1 00:00:00,000 --> 00:00:05,520 Hey everyone, welcome to the Data AI Productivity and Business with a Little Personality podcast. 2 00:00:05,520 --> 00:00:10,960 I'm your host Mukundan Sankar and today I want to talk about using AI for customer lifetime value. 3 00:00:12,160 --> 00:00:15,840 So, you know, you may be wondering what is customer lifetime value? 4 00:00:17,120 --> 00:00:22,240 Well, to explain it in a more simpler and more relatable way, it's like figuring out, 5 00:00:22,800 --> 00:00:29,120 you know, which of your friends you can count on to show up for like a free food night, right? 6 00:00:29,120 --> 00:00:35,840 Like you got that friend who comes up once, he'll eat half the food and vanish until next 7 00:00:35,840 --> 00:00:42,560 time, you know, you have free food. And then there's the other friend who will be there every time 8 00:00:42,560 --> 00:00:49,360 and then they will pitch in for food, they'll pitch in for drinks and basically be a good friend. 9 00:00:51,280 --> 00:00:54,480 And that's what customer lifetime value is. Like it helps you focus on 10 00:00:54,480 --> 00:01:00,880 this other friend who is helping you with everything versus the friend who's just showing up for the 11 00:01:00,880 --> 00:01:09,440 free food. So that's the easiest way to explain this. But like specifically, if I want to talk 12 00:01:09,440 --> 00:01:16,720 about customer lifetime value in marketing, and that's the main focus of this, it is like helping 13 00:01:16,720 --> 00:01:22,160 you identify which customers bring you more value, right? And that's what the example I mentioned 14 00:01:22,160 --> 00:01:29,840 above was. Customer lifetime value is showing which customers will bring you lifetime value 15 00:01:30,560 --> 00:01:38,800 over a period of time, how much revenue they will bring to you, right? And the old ways of traditional, 16 00:01:38,800 --> 00:01:44,800 you know, customer lifetime value essentially were very, very outdated. The way we are doing it 17 00:01:44,800 --> 00:01:52,240 is it's not, you know, completely efficient. I feel like we can always optimize this. And 18 00:01:53,520 --> 00:02:00,400 that's why using AI, which we have available to us is the next step in this process. 19 00:02:02,320 --> 00:02:09,200 So what could be some, you know, challenges? Why am I even talking about this today? In the old school 20 00:02:09,200 --> 00:02:15,040 way for calculating, you know, customer lifetime value, it is like, you just think about it like 21 00:02:15,040 --> 00:02:22,560 using your flip phone in 2024, right? Today's day and age, you're using a flip phone. You can definitely 22 00:02:22,560 --> 00:02:29,280 use it, but you're not exactly going to be able to do anything with it in today's age, right? Like 23 00:02:29,280 --> 00:02:35,920 you won't be able to use, you know, maps, not saying Apple maps, but Google maps, because everybody 24 00:02:35,920 --> 00:02:43,040 uses that or Waze, right? For that matter. So you're typed in, let's say hello, and the group chat 25 00:02:43,040 --> 00:02:51,760 is not going to be as effective on a flip phone. So basically that's what a traditional model 26 00:02:52,400 --> 00:02:58,640 is indicating here. So what you need to be doing is like using what you have available to you, 27 00:02:58,640 --> 00:03:06,240 which is now AI in the last year and year and some change, use that to, you know, go to the next step. 28 00:03:07,920 --> 00:03:14,240 Again, with the traditional CLV of calculating CLV, which is customer lifetime value. It relies 29 00:03:14,240 --> 00:03:20,400 on like old static models, like spreadsheets. I'm sure you're probably doing the same. You're 30 00:03:20,400 --> 00:03:25,840 probably typing in these numbers of like how many customer orders have come through, 31 00:03:25,840 --> 00:03:31,200 how many orders this customer has placed over time, and you're calculating the revenue that this 32 00:03:31,200 --> 00:03:38,080 customer has given in that order, how frequently have they made a purchase, 33 00:03:39,600 --> 00:03:44,640 and how many months has it been since they first joined. So those are usually the parameters that 