AI & Marketing Research with Dr. Eva Wolf
Not another AI news podcast. This is a research radar — a twice-weekly briefing that surfaces peer-reviewed studies on AI and marketing, tells you what the evidence actually says, and helps you decide what's worth a deeper read.
AI & Marketing Research with Dr. Eva Wolf
AI Marketing Performance Research: What Actually Creates the Edge?
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Here's the uncomfortable question this week. If every marketing team is buying the same AI tools, same platforms, same targeting features, same dashboards, what actually creates the edge? That's what both papers today are circling. And the answer is more inconvenient than most vendors want you to hear. We screened 140 papers this week. Two made the radar, both about AI-driven marketing performance, both with real methodological problems I'm going to flag, both still worth your time. Quick caveat: this is a first-pass research briefing, not a final academic review. I'll tell you what the papers suggest, what they don't prove, and what I'd actually do with each one. Okay, let's get into the first one. Paper one, the question on the table. Is the AI tool itself what gives you a competitive edge, or is it something else entirely? The researchers looked at 15 real AI marketing implementations: retail, automotive, financial services, companies that had been running AI-driven marketing for at least 18 months. They combined a literature synthesis with comparative case analysis, using each company's own historical results as the baseline. And the headline numbers are striking. Conversion rates up roughly 30%, cost per acquisition down around 40%, return on ad spend up over 50%, compared to each company's pre-AI results. Now, before you screenshot that and send it to your CFO, I need to tell you where those numbers actually came from. Every single data point was self-reported under confidentiality agreements, no named companies, no independent audit, and only companies willing to share their implementation details were included, which almost certainly skews toward the success stories. So treat those percentages as directionally interesting, not as benchmarks you can cite. Because here's the thing: the real finding in this paper isn't the numbers, it's the framework underneath them. The researchers argue the AI technology is not the competitive advantage. What creates a lasting edge is the combination of three things proprietary customer data built up over years, analytical talent that knows what to do with it, and a culture that trusts data over gut instinct. They had one example, a fashion retailer that needed three full years of detailed behavioral data to build nearly 850 distinct customer microsegments. Competitors couldn't copy that, not because the AI was special, because the data history was. That's the piece I care about. Your competitors can license the same AI platform you're using by next quarter. They cannot go back in time and collect three years of your customers' behavior. They also flag something I think gets underestimated constantly. In the automotive cases, experienced marketing managers resisted the algorithm's recommendations. The algorithm was overruling their instincts, and nobody senior was visibly backing the data-driven approach. The rollout stalled, not a technology problem, a culture problem. Plain English payoff. Your AI tools are only as powerful as the years of customer data behind them, and the culture willing to act on what the algorithm says. Money move. If your company has been sitting on years of customer behavioral data without building models on top of it, that data is your most defensible asset right now. And the window to act before a competitor catches up is closing. Try this by Friday. Pull up your longest-running customer data asset. Purchase history, email engagement, behavioral events, and ask honestly, are we running AI models on this or are we just storing it? If the answer is just storing it, that's your next conversation. Evidence check. The data is unaudited and self-reported by companies with every incentive to look good. The journal isn't a high credibility marketing venue. Do not cite the specific percentages as established benchmarks, ever. Radar verdict. Use cautiously. The framework, data depth, analytical talent, data culture is genuinely actionable. The performance numbers are not independently verified and shouldn't be treated as proof of anything. Paper two, same territory, AI and marketing performance, but this one zooms in on direct to consumer brands and makes an argument about how AI investment should be structured. The core claim: companies that embed AI across their whole marketing operation grow faster than companies that bolt on a single tool. One channel, one campaign, one chatbot, that's not enough. The growth comes from integration. The researchers used structural equation modeling and neural networks to test this. And they found AI helped revenue growth two ways. Directly by making operations more efficient, and indirectly by improving how well companies understood and personalized to their customers. Customer intelligence was the key link in that chain. They also found neural network models outperformed older statistical methods when predicting customer lifetime value, which, if you're still running CLV estimates out of a spreadsheet or a basic regression, is worth paying attention to. Now I have to be direct about the problems with this one. The paper is published on Zenodo through what appears to be a very low credibility open access journal. The sole author is a practicing CEO of a D2C company, so there's a real conflict of interest question. The full PDF wasn't fully extractable, which means the actual sample size and effect sizes are unknown. And the reference list apparently includes citations that have nothing to do with marketing, which is a sign of either very careless sourcing or something worse. So I can't hand you this paper and say trust it. I can't. What I can say is that the directional logic, integrated AI beats isolated AI tools, and customer intelligence is the mechanism, is consistent with what stronger research shows. This paper doesn't prove it, but it's pointing at something real. Plain English payoff, adding one AI tool to your stack isn't a strategy. Connecting AI across your whole customer journey is what actually moves revenue. Try this by Friday. Map out which parts of your customer journey currently have any AI touching them. Segmentation, personalization, campaign optimization, CLV forecasting. Look at the gaps. That map tells you where isolated tools end and integration could start. Evidence check. This paper has serious credibility problems. Low quality venue, potential conflict of interest, unknown sample size. Use the logic, do not cite the paper. Radar verdict. Use cautiously. And I mean that more strongly here than for paper one. The concept is sound. The evidence behind this specific paper is not. Treat it as a hypothesis worth testing, not a finding worth citing. Okay, so here's what I think is actually happening across both papers this week. They're telling the same story from two different angles. Paper one says the technology isn't the advantage, the data and the culture are. Paper two says isolated tools aren't enough. Integration is what drives growth. Those two ideas are completely consistent. And together they're pointing at something I think most marketing teams are getting wrong. Most teams are evaluating AI tools individually. Does this targeting platform perform better? Does this personalization engine lift conversion? That's the wrong frame. Both papers, despite their different methodological quality, are arguing that AI and marketing is a systems problem, not a tools problem. Here's what I keep coming back to, though. Neither of these papers is strong enough to base a major strategic decision on a loan. Paper one has unaudited self-reported data from a biased sample. Paper two has credibility problems I can't paper over. But they're both pointing in the same direction as stronger work in this space. And that directional consistency is worth something, not as proof, as a signal to test. The tension worth naming. Both papers are implicitly optimistic about AI adoption. Neither seriously interrogates what happens when the algorithm makes a bad call, and there's no experienced marketer left who trusts their own judgment enough to override it. That cultural risk is real. The automotive case in paper one hints at it, but neither paper goes there fully. Here's the playbook from this week. One, audit your data assets before you buy your next AI tool. How many years of behavioral data do you actually own? That's your real competitive moat, not the platform subscription. Two, find out who in your organization is resisting algorithmic recommendations and why. If senior leadership isn't visibly backing the data-driven approach, your AI rollout will stall the same way the automotive cases did. 3. Map your AI coverage across the full customer journey. If it's concentrated in one channel or one campaign type, you've got isolated tools, not an integrated system. That gap is where the growth is. Evidence check on all of that. Both papers today are use cautiously verdicts. One has unaudited confidential data and selection bias. The other has serious venue and methodology concerns. Use these findings to decide what to test internally, not as validated benchmarks to present upward. Links to both papers are in the show notes, along with the full limitation lists I didn't read aloud. Read the originals before making major decisions. These summaries are the starting point, not the finish line. See you Thursday, and if something from today changed how you're thinking about your AI stack, I want to hear it.