Personalization at Scale: An Interview with Mark Abraham, Global Leader, Boston Consulting Group

Mark Abraham is Boston Consulting Group’s Marketing, Sales and Pricing Practice Leader for North America and the co-author of the book “Personalized: Customer Strategy in the Age of AI”.

This interview has been edited for length and clarity.

Right offer – right time – right message.

Tough for most marketers to get right, never mind in real-time. Yet marketers have drooled about that possibility ever since Don Peppers and Martha Rogers first popularized the idea of one-to-one marketing way back in the early 1990s. But until very recently, personalization at scale has been largely wishful thinking. Because, to pull it off, marketers need the right data – right technology – right processes. And that trifecta is rarely found outside of direct-to-consumer companies whose business model is predicated on getting it right.

The cost and complexity of putting in place the required technology infrastructure has always stood in the way of real-time personalization becoming feasible for most businesses. Any attempts at doing so by intrepid marketers is usually foiled by the Sisyphean effort involved. The trade-off just doesn’t seem worth it. And the business case is hard to get past the CFO who scoffs at the rosy sales projections. Which is why most forms of personalization today are limited to “next best offer” messaging, the use of dynamic variables in e-mail, or the particular mix of images and text you might see on a web landing page – what’s called “performative” personalization, as in “Stephen, we handpicked this offer just for you”.

Most marketers are just fine with the fluffy use of personalization, content to keep blasting out campaigns with minimal segment-based variation, because to do otherwise would slow the whole process down to a crawl. And marketers are always in a hurry to push out the next social media post, e-mail or targeted online ad. Complexity is their constant nemesis. It gets in the way of execution. It means more time spent planning, testing, measuring. It puts enormous pressure on content production. And it demands a lot of back-end data engineering to stitch together a unified customer profile. Nothing is ever push button. That’s why campaigns can take weeks – sometimes months – to launch.

Until now, that is. Because soon – maybe even very soon – marketers will be able to step back and hand over the grunt work to personalization engines.

Artificial intelligence is already beginning to revolutionize how marketing creates, develops and produces one-to-one communications. Marketers everywhere are finally able to individualize content, images, offers, web pages and video using AI platforms. GenAI tools can write the copy, design the layout, create the imagery, generate multiple versions of those assets for testing, and adapt it to each channel, taking into account what works best based on past performance, clicks, buying habits and known preferences. And as almost everyone has seen by now, the creative output can be eerily human.

If all of that sounds too good to be true, there is another major inflection point just ahead: the use of agentic AI to autonomously run end-to-end marketing campaigns, all without manual intervention. This will cause marketing talent and resources to invariably shift from producing stuff – even thinking about what to produce – to overseeing its creation and deployment. But here is where it gets even more intriguing: AI agents will automatically guide customers through every step of their online journey – interacting with them in real-time, advising, selling, servicing, prodding, reminding, coaxing, stroking – again without the overhead of marketing involvement. Each interaction will pick up from where the last one left off. The customer experience will be completely streamlined, perhaps to the relief of time-starved marketers who can never seem to give it the attention it deserves.

But doing personalization at scale – even with the help of AI – is still a daunting challenge, as Mark Abraham, head of BCG’s personalisation practice, is quick to acknowledge, pointing out that only about 10% of companies today are true personalization leaders. In the book “Personalized: Customer Strategy in the Age of AI”, he and his co-author David Edelman lay out a framework for delivering impactful personalization that enhances the customer experience, arguing that it should be a strategic priority for every company and a source of competitive advantage.

Stephen Shaw (SS): What led you to start up a personalization practice at BCG?

Mark Abraham (MA): In the 2010s we were doing all this great digital strategy work for companies who then couldn’t scale a lot of the pilot projects we had started. At the time we had brought together this awesome talent – human centered designers, martech experts, data scientists, data engineers – and one of the first big builds we did was Starbucks’ personalization engine. They were one of the first movers among the non-digital native companies. But since then, and especially after COVID, our other clients have recognized the importance of scaling personalized experiences.

SS: Are you still having to sell companies on your personalization vision?

MA: Ten years ago when we started there was a lot of work convincing companies on the importance of that. But it’s been incredible how in the last five years that has shifted. I took over the leadership of our marketing and sales practice right at the start of COVID, so March 2020. One month we were worried nobody was going to have any work and then the next month the phones were ringing off the hook. Literally everyone recognized how critical building a digital AI-driven capability was. Our own team doubled in the span of 18 months. Today I think the imperative to do it is no longer a question.

SS: What does personalization at scale really mean?

