The power of personalisation and how to deliver at scale - Mark Abraham, BCG

27:25So take a beauty retailer, for example, we worked with. They found there were less engaged customers that were shopping across a bunch of brands. They were really only coming in for the holidays. And you could see this in the customer data they had from just their store transactional data. Next, we found a group of customers that were coming in more regularly throughout the year. They were buying less on promotion. They were shopping one category. Then you found people that were shopping multiple categories, not just hair, but lipstick and other products. Next, you found customers that were members of the loyalty program and had the app downloaded and visited with their friends.

28:09And each of these customer segments basically formed an engagement ladder when you put it all together. So as you stepped up the ladder, you had lower churn rates, higher loyalty, higher spend. And that's a great example of how companies can mine their customer data to come up with where can personalization really help nudge customers to higher brand love, but also higher revenues for the company. And I think those are the ways that personalization can really help. Yeah. One of the challenges I imagine, you've got this slightly amusing example in your book of where different data sources are integrated and they couldn't even work out the gender, could they?

28:51which is quite shocking but how actually share that example is quite a funny example yeah so you go back to my where i started with the women's yoga pants i bought my sister uh that holiday gift and then i get tagged the wrong gender and in that case it was the first party data but i've seen clients use third-party appends and we've tested that for a specific client where we used one of the common data providers for demographic data. We matched it up against the company's internal first-party data and what customers self-reported. It was about a 60 % accuracy on gender. So basically, most of the time - Not much better than 50-50, right?

29:36Exactly. You could have just done a coin toss and it would have been almost as good. And this is data that the company was buying. So really ensuring that you're buying the right data, you're testing it for accuracy is so critical. And asking yourself, if it's truly that important for the experience, how do you build it in so that customers give you what I call zero-party data? So you're asking the customer that question and the flow of the experience. You know, again, from an apparel perspective, you have Stitch Fix, for example, that sends customers boxes of apparel and they start with a style shuffle, they call it.

30:18So instead of trying to infer does more like trendy clothes or more traditional style clothes, they'll actually ask you in an interface to kind of rate things you like and don't like. And based on that, determine your style preferences. What I love about that is that's actually making it fun, gathering data. If you make gathering data fun, and then you can make it useful, you get far more likely to get people to adopt. That's right. And I think too often companies think, well, I'll create a preference center over here where customers can go and fill in things. Well, if it's not part of the flow, you find like 1 % or 2 % adoption rates, and then you don't really have enough data to do anything with.

31:02So how do you make it part of that flow and make it clear also how is this going to make my experience better. So in that case, you know that your box that's going to come is going to be informed by the style choices you've just made. One of the online services I sign up to is Buttonup Box here in the UK, which is like a home delivery pet food, basically. First question I ask you is, what's your dog's name? So immediately they're touching on something deeply personal and when's the dog's birthday and things like that. So immediately you've got some data that's going to be kind of personal. And in that context, that's a great question to ask.

31:38And as a customer, I'd want to give that. I think where brands sometimes get into trouble is if they don't think through the experience. For example, if you went into a store and the shop assistant you've never met suddenly greeted you by name, that would be pretty weird. A bit awkward. Yeah, exactly. That's weird. Now, with all the data you're gathering together, what kind of intelligence are you, what's the most useful kind of intelligence you're gathering to help you kind of grow and make better decisions? Well, we now live in a world where we can't rely as much on third-party cookies and third-party data in general.

32:11And so really, in addition to the zero-party data that we just talked about, first-party data has become really critical. and most companies have a wealth of transactional data so obviously they need to fully leverage and mine that and connect it in the right way where I think companies under leverage data is the engagement data so everyday customers even in infrequent categories they might be giving digital signals they might open the app, they might click on things they might not click on things all of that is useful information and it's frankly more data than most companies know what to do with or can even harness and tap.

32:55So really thinking through which of those data signals need to be part of the models, building features on that data set and then ingesting it into the models is critical. But again, the key is to be clear, what are you trying to do for the customer? What data would I need to know and how much of that do I actually have? A good example I remember reading about was Woolworths, the Australian retailer, because I think, as you point out, anyone who's worked in retail will know you operate on crazily tight margins. So how you use your data is the difference between being profitable and not kind of thing.