34 00:03:44,640 --> 00:03:51,440 you use for customer lifetime value. But here's the thing, right? It is just accounting for 35 00:03:51,440 --> 00:03:57,520 something really static. Like, I mean, if you want to account for something which is like a change in 36 00:03:57,520 --> 00:04:03,920 actual behavior, like see now the thing is if you're relying on old static rigid models using 37 00:04:03,920 --> 00:04:10,160 spreadsheets, you're not going to be able to capture changes, like sudden changes in the 38 00:04:10,160 --> 00:04:16,560 customer behavior. What I mean by sudden changes is like, like if customers start coming in three, 39 00:04:16,560 --> 00:04:24,640 four months, and you know, they've just been active for... So you want to be... When I talk about 40 00:04:24,640 --> 00:04:30,320 sudden changes in customer behavior, I'm specifically talking about these customers who come in, 41 00:04:31,760 --> 00:04:36,800 you know, once in six months, and then maybe come in three months continuously after that, 42 00:04:36,800 --> 00:04:42,080 and then again disappear after six months. So it's like a sudden change, right? Like, I mean, 43 00:04:42,080 --> 00:04:50,640 so that that kind of stuff is hard to capture in a spreadsheet. So you are missing out on real time 44 00:04:50,640 --> 00:05:00,400 insights, and that's something I'm hoping to tackle in this podcast, basically using AI, right? How AI 45 00:05:00,400 --> 00:05:08,160 can help solve this problem of, you know, not capturing real time data, which I'll talk about. 46 00:05:08,160 --> 00:05:17,440 So in using AI, what happens is you can do real time data processing, right? Now think about AI 47 00:05:17,440 --> 00:05:26,000 specifically as your friend who always knows when to ghost the group without the session. 48 00:05:26,800 --> 00:05:32,880 How would you think about AI? AI is like your friend who knows when you're about to ghost the 49 00:05:32,880 --> 00:05:38,320 group. So they'll do some kind of an activity to make sure that you're not ghosting the group. 50 00:05:38,320 --> 00:05:44,080 They'll maybe engage with you at the time, and that's what AI is. AI is your friend. So 51 00:05:45,120 --> 00:05:51,280 AI for customer lifetime value, what it does is it will update based on your interactions 52 00:05:52,160 --> 00:05:57,040 with the product or the company and like as a customer. So like let's say you're the customer 53 00:05:57,040 --> 00:06:02,480 and you're interacting with the company, and AI will update your interactions with the 54 00:06:02,480 --> 00:06:09,840 company, will update that model. So basically how the frequent interaction is, how many orders you 55 00:06:09,840 --> 00:06:16,800 placed. So all of that is being tracked with AI. So you're really getting super swift answers, 56 00:06:16,800 --> 00:06:23,760 right? Super quick. And I mean, you're seeing it real time. That's mainly the thing. And based on 57 00:06:23,760 --> 00:06:30,400 that real time interaction, you will now see targeted offers specifically for you. I don't 58 00:06:30,400 --> 00:06:36,560 think there's anything better than getting offers which is relevant to you as a customer. 59 00:06:37,440 --> 00:06:41,360 I mean, versus getting recommended some random stuff, which you're not probably going to buy, 60 00:06:41,360 --> 00:06:46,560 and it's a waste of your time, right? That's what I mean with AI, it can help. And the other piece I 61 00:06:46,560 --> 00:06:53,280 want to talk about how AI can help is with smarter customer segments. So what that means 62 00:06:53,280 --> 00:07:01,600 is like think of your AI power customer segments, which is like your Spotify playlist, right? 