MA: Personalization at scale really means you learn something about a customer in every single interaction you have with them. And you are using that insight and knowledge to make their next interaction better, faster, cheaper, more convenient, and doing that not just for one customer, but if you’re a large brand, across millions of customers and billions of interactions. It’s about thinking how to empower the customer. So I love that we’re on the Customer First Thinking podcast, because that’s what this is all about at the end of the day.

SS: You say the main idea in the book can be distilled into one equation: volume of interactions multiplied by speed to the power of two. How did you arrive at that particular formula? And why is speed so important in the equation?

MA: Yes, so let’s unpack that a bit. It’s a simple distillation of a ton of research we’ve done, benchmarking hundreds of companies, using what we call our “personalization index”. What we found was, number one, if you want to do personalization at scale, you do need a significant number of interactions that you’re capturing digitally. And so that’s where the N in the formula comes from: How are you engaging with the customer so that you actually have more interactions than you have transactions?

But the speed component is the most important and that’s why we underscore it in the equation. Most companies have a campaign mindset. They’re pushing out communications to customers that can often take three, four months, let alone measuring it and trying to improve it. In one bank we work with, for example, maybe 80 people can touch a campaign across the agency, internal marketing, and the tech folks. There might be 15 handoffs or more. But the personalization leaders – the top 10% of companies – are maniacally focused on shrinking that time. That’s why the digital natives – companies like Spotify, Netflix, Uber – operate with a mindset of running thousands of campaigns and experiments, learning from each of them, and optimizing the experience as a result.

SS: So the essence of your formula is the compression of time to accelerate learning and improve the experience faster?

MA: Yes. Before it was more about predicting what the customer might want. There was tremendous emphasis on the data science and the AI used to predict what kinds of marketing messages customers were likely to respond to.

SS: You referenced your “personalization index” – can you describe the elements of that score?

MA: Ultimately the personalization leaders deliver on five promises to the customer that they make. Number one is “Empower me”. How are you using personalization to help me? If I’m Spotify, I’m trying to help you get access to the music that you love. What is the outcome for the customer? Are you making it better, faster, cheaper or more convenient? So that’s where it starts. But in order to do that at scale, you also need to have all the enablers for that. And that’s where the other four promises come into play.

So, the second one is “Know Me”. You need to actually have customer data. And again, back to the Spotify example, they have not just my profile as a customer when I log into the app, but all the songs I’ve listened to in the past, what genres they are, all the metadata about those songs, how long I listened to them, which ones I skipped over, who my friends are that I’m connected to on the platform, et cetera. That is the key foundation on which personalization is built. Spo there’s a lot to think about from a data architecture standpoint.

The third piece then is “Reach Me”. So once you have all that data, you’ve got to know how to make it easy for customers to access that experience. In the case of Spotify, they have this amazing music library. They use AI to curate playlists that are incredibly personalized for the listener.

Next comes “Show Me”. You’ve got to have the right content. And so obviously in the Spotify example, that’s not just their catalog, it’s also new songs and podcasts and so on. Or even notifying customers when their favourite artist is coming to town so they can get tickets and shop for merch.

And lastly, and this is the most important enabler actually, because it comes back to this point around speed, “Delight Me”. Are you making the experience better every time I come back to you? In all of our consumer research work, we found that 90% of customers are willing to share their data when they have a better experience. So there’s tremendous appetite for personalization, even with all the concerns around data privacy, as long as companies deliver on these five promises. So that’s how the index works. And we score companies from a simple 0 to 100. We find that only 10% of them score above a 75 or so, making them a personalization leader. The average company only scores a 49. So there’s a lot more work to do on personalization, which is one reason I wrote this book.

SS: Who are some of the companies you’ve deemed to be personalization leaders?

MA: The digital natives, not surprisingly, still score the highest on the index. Uber and Spotify are doing this really, really well. I love what Spotify has done with “Daylists”(1) because the name and tagging that’s personalized to you is so iconic that it’s become a way for Gen Z and Gen Alpha to interact with each other and get to know each other quickly. On the brick and mortar side, there’s a lot of interesting companies, whether it’s Fidelity or Voya in financial services, or the likes of Woolworths, Tesco and Home Depot.

SS: What are the main reasons that companies are struggling with the transition to personalization at scale?

MA: About 70% of digital transformations fail mainly because of people. So I would start with fixing the operating model. Especially for personalization. It is the most cross functional initiative any company can undertake. It requires customer first thinking. So, the marketing team, the insights team, have to be at the table. But in order to actually deliver at scale, you need analytics, data, engineering, tech, and digital experience to come together. And you need to identify key use cases that will deliver value to the business. Because there are real investments that have to be made to make a meaningful difference for the customer. So that’s really the art of getting it right.

