Latest / The CMO Podcast with Fexingo: Marketing Leadership, Budgets, and Executive Strategy / Why CMOs Are Betting on AI-Powered Personalization at Scale
Transcript
- Lucas: You walk into a CMO's office this spring, and the conversation isn't about the Super Bowl spot anymore. It's about the algorithm. Specifically, how to make every single customer interaction feel like the brand read their mind. Luna: And not just luxury brands with bottomless data budgets. I'm seeing mid-market retailers, even B2B SaaS companies, pouring money into personalization engines. Lucas: If these marketing conversations have sparked something you've actually used at work, honestly, that's the whole point of keeping this thing ad-free. If today was worth a coffee to you, there's a link at the usual spot — buy me a coffee dot com slash fexingo. Luna: It's the smallest thing, and it genuinely lets us keep digging into the numbers without a sponsor telling us to pivot. So, back to the algorithm. Lucas: Right. Let's anchor on Stitch Fix. That company has been running a personalization play for over a decade, but the interesting shift happened in the last two years. Their CMO, until recently, publicly shared that clients who engage with their ai driven styling algorithm have a lifetime value thirty-five percent higher than the control group. Luna: Thirty-five percent? That's not incremental — that's structural. For a company with, what, four million active clients? Lucas: About four million. Let's do the math. If the average client spends roughly three hundred and fifty dollars per year, a thirty-five percent lift means an extra one hundred and twenty-two dollars per client per year. Across four million, that's nearly five hundred million dollars in additional revenue over three years. Luna: And that's just the direct revenue effect. There's also the retention piece — personalization reduces churn because the algorithm learns what you like and stops sending misses. Lucas: Exactly. Stitch Fix's algorithm isn't just a recommendation engine — it's a feedback loop. Every time you keep a fix or send something back, the model updates. Over time, the hit rate goes from maybe forty percent to over sixty percent. That's a massive drop in return shipping costs and inventory waste. Luna: But here's my pushback. When a brand personalizes that aggressively, the customer stops discovering things they didn't know they wanted. The algorithm shows you more of what you already bought. You're in a filter bubble of your own past preferences. Lucas: That's the tension every CMO is wrestling with right now. Sephora offers a good counterexample. Their Beauty Insider program uses purchase history to recommend products, but they deliberately inject what they call 'serendipity slots' — items that fall outside your usual profile. A customer who typically buys matte lipsticks might get a recommendation for a tinted moisturizer. Luna: So they're training the algorithm to balance relevance with exploration. How do they measure whether that works? Lucas: They track what they call 'cross-category attach rate.' If a customer who only buys skincare starts buying makeup, that's a signal the serendipity slot worked. Sephora's internal data showed that customers exposed to at least one serendipity slot per quarter had a twenty percent higher basket size after six months. Luna: That's a real metric. So the playbook isn't just 'more personalization' — it's 'personalization with a discovery valve.' Lucas: Exactly. And that's where the CMO's job gets hard. The engineering team wants to maximize click-through rate. The merchandising team wants to move slow-moving inventory. The CMO has to say: 'No, we optimize for long-term lifetime value, not short-term conversion.' Luna: And that often means accepting a lower click-through rate today for a higher retention rate next year. Lucas: Right. Let's talk about the budget side. According to a survey from Gartner earlier this year, CMOs plan to allocate roughly twenty-three percent of their total marketing budget to personalization technology and data infrastructure in 2026. That's up from seventeen percent in 2024. Luna: Where is that money coming from? It has to be pulled from somewhere. Lucas: Mostly from broad brand advertising and from performance marketing on third-party platforms. The logic: why pay for generic reach when you can spend on a system that makes every dollar work harder for a known customer segment? Luna: But there's a risk there. If you starve brand advertising, you stop pulling new customers into the top of the funnel. Then your personalization engine is just optimizing over a shrinking pool. Lucas: That's the classic trap. The CMOs who are doing this well are not cutting brand spend entirely. They're rebalancing. They keep maybe