Mapping the Consumer Genome: Ep 6
Featuring a conversation with Nikhil Chandurkar, formerly of Kohl’s on:
- Automating customer retention strategies in a world of machine learning
- Creating differentiated experiences for each unique shopping journey
Conversation with Nikhil Chandurkar, formerly Director of Product Management at Kohl’s
00:07 | Debjani:
Nikhil, welcome. Before we start, I’d love to give our viewers a little bit about your background and your experience over the last few years.
00:17 | Nikhil:
Certainly! Thanks for having me, much appreciated. I was with Kohl’s for about four and a half years, primarily focused on personalization, marketing automation, and cart and checkout experiences on the digital side, of course. Prior to Kohl’s, I was with Symantec as part of their Norton.com e-commerce platform, and prior to that, with Yahoo and eBay, doing mostly e-commerce related work. But personalization has been a common theme across all my different stints at different companies, so I really enjoy working on personalization – that’s my passion.
What technological shifts have transpired recently?
01:06 | Debjani:
Excellent. I think of yourself as one of the pioneers in what I think about as in-session machine learning and what you can do with it. Talk to me a little bit about your experiences there, what you were trying to do, and what you achieved with that.
01:23 | Nikhil:
At a broad level, differentiated experiences are a transformative way to personalize in real-time to a visitor’s unique shopping journey. As you can imagine, in a retail situation, typically, we assume a common pattern for a given customer, but that happens not to be true.
Sometimes, the customer is in a hurry to make a decision and wants to just get the things that they want in a hurry, so urgency is of the essence there. In some cases, they are safety-driven; let’s say, they’re buying a car seat for their child, so safety is paramount in those scenarios. And then, there could be other shopping journeys which have different nuances; it could be the same person, and they may have the same overall tendencies, but they do have a narrower focus. And maybe you need to focus for a given shopping journey. So how do we tap into those? How do we understand what those nuances are for this starting shopping journey to personalize it in real-time?
Over the years, that’s been my focus, especially at Kohl’s; and we came up with a solution called ‘dynamic real-time offers’. That was based on identifying an ‘on-the-fence shopper’, a shopper who is not yet quite sure that they want to make a purchase, incentivizing them based on what we feel is their current shopping journey insights, and giving them a meaningful offer to have them convert. There were use cases that we looked at in terms of, hey, we probably think this customer is in need of value amplification, so we would message them about the value of the product, in terms of, hey, this is a product that’s selling fast or this product is on sale. Those were the insights that we leveraged and found a way to embed in the experience to make it more compelling, more personal in that specific customer journey.
Creating differentiated experiences for each unique shopping journey
03:32 | Debjani:
I think what you are talking about is predicting a whole range of consumer behavior in-the-moment, and that predictive capability is leading you to deploy these experiences at that moment. What are the things about consumer behavior that are most valuable in-session?
03:53 | Nikhil:
There are two things that I’ve used extensively in my experience. One is, in the current shopping journey, is it a more value focused journey in the sense that – is price a big factor, or the only factor, or a major factor? Or, is it the selection of the product -is the product itself, you know, some uncertainty around that. If we could be proactive in determining whether a customer is price-sensitive or product-sensitive, you amplify different aspects of the product or the value to help the customer make the decision. Those were the two key insights that we used, both historical assessment of the customer which is the typical aspects in terms of propensity scores and whatnot, but also real-time because, like I said in the beginning, a journey is kind of unique even though the customer might be the same. And then using those real-time signals not just based on the current shopper’s journey, but also broadly, based on what is happening in real-time right now. Things like weather, or maybe even their location. Those are the signals that we have used to try and get more personal with personalization as part of these differentiated experiences.
Predictive capabilities at the top of the funnel vs. mid-funnel
05:19 | Debjani:
Excellent. Excellent. You know, what you and I are talking about right now is in-session mid-funnel and the value therefore it creates. Talk to me a little bit about the trade-off between using it at the top of the funnel, which has been done for years and years and years, and those predictive capabilities mid-funnel. How does one make that determination, and what is the value trade-off thereof?