SS: You say that personalization at scale comes down to content. Is content the biggest choke point right now?

MA: Initially the big bottleneck was the data and what I would call predictive AI. So getting the algorithms right and fine tuned and making sure the data is set up and clean, and I would say very, very few organizations have that fully right. It’s a constant evolution. But once you get that to an okay place, content becomes the bottleneck.

What’s interesting in the last couple of years though, it is way easier to create content with GenAI than it has ever been and so we’re in the midst of a content explosion. So you can create not just static images or copy, but even video and human characters, all with GenAI tools, in a matter of minutes and hours, whereas that used to take really expensive photo shoots and many months of planning. And it also enables you to take one campaign, let’s say an ad for a product, and create 100 versions of it. So you can do things like a global launch across 50 markets aimed at five different personas. What’s exciting is you can set up these content libraries that are geared towards your target customers with GenAI and pair the predictive AI with that to serve up the right content at the right time.

Jesper, Dall-e, et cetera are great tools for images. When it comes to video, Google’s VO2 just launched. But then you have other ones like Flare.AI, Flickr, etc. that solve different elements like motion. Adobe’s Firefly allows you to take copyright approved content for things like images of humans and characters. So there’s a plethora of tools and different ones are good for different tasks in the content creation space. There need to be teams internally that stay on top of the evolution of those tools – that have the knowledge of how to mix and match them. But it does take a partnership across clients and agencies to leverage these tools fully, thinking about where they can take cost out and invest back in more content creation.

SS: I do want to touch on one other challenge that companies are wrestling with and that is finding more agile ways of working. You say that’s the hardest thing to get right. And to go back to your campaign mindset reference, is that because most companies just aren’t set up that way?

MA: So a great example of this is a travel and airline client that had 50 different products. They were trying to sell the customer everything from seats to ancillary flight products, like an upgrade to a credit card, travel insurance, and other products. And each of those products was set up as a separate P&L with an owner who was responsible for hitting quarterly targets. So of course they were creating great content and pushing that out to customers through email and on the website and the app in order to hit their targets. And they were even using predictive AI to figure out which customers would be interested in which products.

The problem, of course, was that the highest value customers were actually getting hit with the most product offers while the least engaged customers were not getting anything. So some customers were getting 10 emails a week and a lot of outreach through the app and others were getting nothing. When you flip that on its head and actually ask, what’s the best for the customer, you get to a very different approach. But in order to make that change we had to set up a personalization lab, assemble a team, carve out a few hundred thousand customers and show the organization that this is what can happen when you take a customer first approach. We proved that we could reduce the number of emails we send by 30%. We proved we can drive 10% more cross sell revenue. But we said that we needed a team that’s actually going to think customer first and act as air traffic control and build the technology platform in a way where it can operate seamlessly. So that was a huge change. It required rewiring their organization, their incentives, as well as doing integration of the backend technology.

SS: What are the main elements of the technology infrastructure to do personalization right?

MA: Number one is the data. You have to have accurate data. So you need a data platform, and you need to understand what you may be missing based on your use cases. Back to the Starbucks example. Like many retailers they have many different product hierarchies. And so in order to get product recommendations right we had to really understand what is the right product hierarchy from a customer standpoint. So again, start with the use case and line up the data accordingly.

Next, there is a whole set of components that have to come together from a martech standpoint: the tools to be able to set up lots of experiments. There’s also the actual delivery of that experience into given channels. There are a lot more tools that are available now whereas in the past you had to custom build. But there is always some customization required to make sure the solution is tailored to your business, versus just relying on out of the box models.

SS: Let’s move on to organizational structure. In the past personalization has been first and foremost a marketing tool. But importantly it’s also a way of enhancing the customer experience. So who should have point on transformation?

MA: It’s one of the biggest issues in organizations. I think ultimately personalization rises to the level of a CEO. It should be part of corporate strategy for a company. AI with personalization can be used to drive growth. In fact, personalization leaders grow 10 points faster than laggards. And so, at the end of the day, the CEO should have it on their agenda.

