Latest / The CMO Podcast with Fexingo: Marketing Leadership, Budgets, and Executive Strategy / How CMOs Are Using Digital Twins for Campaign Simulation
Transcript
- Lucas: So I was talking to the CMO of a midsize consumer brand last week — off the record — and she told me her team now runs every campaign through what she called a 'digital twin' of their target audience before they spend a single dollar on media. Luna: A digital twin of a person? Like a synthetic version of a real consumer? Lucas: Exactly. It's not a real individual, but a statistical replica built from first-party data, behavioral signals, and sometimes third-party demographic clusters. Think of it as a simulated persona that behaves — clicks, scrolls, converts — like a real segment. Luna: And companies are actually using these to make budget decisions? That feels like a big leap from classic A/B testing. Lucas: It is a leap. But a growing number of CMOs I talk to say it's becoming standard practice for upper-funnel testing. Instead of running five different ad creatives to 10,000 people each and waiting two weeks for results, they feed the creatives into a digital twin environment and get predicted response curves in hours. Luna: So it's basically a simulation sandbox. How accurate are those predictions? Lucas: That's the billion-dollar question. The vendors selling these platforms claim 85 to 95 percent correlation with real-world outcomes. But I've heard from marketers who say it's more like 70 to 80 percent — still useful, especially for ruling out duds before you spend. Luna: And I imagine it's a huge time-saver. Two weeks to a few hours is a massive efficiency gain. Lucas: Right. And the cost savings are real too. One report I saw from a CPG company — I think it was PepsiCo, though they didn't confirm it publicly — said they cut their creative testing budget by about 30 percent in the first year using digital twins. Luna: Thirty percent? That's enough to get any CFO's attention. But let's talk about the elephant in the room — data privacy. You're building a replica of a real person. Is that crossing a line? Lucas: It's a fair concern. But here's the nuance: the digital twin isn't a copy of a specific individual. It's an aggregate model that captures the statistical patterns of a segment. No actual personal data is stored in the twin. Think of it like a synthetic data set that reflects the distribution of behaviors without referencing any one person. Luna: Okay, so it's privacy-safe in principle. But how do you validate that the twin actually represents your audience? If the model is biased or built on stale data, you could be making decisions on a fantasy. Lucas: That's the Achilles' heel. A lot of these twins are trained on historical data — past campaign responses, CRM data, maybe some third-party panels. If consumer behavior shifts, say, after a major economic event or a platform algorithm change, the twin might not reflect reality. So smart CMOs are constantly recalibrating with real-world holdout groups. Luna: So it's a supplement, not a replacement. You still need some real testing to keep the models honest. Lucas: Exactly. And that's the honest take I hear from practitioners. Unilever, for example, has been running digital twins for their Dove campaigns, but they always reserve about 10 percent of their test budget for live validation. They use the twin to narrow down from 50 creatives to 5, then test those 5 in the real world. Luna: That makes sense. It's a filtering tool, not a crystal ball. So who are the main vendors in this space? Lucas: There are a handful. One is called Synthetic Minds — they started in gaming, simulating player behavior, and now do consumer twins. Another is Replica AI, which focuses on email and push notification response. And the big platforms like Adobe and Salesforce have started integrating twin-like features into their marketing clouds. Luna: Interesting. And are we seeing this more in B2C or B2B? Lucas: Definitely more B2C so far, because the data volumes are larger and the purchase cycles are shorter. But B2B is starting to creep in. A few enterprise software companies are building twins of IT decision-makers, simulating how they respond to whitepapers and demo requests. The lead time is longer, so the simulation horizon is measured in months rather than days. Luna: I could see that being powerful for account-based marketing. You simulate the whole buying committee's reaction before you even reach out. Lucas: Right. And one company I read about is using digital twins to simulate 'what if' scenarios — like what happens to conversion rates if you change your pricing page, or if you run a LinkedIn ad versus a podcast sponsorship. It gives the CMO something to take to the CFO that's more than a gut feel. Luna: That kind of data-driven argument