05:41 | Nikhil:
If you look at top of the funnel, you’re likely to have a lot less real-time data about the customer, and you might have no data about the customer because it could be an unknown customer, a new customer, a first-time customer, whatever you want to call it, right? So top of the funnel is a bit of a challenge, and most of the time a typical pattern is what I call ‘wisdom of the crowd personalization’ – what’s the most popular item for newbie customers. We tend to use those kinds of insights to deliver experiences top of the funnel.
As you go a little deeper into the middle of the funnel, you have some journey aspects of the customer. They could have used certain products, they could have engaged with certain parts of the site that would give you some clues as to what the shopping journey is, and those insights, along with if we know the customer really well, when you put them together could create a compelling experience. So definitely, you have a lot more data, both real-time and historical, as the customer goes deeper into the funnel as opposed to in the beginning part of the funnel.
Now, at the top of the funnel, a lot of new data points have been used: location is one thing that is critical, especially with people on their mobile devices; weather factors are used quite a bit nowadays. So in addition to just the ‘wisdom of the crowd’, knowing where the customer is and what weather is in there, what weather they are experiencing, could be two great inputs that could allow you to be more personal in-the-moment for top of the funnel personalization.
07:32 | Debjani:
Yes, I think the focus has moved more and more towards mid-funnel engagement, loyalty, and in-session behavior as technologies have caught up to being able to react to those events as they’re happening. Those capabilities are becoming more mature in regards to event-driven, in regards to mapping the genome, as we call it. What’s the trade-off between what we’ve seen typically being deployed for marketing vs. these mid-funnel type things which are a lot more complex? How do you, as a product owner, enable that?
Automating retention strategies in a world of machine learning
08:11 | Nikhil:
Right. If I put my marketing hat on, there is the acquisition funnel and there is a retention funnel. More and more retailers like Kohl’s are paying equal attention to retention as they did with acquisition. As the landscape is changing, it’s very clear that retaining our existing customers is paramount. The cost of acquisition is also going up because customers have lots of choices, and “normal” retailers are competing with some really entrenched players like Amazon and Walmart.
08:54 | Nikhil:
Obviously, embedded in retention is loyalty, in the way you spoke about it. Traditionally, what has been happening is: marketing creates some course of life cycle segments that are manually curated based on some data pool from their data mart or big data systems, and they launch campaigns against those customer audiences. But there is no feedback loop, and everything is very manual and results in a lot of customers just falling through the cracks.
With the advent of automation and machine learning, both of these aspects are getting completely automated in the sense that you have life cycle segments being automatically managed. When a customer is moving from, let’s say, a newbie customer to a retained customer, the system is automatically aware and, sort of, maintaining the state of the customer. If the customer isn’t moving in the right direction, triggers or signals are sent out; these signals can be then acted upon through experiential changes or incentives that you want to deploy. That’s the newer way of doing things. It’s no longer about doing the heavy lifting of building these segments, targeting these segments, running campaigns. This is a point I want to emphasize as people are moving away from campaigns at large.
Imagine a world without campaigns, where things are completely automated and no customer slips through the cracks. Once we get the signals from these automated systems that leverage machine learning, what are the tactics that we should put forth to try and gain the confidence of a customer who is about to churn? It depends on which part of the life cycle they are, it could depend upon customer life cycle value, it could depend on their propensity for certain things. That’s the paradigm shift I’m seeing on the marketing side, especially when it comes to retention.
The role of data for customer acquisition
11:07 | Nikhil:
When it comes to acquisition, like you said, the data is a big challenge. In a company where I had a situation where I had worked on acquisition funnel, we used third-party data and first-party data to try enrich the experiences when it comes to acquisition. But it’s all about how much data you have that will allow you to personalize; and the fact that data isn’t quite commonly available, and people are sensitive to security concerns about their personal data being shared and whatnot. I think acquisition is going to be a lot more challenging going forward; retention is going to be a lot easier. Typically, marketing people will tell you that every dollar you invest in retention, it requires 5x that much money to acquire a customer. So, retaining a customer is a lot cheaper, primarily because you have so much data about the customer and it’s a lot easier, you have the ability to influence them.