33:27But they've invested heavily, haven't they, in how they use their loyalty card data? It's a massive area we're seeing with grocers around the world. Personalized offers is an amazing tool for getting better returns from the billions and billions that are spent on promotions, mostly in a mass way today. In fact, in most markets around the world, only a couple of percentage points of that promotional spend is personalized, and most of it is spent on mass events. So what Woolworths has done, and frankly, many leading grocers around the world are doing right now, is really building personalized offers at scale, both for their own promotions.

34:12So now they can have for any given product at the SKU level or brand level or at the store level, they can set how much of the discount do they give to this customer in this week. They've even got challenges where they make it fun. They ask you to buy a few things or come a few times or a period of time, and then you get more points. And points is a great way to engage customers in many markets around the world as well because people just value points more and they want to see that balance build. One of the key reasons people go to that app in the case of Woolworths is to check their points balance.

34:51So they've gamified that experience. And then the other cool thing they've done is they've engaged suppliers in this. So to your question on if I'm selling sodas or fizzy drinks, actually a great use case for a personalization is tapping into this kind of capability with a retailer like Woolworths. If you can spend your promotional dollars targeted the right customers that are going to either discover you as a brand, switch maybe from another brand, or buy you on occasions they might not have before, those are great opportunities for you to get in fact 3x the returns we found in all our work versus just spending it in a mass way i had a similar experience actually client side with tesco in the uk um had a bit of a challenge on my hands i was overseeing uh luco's aid and we were going through a big reformation sugar reformulation and we ended up with a big backlash and uh what was interesting is my uh club card data tesco club card data was immediately reflective of the problem we had, my brand tracking data was six months behind.

36:00It was really fascinating. So I could see in the Club Card data, because it updates daily, I could literally see people not responding to the normal promotion, where they were, who they were, what their buying habits, what they're also buying. So when it came to do something about our issue of having lost a big chunk of our customer base, we could be super targeted in who we're targeting, where they are, what kind of offers would get them back. and I found that the club card data, back to your speed point, we could operate so quickly and we could adapt our kind of promotion to suit it compared to the traditional brand tracker.

36:32I think it was about nine months before it even suggested we had a problem. Not only are you seeing that data faster, but I think what's changed even in the last 12 to 14 months with these retailers is we've seen them scale the content and automate it. So now big grocers can launch hundreds of thousands of variants of these promotional vehicles and actually change that every week to who they're targeting based on the data. And that's where you could never do that with a mass promotion that you've printed on all the labels and the signs in the store. Well, this is going to be my question because before that or back in the day, you'd often be at the till and then it would print out a coupon.

37:18Right. At CVS in the US, it actually prints this mile long receipt with all your coupons. The crazy thing about it is I've never been more excited to have 10p off carrots. You know what I mean? It didn't matter what the offer was. The physical experience of being handed this long strip of multiple offers on random things. Anyway, a bit of nostalgia, but sometimes the way it's delivered, it doesn't matter how much discount it is, just the theatre in which you deliver it can be important. But you touched on a really, really good point, because if we set up organisations to allow us to change our message to every consumer on every occasion, in every moment or whatever, suddenly the marketing department has got a challenge on their hands, because suddenly the campaign they created, their single campaign, has now got to be adopted, you know, thousands, maybe even millions of times.

38:08so how do you kind of connect those two things together without kind of losing the quality of what you're doing because i guess the danger is you end up dumbing down to lowest common denominator just changing that changing the kind of headline but how do you kind of solve for that problem of now we need to communicate lots more messages to lots more people and still with the kind of engagement levels that we might have done before so this is one of the first problems we solved when i launched our personalization team uh starbucks at the time had their Frappuccino happy hour offer, right? In May, we always talk about Frappuccinos to everyone.

38:43And I think that's emblematic of most organizations. Category teams will try to push their products. Well, only about 30 % of Starbucks customers actually ever want to buy a Frappuccino. And so you're really sending an irrelevant message at best to the other 70%. And this is where if you can create a flexible, templated, dynamic message that actually has space in it for every product, in the case of Starbucks, there's 87 ,000 plus products you could talk about, maybe with an offer or maybe it's just a recommendation. you multiply that out you could have hundreds of thousands of variants of dynamic messages going out to customers.

39:34And it's actually, I mean this has been possible for a decade now to do this with technology. With Gen.AI it's even easier because you can even have the images varied and the persona of the customer come into the email you create or the ad you create. And that's what companies are doing. And again, with that approach, we have seen triple the ROI on promotional investments, ad investments in this area. It makes sense because actually my local barista knows that there is no temperature that's so cold, but I won't also have a frappuccino. So honestly, if you're appetizing frappuccinos in May, I was buying them in the middle of winter, and it was minus 10.