is gold in budget meetings. So, what's the catch? There's always a catch. Lucas: The catch is that building a good digital twin requires a lot of clean, rich data. If your data is siloed across departments or full of gaps, the twin will be too. And the upfront investment — both in software and in data engineering time — can be significant. Smaller brands without a data science team might struggle. Luna: So it's a tool for the enterprise right now. But I imagine as the platforms mature, it'll trickle down to mid-market. Lucas: That's the trajectory. In fact, I've seen a few startups offering 'digital twin as a service' where they build the model for you using your existing CRM and ad platform data. Pricing is usually a monthly subscription plus a per-simulation fee. So it's becoming more accessible. Luna: And with that accessibility comes the risk of misuse. If every brand starts using the same twin models, won't campaigns start to look homogenous? Lucas: That's a smart observation. If everyone optimizes for the same predicted response, you could end up with a creative monoculture. But the best CMOs I know use twins to explore unconventional angles — they feed in wild card creatives alongside safe ones and see what the twin predicts. Sometimes the counterintuitive idea wins. Luna: Because the twin isn't biased by human assumptions. It's just pattern matching. Lucas: Exactly. And that's where the real value is. Not just saving money, but discovering insights you wouldn't have tested otherwise. One CMO told me her team found that a slightly awkward, unpolished testimonial video outperformed a slick produced one — something they never would have believed without the simulation data. Luna: That's a great example. It's like the Dollar Shave Club ad — raw and authentic. But that was a human intuition. Digital twins might help you find those moments systematically. Lucas: If these marketing conversations have sparked something you've actually used — an idea, a tactic, a framework — consider tossing a few dollars toward keeping them going. A couple of dollars a month is genuinely what keeps these shows ad-free and independent. If you've gotten something out of them, buy me a coffee dot com slash fexingo. Luna: Yeah, it really does make a difference. We hear from listeners who say they've applied something from an episode and saved thousands on a campaign. That's the kind of impact that keeps us going. Lucas: Alright. Back to digital twins — I want to talk about where this is headed in the next 12 months. I'm hearing whispers of real-time twins that update in near real time as consumer behavior shifts. Luna: So instead of a static model, you'd have a living simulation that adapts week to week. That's a whole different level of granularity. Lucas: Exactly. Imagine you're a CMO at a retailer and you see a sudden dip in in-store traffic. A real-time twin could instantly simulate the impact of pushing a new promotion versus adjusting your media mix. You'd have a recommendation in minutes, not days. Luna: That sounds like it would require an enormous amount of compute power and data integration. Is that realistic for most companies? Lucas: Not yet. But cloud providers like AWS and Google Cloud are offering specialized instances for these workloads, and the cost is dropping. I think within two years, real-time simulation will be a standard feature in enterprise marketing suites. The question is whether the quality of the data keeps up. Luna: And whether consumers are comfortable with their behaviors being constantly simulated, even if anonymized. There's a trust angle here that CMOs can't ignore. Lucas: Absolutely. Transparency is going to be key. Some brands are already experimenting with opt-in models where consumers get a small reward for allowing their anonymized data to be used in twin training. It's a value exchange. Luna: That could actually build trust if done right. 'Help us serve you better ads by letting our AI learn from your patterns' — but it has to be clear and not buried in terms of service. Lucas: Right. And that's the challenge. But the potential is huge. If we can simulate campaigns accurately enough, we could reduce ad waste from the current estimated 30 to 40 percent down to single digits. That's billions of dollars in efficiency. Luna: And a lot fewer annoying ads. That's a win for everyone. Lucas: Exactly. So digital twins aren't a gimmick. They're becoming a core part of the marketing toolkit. If you're a CMO or a marketer, I'd start exploring the data readiness in your own organization. Because the companies that get this right will have a real edge. Luna: And the ones that don't might find themselves simulating campaigns that never run. Something to think about. Lucas: That's a good note to end on. We'll be back next week with another angle.