63 00:07:02,720 --> 00:07:11,520 Spotify is helping you get recommended to the songs which you like, right? Like I mean, 64 00:07:11,520 --> 00:07:17,200 it's obviously a recommendation engine. So they know which kind of music you would like. And 65 00:07:17,200 --> 00:07:22,880 that kind of recommendations is what we're aiming for. So now if you think about like more in a 66 00:07:22,880 --> 00:07:31,040 e-commerce base, so it knows which customers are in your big spending category or someone who's 67 00:07:31,040 --> 00:07:37,200 like in your, you know, they recycle their Ziploc bag. So there's two different kinds, essentially 68 00:07:37,200 --> 00:07:44,880 your one is a high budget customer and one is a moderate, I would say. But like that kind of 69 00:07:44,880 --> 00:07:53,360 segment is identified using AI. So an AI, it will go beyond demographics. And I mean, obviously 70 00:07:53,360 --> 00:08:03,360 demographics meaning their age, gender, location, and it will use behavior, the customer behavior 71 00:08:04,160 --> 00:08:11,840 for better targeting, right? And you want as a company, you should be looking at the customer 72 00:08:11,840 --> 00:08:17,760 behavior, but like AI would be helping you with that because doing it by yourself and looking at 73 00:08:17,760 --> 00:08:25,280 a spreadsheet, that's not the future. So AI obviously means it is looking at predictive capabilities 74 00:08:25,840 --> 00:08:30,880 because that's how it's learning about you. So it was predicting your next move. So yeah, 75 00:08:30,880 --> 00:08:37,520 you want somebody to like actually learn about you if you're comfortable with that, that is. 76 00:08:37,520 --> 00:08:42,800 So the traditional model won't allow you to do that. It will be giving you a weather report from 77 00:08:42,800 --> 00:08:47,680 last week. It's like a weatherman which is giving you weather report from last week. Think about that 78 00:08:47,680 --> 00:08:54,000 way, but AI would be giving you a future weather report for tomorrow. Like I mean, tomorrow's 79 00:08:54,000 --> 00:09:00,080 weather report would be predicted more accurately. And that's something you want to be aiming for. 80 00:09:01,200 --> 00:09:06,080 Basically you're forecasting what you're going to be doing tomorrow, what you're going to be wanting 81 00:09:06,080 --> 00:09:14,080 right now. It should know about you to be able to be able to give like a proper lifetime value 82 00:09:14,080 --> 00:09:21,040 prediction. Now, if I think about AI, right? Like, so think about when you're having like a quiet 83 00:09:21,040 --> 00:09:30,320 customer, like customer who is, you know, buying every few months and then is quiet for a few months, 84 00:09:30,320 --> 00:09:36,880 but they don't leave you any reviews, like a five-star review at least. So now with AI, 85 00:09:36,880 --> 00:09:44,240 what happens is like even your quiet customer, the customer who is, you know, buying three, 86 00:09:45,280 --> 00:09:50,960 buying a product like for three continuous months and then keeping quiet for six months and then 87 00:09:50,960 --> 00:09:55,120 buying after three months. So even for those customers who are quiet in between, 88 00:09:55,120 --> 00:09:59,520 what AI is doing is it is helping identify those customers in that segment. 89 00:10:00,560 --> 00:10:06,720 So nobody is getting, you know, missed because those could be a valuable customer and AI is 90 00:10:06,720 --> 00:10:12,400 helping you identify those. That's what we should be aiming for. Basically you're looking at 91 00:10:13,120 --> 00:10:18,800 all kinds of data that they provide, this customer is giving to you, like any kind of review 92 00:10:18,800 --> 00:10:26,720 information, if they're given social interactions. So that would be also helping you to identify the 93 00:10:26,720 --> 00:10:36,560 customer's value. So do make sure that you're, you know, utilizing those channels to leverage AI 94 00:10:37,440 --> 00:10:44,960 more correctly for customer lifetime value. Now, all this in theory might sound great, but it's 95 00:10:44,960 --> 00:10:50,240 like you would be wondering, has this been practically implemented? There is, yes, obviously 96 00:10:50,240 --> 00:10:56,000 there is a case that I want to talk about specifically today and this one was a Starbucks 97 00:10:56,000 --> 00:11:04,320 example. So Starbucks, what it does is, and I'll link this study in the show notes, but Starbucks, 98 00:11:04,320 --> 00:11:10,880 what it does is it will not just know your favorite drink, they will also ask you, you know, 99 00:11:10,880 --> 00:11:18,240 what's up, basically like just to see if