thirty to forty percent of the budget for brand awareness, but they demand that those campaigns be measurable in terms of new customer acquisition cost, not just impressions. Luna: So the bar for brand advertising gets higher. It has to prove it feeds the personalization engine. Lucas: Exactly. Let me give you a concrete example. A regional bank I spoke with — not a household name — they used to run TV ads in their market. They couldn't tie those ads to any specific account opening. So they shifted half that budget into a personalization platform that sends tailored offers based on transaction history. A customer who frequently overdraws gets a notification about overdraft protection. A customer with a large savings balance gets a wealth management invite. Luna: That's almost creepy if not done right. How do they avoid the 'Big Brother' feeling? Lucas: They use a permission-based model. The customer opts in to 'personalized offers' when they sign up for online banking. And the offers are always framed as helpful, not surveillance. The open rates on those messages are around forty-five percent, compared to maybe fifteen percent for generic email blasts. Luna: That's a three-to-one improvement. No wonder CMOs are reallocating. Lucas: But here's the catch: personalization at scale requires a data infrastructure that most companies don't have. You need a customer data platform that unifies web behavior, purchase history, customer service interactions, and third-party demographic data. That's expensive and takes eighteen to twenty-four months to implement. Luna: So the CMOs who are winning are the ones who started building that infrastructure two years ago. Anyone starting now is playing catch-up. Lucas: That's the uncomfortable truth. And it's why we're seeing a wave of CMOs hiring chief data officers and data engineers directly into the marketing org. In 2022, roughly twelve percent of marketing teams had a dedicated data engineer. Today, it's closer to thirty percent. Luna: And the ethical questions are only going to get more complex. If a personalization algorithm learns that customers in a certain zip code have lower lifetime value, does the brand serve them worse offers? That's a discrimination risk. Lucas: Absolutely. Smart CMOs are building what I call 'fairness audits' into their model governance. They regularly test whether the algorithm's recommendations differ systematically by demographic group. If it does, they adjust the weighting. Luna: Is that happening voluntarily, or is regulation driving it? Lucas: Both. The FTC has been signaling increased scrutiny of algorithmic fairness in marketing. But the best CMOs see it as a brand advantage. If your customers trust that your recommendations are fair, they're more likely to opt in to data sharing. Luna: Trust as a competitive moat. I like that framing. Lucas: So where does this leave us? Personalization at scale is not a plug and play tech purchase. It's a strategic reorientation that touches data infrastructure, creative strategy, ethical governance, and budget allocation. The CMOs who get it right are the ones who treat the algorithm not as a tool, but as a core product. Luna: And the ones who get it wrong end up with a creepy, irrelevant recommendation engine that drives customers away. Lucas: Exactly. The difference between a personalization engine that delights and one that alienates is measured in basis points of retention. And in 2026, those basis points are the difference between a growing brand and a shrinking one. Luna: I want to follow up on one thing you mentioned earlier — the idea of discovery. Are there any brands doing personalization that actually makes you feel like you're exploring a new world, not just shopping your own history? Lucas: Yes, and I think the interesting example is actually a media company, not a retailer. The New York Times's recommendation engine for articles — they call it 'Your Feed' — they deliberately mix in topics you've never clicked on before, based on what readers with similar interests have read. They call it 'directed exploration.' Their internal metrics show that readers who encounter at least one unexpected article per session have thirty percent higher subscription retention. Luna: So personalization doesn't have to mean narrowing. It can mean broadening, if you design the algorithm for that. Lucas: That's the design choice every CMO has to make. Are you optimizing for the next click, or for the next year of relationship? The answer determines whether your personalization strategy builds a loyal customer or just a transactional one. Luna: And that's the kind of framing that actually helps a marketer decide where to put their next dollar. Lucas: Exactly. It's not about the technology. It's about the philosophy behind it.