Operationally you’ve got to think about alignment around a shared roadmap and vision. Where are we going over the next three to five years? And what are the use cases that we’re going to tackle with what payback and investment? Someone needs to own that roadmap and hold the organization accountable even though they’re not going to control all the resources, because some of it will sit in tech, some of it will sit in marketing, some of it will sit in analytics. That leader typically is at least a VP – often an SVP – who has the buy in and respect of the organization. That SVP or VP can sit under a number of leaders – the Chief Marketing Officer, the Chief Customer Officer, the Chief Digital Officer, even the CTO if the technology group owns digital. But the key is cross functional senior alignment and a mandate to drive change. They need to be able to flag issues, get rid of roadblocks and hold the organization accountable.

SS: You also call for a Personalization Officer.

MA: What that looks like will be very organization specific. But that role in itself can’t solve everything. So it’s also about how you set up “tiger teams” – the “two pizza teams” that can make progress in three to six months against a set of KPIs, show the value and then line up the next set of investments after that.

SS: You say the future org structure is going to look more like a jazz ensemble instead of a symphony orchestra. Can you explain what you mean by that?

MA: So instead of command and control trying to set the direction from the top down, this is where those tiger teams are run by a conductor who has the ability to improvise. If every Monday morning that tiger team is looking at the customer dashboard and looking at whether they drove engagement or not, the conductor can say, well, the content guru needs to go and make some tweaks, or the data and analytics person needs to tweak the targeting. Now let’s launch that campaign in two days and see what happens next week. So it is much more akin to a jazz group improvising than it is performing sheet music.

SS: I want to move on to a discussion about agentic AI because obviously that is one of the big game changers now. Is it the magic potion for taming the complexity of personalization?

MA: We are on the cusp of a massive revolution. What we’ve witnessed so far is push personalization. Now we are now entering this phase of pull personalization. Customers are going to be in the driver’s seat, and they get to decide when they want a personalized experience. Look at Delta and what they’ve announced with Delta Concierge (2). The Delta app will be your virtual travel assistant, anticipating when you need to go to the airport, surfacing exclusive YouTube TV content for you on your phone while you’re in the plane, and then predicting when to have that Uber show up at the gate when you land. I think this is the direction we’re going. The old world of email where you’re clicking through an experience will be transformed by these virtual assistants that are guiding you through voice and chat and video.

SS: You also talk about this idea of a brand serving as a gateway to a bunch of value added partners.

MA: Yeah, I think that’s incredibly important. It is not just one company trying to provide all these services. That’s what I love about the Delta example – tying together YouTube and Uber and Joby and all their partners to try to provide a seamless travel experience.

SS: In a world where customers will be hiring agentic AI tools to do their shopping for them, you certainly have to question the role of marketing.

MA: Yeah, yeah. I mean one of my CPG clients is revamping their entire marketing process and organization because they’ve told the Board in the next three years a third of their multi-billion dollar marketing budget will be marketing to AI-powered agents. And so in a world like that, how do you ensure the virtual assistant recommends your products? So big, big change coming.

SS: Some of the personalization challenges you lay out in the book do seem formidable, if not insurmountable. What’s your answer to the CFO who says, yeah, yeah, nice theory, but I don’t think it’s worth investing in right now?

MA: Yeah, unfortunately there are many jaded leaders out there who’ve been told just make this $10 million investment in this martech tool and it will solve everything. And then they figure out it’s made no real difference. That’s because all the “five promises” have to come together – one leg of the stool is not enough.

No company I work with has a shortage of ideas on how to improve the customer experience. So you need to get your cross functional team together in a room. In an hour you can fill the whiteboard with use cases for how you can improve the customer experience. Start with a specific channel or start with a specific step in the customer journey, or start with a specific market or set of products and map it out. What KPI are you going to move over what time period? Challenge yourself that it has to be in field within three to six months because the CFO is not patient. Explain that a proof of concept pilot will make it easier for the organization to swallow the change. Start with use cases that are going to build the confidence of the CFO and the organization.

The truth is, there are many, many proven personalization use cases that pay back within the year in every single industry. That is the biggest change actually. When I first did the personalization index benchmarking 5-10 years ago, it was really just retail and digital natives that were far ahead of everyone else. Today it’s banks and healthcare companies and B2B companies as well. So pick those use cases and win the trust of the organization that way versus spending too much time on the grand vision.

1 – Spotify’s “Daylist” is an algorithmically generated playlist for subscribers that plays the music they like to listen to at specific times.

2 – Delta Concierge is an AI-powered travel assistant from Delta Air Lines designed to personalize and streamline the travel experience.

Stephen Shaw is the Chief Strategy Officer of Kenna, a marketing solutions provider specializing in delivering a more unified customer experience. He is also the host of the Customer First Thinking podcast. Stephen can be reached via e-m