12:11 | Debjani:
I love the notion of the campaign-less world because essentially what you are saying is that the ML (machine learning) or the intelligence determines what the state of the customer is, and we will have predefined actions for each state. So it’s the state that determines the actions, and the state is determined by intelligence, behavior, location, weather, whatever the case may be.
12:39 | Nikhil:
Yep, near-real-time actions that the customers are taking or maybe not taking and, of course, the historical information that you have about the customer. I think that’s a really powerful area where lots of investment is happening right now.
Creating differentiated experiences
12:59 | Debjani:
I’m assuming this campaign-less vision of the world is actually omnichannel, cross-channel longitudinal views, right?
13:14 | Nikhil:
Exactly. In the beginning, the focus of marketing was on how I can convert the customer right now. But from a marketing and retention standpoint, it’s about the next transaction and the next transaction… not necessarily about the current transaction. So how can the experiences be modified in real-time through the concept I call differential experience that will allow for that to happen? What can I do for Debjani right now? I know she’s interested in this purse, but I want her to come back and maybe buy more. What can I do in this session to influence that? Not just the current purchase, but think about the future purchase as well.
14:02 | Debjani:
The notion truly is that in this purchase, context decays with time, and the immediate context is the most powerful in determining what that user wants to do in this moment. So if you can harness that, that really gives you the highest ROI in the type of experiences you deploy.
14:26 | Nikhil:
What does the next generation marketing cloud look like?
14:28 | Debjani:
So what’s next, Nikhil? At ZineOne, we think about the next generation cloud as being predictive, ML-driven, real-time, event-driven, location-driven. If you were designing the next-generation cloud, what would it be?
14:46 | Nikhil:
From a customer experience point of view, I think there is this whole thing about conversational experiences now. When I say conversational experiences, I mean more than just the Amazon Echo’s of the world or Siri’s of the world, having a voice recognition way of servicing customer requests, it’s more than that. I know, for example, there were some tests done by Amazon to recognize patterns of the customer’s voice. If the customer’s voice sounds stuffy, maybe they have a common cold, and maybe you could recommend some common cold medication… getting more personal with the conversation. Another example, CoverGirl has this virtual try experience, where somebody could try out makeup with their face in a virtual setting. Conversational experiences go beyond just digital; it is also what can happen inside a store or outside of the digital space. That is one area I feel like lots of innovation is happening.
Empathy at scale
16:10 | Nikhil:
The other piece is what I call empathy at scale. How can we become even more personal with personalization? What I mean by that is, to empathize means to actually know what the customer is feeling, and to be predictive about it. One simple example is, I think Amazon does this, if for example, they’re unable to deliver a certain product on the timeframe that they have committed, they promise aggressive shipping. For some reason, the product doesn’t arrive. They know it’s not going to arrive in the timeframe that they promised. So they will just cancel the order, refund the value of the order, and maybe give you an incentive or a recommendation for a different product. All done proactively. Again, not really machine learning but it is empathy at scale. You’re trying to do the simple things before the customer can experience any negativity. I think that is where the next focus is going to be: how can we get empathy at scale and use that as a part of personalization.
17:28 | Debjani:
That’s amazing, Nikhil. In a campaign-less world, empathy at scale. The world could do a lot with empathy these days, with everything that is going on.
17:38 | Nikhil:
17:40 | Debjani:
That’s awesome. I wish you all the best in your journey as you work on some of these very cutting edge things.
17:47 | Nikhil:
Certainly. Thank you so much, appreciate it.
Imagine a world without campaigns, where things are completely automated and no customer slips through the cracks.
– Nikhil Chandurkar, Product Management Leader