you're interested in some kind of a drink that your 100 00:11:18,240 --> 00:11:26,480 barista made the last time. So because it has tracked your interactions, whenever you've gone 101 00:11:26,480 --> 00:11:31,280 in and made a purchase, so AI knows maybe, right? Like, you know, around that time to ping you 102 00:11:31,840 --> 00:11:36,880 on your, the Starbucks app and let you know that, Hey, do you want to get something today at 103 00:11:36,880 --> 00:11:43,760 because it has looked at your past interactions that you've gone in at 3 PM every Thursday 104 00:11:45,120 --> 00:11:50,400 to get your Starbucks pink drink and it would pink you, Hey, do you want to get a pink drink? 105 00:11:50,400 --> 00:11:59,120 So that kind of stuff is being tracked in the Starbucks app. So it is using AI to track the 106 00:11:59,120 --> 00:12:08,160 customer habits through their app loyalty program. So AI will basically flag the company 107 00:12:09,120 --> 00:12:15,040 if there's any changes in the behavior. So you're getting more personalized offers based on 108 00:12:15,040 --> 00:12:24,240 interactions and that's the world I would want to live in. Now, so like I wanted to also talk about 109 00:12:24,240 --> 00:12:30,480 how do we, if we wanted to do it ourselves, right? If you want to do this process ourselves, how do we 110 00:12:31,200 --> 00:12:38,880 go ahead and implement AI for customer lifetime value prediction? So think about it like this, 111 00:12:39,680 --> 00:12:45,600 getting started with AI in general is like cleaning up your room before your guests arrive. 112 00:12:46,560 --> 00:12:51,600 And what that means is your data needs to be clean, right? Like, I mean, AI needs to be clean, 113 00:12:51,600 --> 00:12:57,520 your data needs to be clean, right? Like, I mean, AI can only be as good as your data. So if your 114 00:12:57,520 --> 00:13:04,480 data is not clean and there's inconsistencies in the data and all that, AI is not going to work. 115 00:13:07,440 --> 00:13:13,840 You would, yeah, I know you want to basically give all access to the data, 116 00:13:13,840 --> 00:13:22,720 to the AI, but AI will definitely be more effective when you have cleaner data. So make sure data is 117 00:13:22,720 --> 00:13:29,040 organized, you're choosing the right tools, like the ones I can recommend, I've seen, you know, 118 00:13:29,040 --> 00:13:35,760 being used effectively, well, Google AutoML or Salesforce Einstein. So they have very 119 00:13:35,760 --> 00:13:43,680 user-friendly AI integration. So make sure you're looking into that. And I would say start like 120 00:13:43,680 --> 00:13:49,760 gradually, right? Like, I mean, start with a small focus group before scaling up with this AI process. 121 00:13:50,640 --> 00:13:58,160 So sometimes AI may not be for you yet, but I would always suggest to start small, 122 00:13:58,160 --> 00:14:03,680 definitely take your time. But like I would say, do look into this process of using AI for your 123 00:14:03,680 --> 00:14:09,680 customer lifetime value. So that's the key takeaway really. AI driven customer lifetime value 124 00:14:09,680 --> 00:14:16,880 prediction is essential if you want to progress ahead in a more modern marketing world. 125 00:14:18,000 --> 00:14:23,680 So you can get more personalized campaigns, reduce your customer churn and build your 126 00:14:23,680 --> 00:14:30,000 personalized campaigns, reduce your customer churn and build solid customer relationships 127 00:14:31,120 --> 00:14:37,120 that last you maybe in your whole company lifespan. So that's something you want to be aiming for. 128 00:14:38,240 --> 00:14:43,840 And just think about why, and I'm sure you know this, why AI is the future. 129 00:14:46,720 --> 00:14:51,840 AI is everywhere around us. And if we are not adapting it, we're definitely lagging behind. 130 00:14:51,840 --> 00:14:57,760 So I keep thinking about areas where we can use AI and customer lifetime value 131 00:14:58,800 --> 00:15:03,200 is definitely a strong area to do that. So make sure you're using AI and 132 00:15:03,200 --> 00:15:20,720 that's all for this podcast. Thank you and I will see you in the next episode.