40:14So I'm living proof that that's the case. Yeah. And actually, it's interesting. The whole coffee market has moved towards cold and more indulgent beverages in that regard. So part of it is our ability to discover all these beverages. It's like the macro versus micro, isn't it? At the micro level, of course, Frappuccinos peak in summer, of course. But then there'll be a whole bunch of people like me that are very happy with Frappuccinos in the winter. Yes, and young people now order cold drinks throughout the year on average. No, exactly. But then by individual, it could be different. It's good to say I'm on trend then.

40:47Exactly, there you go. But also actually in my day job at System One, there's some good evidence for how this works as well. I mean, nowhere near, it's a complexity you're talking about, but we did this project for a regional airline in the UK called Jet2. And we measured the emotion that people felt looking at something that was generic, so targeted everybody, something that was regionally targeted, regionally and personally targeted. So imagine, I mean, in fact, they've got a really good campaign. It's kind of a couple going on holiday. It's actually filmed from the first person. So you see the girl dragging the boy's arm into the pool, you know, to the restaurant, down the shops, you know.

41:23So it's her perspective looking at him kind of, you know. So it's beautifully done. It's got a great soundtrack. And then these are round numbers. So I won't give the precise numbers, but we've got a five-point scale. And they got three-star, which is pretty good for the generic ad. When they then, I think they then went regional and said, oh, now flying from Leeds Bradford Airport and then served it in Leeds, right? went to four star and then when they targeted young couples because they had in fact they had loads of different versions of this there was an older couple there's a family there's the you know they had lots of different version execution so when you then married the audience who saw themselves in it the airport they lived near to and and the time they want to go on holiday which actually is is the winter so advertise holidays in the winter because that's what everyone's thinking you know i don't want to get away from here all those things took it from three star to five star.

42:13So, you know, there's a power in terms of, you know, making a correct emotional response, the memory that comes from getting this right. That's so funny, because we did the exact same thing with Asia Pacific airline, and we saw 20 % increase in conversion from those exact things, you know, where you were flying to, and what type of persona, where your weekend warrior family and the background, etc. It makes a huge difference. Yeah, it really does. It really does. Now, And let's talk about some of the barriers because, you know, as I read through your book, I mean, the business case is very strong.

42:47Of course, we want to personalize, but there are lots of reasons why we don't get this right, don't we? So let's break down some of the reasons. I think the first one that you talked about in the book, which I think is enormously important for marketers out there, is the fact that businesses are siloed in terms of department. Because as you've talked about the solutions, the solutions involve more than marketing. So if I'm talking about my crazy development or if I'm talking about my media buy, that's on me as the CMO, right? I own that and what I say can be implemented straight away. You're talking about the customer experience.

43:16Now that touches so much more of the organization. So how would a CMO get through that kind of having to align the whole organization behind this? Yeah, I think that's the critical bit is that personalization is a new way to compete. And so that's not just on the CMO. It's really a CEO-level agenda, and it impacts the entire C-suite. So in the book, we go through the role of the C-suite, and every single leader has a role in this. I mean, just think of the CFO. They're underwriting the business case, and they need to be able to stage-gate this and understand should they release that next tranche of investment, because we're typically talking millions of dollars in people, technology, data costs.

43:59But then you've got the other parts of the organization too. The chief data and analytics officer has to make sure the data is flowing right. There's data governance. There's a common definition of simple terms like who is a lapsed customer? So many clients they have, they have five definitions and five teams going about it in different ways. You've got a disjointed experience. You've got the CIO. oh, there's always a shortage of technology resources, and so much of this does depend on technology at the end of the day. So how are they aligned with the roadmap and the vision and allocating the resources against this?

44:38Those are just a few. But again, in my experience with System 1, actually, there's a magic between the person in charge of customer or CMO and the product development team. so the people that can make it and understand what's possible and how to use the data and how it can all turn into a product and the people that understand the needs of the customer. I mean, you know, other departments have a really important role and need sponsorship, but I found those two. It's almost you need to create an alliance with your CIO and put those two things together because often they're, you know, one's a supply and one's a demand and those two don't meet.

