.

So, I have the preferred day of a customer and I have the preferred timing of a customer, but what do I do with this? How do I translate it to actual or to practical use? So, with this, we have two approaches. We can either target the customer when his chances to react are organically higher because we know that that’s his preference, which is for example, Thursday. Or we can try to cross-sell this customer to a different day and send the communication, let’s say on Monday. Here, I would recommend to use additional segmentation. So, you probably segment your customers by life-cycle stages, right? New, active, churn. With customers who are active, or we just saw recently, I would try to, you know, expand this bin to a different day and send this communication on Monday. However, with a churn customer, as his chances to react to my campaign are naturally lower, I would send the campaign on Thursday in order to increase his chances to react and reactivate. However, we will need to test, okay. So, let’s try to send all those customers whose preferred day is Thursday, a test once on Monday, once on Thursday, and make sure to keep a portion of the customers to serve as a control group so we make sure that we can, you know, actually reach conclusions.

– So, we have the right timing. We have the right channel. Now, what are we going to offer to our customer? What will be the next promotion? So, does your customer appreciate all kinds of discount or only discount that are greater than 30%? Or, does your customer buys only when there’s a sale? Now, this is a place where we want to understand what works for each customer. Well, we have a lot of, you know, promotion types out there and we’ve listed here only few, but first in order to understand what works better, we want to start with a general A/B testing. Now, we can test various things. So, for example, we can test the percentages of discount that we give. For example, 30% versus 50%.

Another A/B test that we can use, that we can do is the promotion type itself, so we will test what works better, free shipping over $50 or get free gift over $50. And additional tests that we can do is testing the actual content of, you know, the message that we send, will it be two plus one versus get the third one for free? So, after we understand generally how our customers behave, we want to dig it in deeper. You know, for each one of the customers, what works better individually? So, if you have a lot of promotion types, then it makes more sense to group your promotion into promotion types, and then for each one of the promotion types, we will calculate the promotion redemption rate, so that will be number of promotions redeemed divided by the number of promotion that were granted to each customer. And that way we can understand what works better for this customer. Will it be one plus one as we see over here, percent of discount and so on?

So, this one, you know, this customer actually prefers more the one plus one. And then we have it an additional attribute for each customer, the favorite promotion. Now, we do need to keep in mind that if we’re not equally offering the promotions and we were sending out more one plus one than we do percentage of discount, then we will calculate the same rate. We will just also use the weighted average based on how you hand out your promotions. Now, if you have more, you know, discount, you don’t have a lot of promotion types, you have more discount offers, for example, then we’ll like to use over here, the discount sensitivity, that will be the minimum discount that the customer has redeemed.

So, for example, if I’ve sent out, you know, many percentages of discount to my customer, I will check what is the lowest offer that, you know, lowest promotion that the customer has actually responded to and made him come and make a purchase. Now, we can also add the average discount and the discount sensitivity, but what, you know, matters the most is taking all those power measures that we have discussed, that help us realize what the customer has preferred and combine it with the customer’s purchase history. Understanding, you know, the quality of my relationship with the customer combined with those power measures, will help us and support our decision of what promotion we can give him next based on his preference and its value.

– Understanding these three elements is the key to right communication. So, for each one of the customers, we’ll create a single customer view and we’ll add to the aggregated, you know, attributes that we already have for this customer. Also, attributes are related to these three, what is his preferred channel, what preferred timing, and what is the preferred promotion in order to actually reach 100% of my customers. This is the least we can do in order to have an impact. We hope you gain some insights and thanks for listening.