45:14But I think a lot of this is getting those two departments together, isn't it? It's so true. True. And I think it plays out differently in different industries, which is why some companies have made more strides than others historically. And digital native companies, typically you'll have a digital product organization that can build stuff fast and launches things fast and has that nimbleness built into it. But then they might not think about the marketing or communication aspects of it. So how do you get those teams partnering? In a large bank, typically every channel is put in a different part of their organization, and then you've got the matrix of different categories and products against that.

45:55So how do you bring that together? In a retail example, you might have the CMO, again, controlling the customer, but then the digital experience like the website or the app sitting in a separate team. And then you might have the core systems even sitting in separate teams. So again, how do you bring that together? I think consistently the answer to me is develop a clear roadmap, align the senior leaders, and then break it down into small use cases. This is why we go back to Empower Me. How are you going to help the customer in the next six months in a scalable, demonstrable way that actually drives impact for the customer and the business?

46:38And then align that team against making progress. And I think almost in every case study you talk about in the book, there is this, we set this vision out, we want to do this, but then there's this test and learn phase, isn't it, where they're kind of experimenting on one thing, learning about it, and that builds confidence in the business then to kind of roll it out, roll it out and scale it up. Yeah, for two reasons. One is oftentimes you need to build the case for change within the organization. Just going back to my airline example that I gave earlier, we had to change the way all these 50 product groups were communicating to the customer.

47:11That was massively risky. I mean, you maybe some product groups actually could communicate less and might be worse off, but even if the enterprise was better off as a whole. So by having this personalization lab approach, we called it in that case, we could test and learn what the impact would be for 5 % of the customer base with this new approach where it was much more next best action and leveraging a common content library. and then go back to the organization to say, this is the impact, this is how we need to adjust our targets by product. And by the way, we learned a lot about how we need to re-engineer the content production process, as an example, to enable this kind of approach.

47:57So when we had to drive the change with hundreds of people instead of a dozen people, we knew what the changes were. Yeah. Now we often think that the block is technology, Whereas actually you had this quite neat 70-30-10 rule. No, 70-20-10. Sorry, get my maths right. 110 % would be quite impressive. But yeah, the 70-20-10 rule, which actually puts the spotlight on it. It's often people in process, isn't it? That that's the bit you've got to fix first. Yeah, at BCG, we've done a lot of AI transformations and this is the rule. It's 70 % of the time, these transformations fail because of people.

48:3220 % technology, 10 % data. So it doesn't mean data and technology are simple. We'll get to that and there's a lot to get right there. But it does underscore just how complex the people changes and upskilling the people, getting them bought into a change, re-engineering these processes for speed. Now I think probably the other thing that people sometimes worry about in this space is the whole risk, data risk, compliance, data quality. There's a whole world of, I mean, I know I've been in a few deals with some of our larger customers. And it'd be months of legal negotiation about who owns the data and like, you know, when, you know, consumer data is transferred between one organization to another, what does that mean?

49:16And who can see it and who can't see it sort of thing. There's quite a lot to consider, isn't there, in terms of both data quality, but also the risk and compliance to the law and so on. Yeah. So, I mean, I deal with that all the time in terms of the data access. I think the key there is, obviously, there's the regulations that vary by geography, and they're constantly evolving. I think what the smart brands are doing is giving customers transparency. Marriott is the example we give in the book, for example, where they make it easy for you to go in your Bonvoy app and see what data they have about you.

49:52Every customer has a right to be forgotten under GDPR and many other jurisdictions. but actually giving people an easy place to see what even they have about you is a great way to build that trust with customers. LinkedIn is another example of a company that actually you can go in and download all your data that LinkedIn has about you and easily transfer that and leverage that as a customer. Another way for them to build that trust with customers. I think where it's evolving even more though is Gen.ai. I see a lot of companies taking different approaches in terms of how open they are to innovation around Gen.ai and the security around that, the risk management of that.

50:39I think the worst thing you could do is, like some of my clients have done, implement a blanket ban on these tools. I think companies have to force themselves to find ways to safely experiment with the tools because the truth is the skills of their own workforce and how relevant they are in the marketplace will be driven by their ability to leverage these tools and i'm seeing a huge bifurcation in that already that's really interesting because i've spoken to a lot of the world's biggest tech companies all of whom want to lead on ai most of whom ban the open source ai platforms as well so there's a conflict coming there presumably isn't there in terms of who's going to win out and who's going to, I suppose they're all playing a game of, you know, they're backing their horse, aren't they, to be the winner.

51:29But presumably only one or two platforms will end up being the dominant in the future. Yeah, there's definitely a network effect here. I think fundamentally, though, here, there is a piece on the LLMs can be leveraged for understanding English language or whatever language requests. But then the source of that data needs to be the company's proprietary data in many cases and when you get to the specific use cases. So if you're wanting your customers to discover products for your beauty needs or travel needs, well what is the source of that inventory of products, experiences that you're serving up to them?

52:11that should come from proprietary content and data systems that the companies have. So it's architecting those in the right way where you can resolve some of these concerns about security, safety, etc. Yeah, but back to your previous point, the opportunity might be then collaborating with other systems as well. And then also where the future may go to, which is more open source or people creating their own personalized experience and you need to be part of it rather than pulling everyone onto your own. Yeah, and that's where I think just working with companies, for example, in the entertainment domain, there's very much clear concerns from artists and others creating content around how is Gen AI going to be used to infringe on their copyright and all the work they've done into their products.

53:03So some of these companies are taking very conservative approaches to these tools. I think how you get around that is to say, well, you can use your LLMs to the large language models to understand the requests. You know, here's what I'm asking the tool. But then, again, query the actual copyrighted content and database underneath. So, final question. You've written a book on this. You've had over 100 customers, I believe, kind of, as BCG has partnered with over 100 customers, deliver personalization programs globally. What would be your advice to somebody that's listening to this thinking, I need to do this better.

53:44I need to lean into this. What should they be thinking about? What should they do? Obviously, read the book and give you a clue. That's taken us ready. Yes, but also ask yourself the questions in the book. We tried to make it really practical. And in each chapter that goes through the five promises, we arm you with questions to ask yourself and ask your team. Because most companies I find say, we're already doing personalization. But when you ask yourself those questions, I think you might find maybe there are some things to do better. And that's what the personalization index results suggest as well.

54:21I think the most critical ones are, do you have a clear roadmap? And is it broken down into these six-month chunks or more often where you can show real business value as well as customer value to fund the investments? and then what are you doing next and what are you doing now? There's no shortage in any client that I go to of ideas for how to do more personalization. Typically we'll get a cross-functional team in a room and within a couple of hours we'll have the entire wall full of stickies with use cases and ideas. But which of these are truly going to change the customer experience? Where is personalization really critical in that journey for the customer?

55:07And which of these will drive material business results so you can fund the journey? Those are the questions we ask. And I would say pick those few big wins and get started on them now. Once you build the momentum, and most companies are already on this journey, but some of them have lost a little bit of the momentum because maybe those first use cases just weren't big enough material enough or they didn't make progress on them fast enough so i think there is to your where you started this whole discussion some ceos i talked to are jaded with well i heard about this personalization effort the last five years and we've spent all this money but what have i gotten for it i think that's where again having that clear owner uh who's holding the organizational accountable can be a game changer yeah well it's a bit like the barcode 10 years ago and sorry not the barcode, the QR code 10 years ago.

56:02Everyone was saying it's going to be the big thing. We then went off it and then pandemic hit and suddenly everyone's using it sort of thing. So there's definitely a timing element, isn't there? And I think the same risk exists with Gen AI. If we only focus on using AI to drive efficiencies and cost savings, I think we will tire of it very quickly. So how do we use it to drive growth, to improve the customer experience? and personalization is the way. There you go. Well, couldn't put it better myself. So what a great place to end. Mark, thank you so much. Really appreciate it. Thanks for your book, of course.

56:36And thanks for sharing your wisdom on the podcast. Thanks for having me and excited to share it with the world. By the way, it's also bright yellow, which is a color that I endorse heavily. Yes, I love walking around. We just launched in all the Hudson's bookstores across the airports in the US. So wherever I fly around the country, it's fun to see it. It's like a big highlighter pen through the bookstore. So congratulations. My friends are sending me texts with it. So it's super fun. Brilliant. Well, enjoy. Take care. Thank you very much for listening or watching Uncensored CMO. I hope you enjoyed that.

57:06If you did, please do hit the subscribe button wherever you get your podcast. If you're watching, hit subscribe there as well. I'd also love to get a review. Reviews make a big difference on other people discovering the show. So please do leave a review wherever you get your podcast. If you want to contact me, you can do. I'm over on x at Uncensored CMO or on LinkedIn, where I'm under my own name. John Evans. Thanks for listening and watching, I'll see you next time.