Go to Section

The Innovator’s Dilemma in Marketing Science

Avi Goldfarb

Published Online: October 20, 2026

Abstract

Artificial intelligence is creating extraordinary uncertainty for marketing managers. Yet, just when our insights as marketing scientists should be particularly valuable, our research area is again asking whether our work is sufficiently relevant. That anxiety may be exactly what we should expect. As Clay Christensen’s The Innovator’s Dilemma suggests, technological disruption makes relevance harder to recognize because the audience that matters going forward may be different from our core audience today.

---

The marketing science discussion about relevance often highlights that our work is too technical, too focused on causal inference, or too backward looking. In the process, many highlight an apparent trade-off between rigor and relevance. The claim is that as we focus more on rigor, the questions we can answer become narrower and less relevant to our audience of practitioners.

The Oxford English Dictionary defines relevance as “bearing upon the matter at hand” (Oxford English Dictionary 1988, s.v. "relevance"). Research relevance is a statement about demand, whether our work bears upon the matter at hand for a current or future audience. In my view, there is no trade-off with rigor. Careful, well-executed analysis is essential to what we offer practitioners. When rigor fails, the entire field suffers. The best-case scenario when our work is not rigorous is that managers, policymakers, and scholars in other fields ignore us. Worse, they do take us seriously, leading to bad policy, billions in losses, or both.

I do not mean to imply that our work is always relevant, nor that we cannot do better. Instead, I think that the challenge around relevance has far less to do with rigor than with finding the right audience.

One solution that appears frequently is “direct interaction with marketing stakeholders” (Lilien 2010, Kohli and Haenlein 2021). By talking to our core audience, we can anticipate their needs and develop research that benefits practitioners as well as others. This often works well. Several influential papers grew from direct interaction with practitioners, using rigorous and appropriate methods to help marketing managers solve specific problems (Misra and Nair 2011, Johnson et al. 2017, Schwartz et al. 2017, Gordon et al. 2019).

This approach, however, can fail during times of technological disruption. Clay Christensen’s The Innovator’s Dilemma highlighted that “Precisely because these firms listened to their customers…they lost their positions of leadership” (Christensen 2005, p. xv). This is a central aspect of how Christensen describes the challenge of managing disruption. A company’s existing customers are not initially interested in what a new technology makes possible. Even when they are interested, existing companies are not focused on the transformative potential of a new technology and instead focus on doing what they have always done, but better. Marketing science scholarship risks disruption if we focus only on our current audience.

Marketing is on the cusp of a new transformation through artificial intelligence (AI), which is likely the next general purpose technology similar to the steam engine, electric power, semiconductors, and the internet (Bresnahan and Trajtenberg 1995, Goldfarb et al. 2023). As marketing scholars, the open question is whether satisfying the needs of our current audience of marketing practitioners will mean we continue to be relevant many years from now.

Through the 1980s and 1990s, our core practitioner audience worked for companies selling consumer products and for large retailers. The papers in the very first issue of Marketing Science in 1982 discussed consumer packaged goods, consumer durables, retailers, restaurants, and real estate. The first issue of Marketing Science in the 1990s used examples from automotive, ready-to-eat popcorn, and consumer packaged goods.

As the internet diffused, our field debated the relevant audience. Much of our toolkit had been developed for consumer packaged goods, retail, and other well-established industries. We continued to serve that audience faithfully. Between 2003 and 2012, the Practice Prize was awarded to marketing scientists working with Procter & Gamble, Bayer, IBM, and CVS (Lilien et al. 2013). The 2012 Theory+Practice conference included the former CEO of Procter & Gamble, the CEO of Unilever North America, the CEO of Mastercard, and the CMO of Best Buy Europe.

At the same time, the future of marketing was being invented at companies like Google, Microsoft, Yahoo, and Amazon. Economist Hal Varian played a role in several of Google’s marketing efforts, examining the selling of advertising through auctions and the importance of causal inference for decision making (Varian 2009, Varian 2016). Economists also played a central role in Microsoft’s marketing products in the 2000s, including Susan Athey’s work developing ad tech auctions. Yahoo’s economists worked on auctions and ad measurement (Edelman et al. 2007, Lewis and Reiley 2014). Pat Bajari worked on demand estimation, among other things, at Amazon, and Steve Tadelis helped reset eBay’s advertising strategy and helped develop their reputation systems (Blake et al. 2015, Nosko and Tadelis 2015). Economists helped tech companies develop marketing products, including designing reputation systems, estimating the returns to advertising, and designing advertising markets (Athey and Luca 2019). These were not peripheral marketing problems. They became central questions in channel design, advertising, pricing, and product development.

Marketing science scholars risked missing this change. To be sure, there was plenty of prescient work such as the first Marketing Science issue in the 2000s, a special issue on the internet covering many forward-looking topics for new audiences including online retail, consumers and online information providers, search engines, and sellers of digital-only goods (Häubl and Trifts 2000, Lynch and Ariely 2000, Novak et al. 2000, Bradlow and Schmittlein 2000, Bakos and Brynjolfsson 2000). This issue anticipated several ideas that would be essential for understanding digital marketing, including endogenous search costs, massive bundles, and digital addiction. Over the next few years, Marketing Science would publish many more papers that anticipated potential changes that would come with increased digitization (Godes and Mayzlin 2004, Acquisti and Varian 2005, Iyer et al. 2005, Zeithammer 2007, Fay and Xie 2008, Tucker and Zhang 2010). Much of this work was highly technical, including economic theory, structural econometrics, and causal inference (Katona and Sarvary 2008, Gordon et al. 2011, Hartmann et al. 2011). But this work was often viewed by many in our community as less relevant, at least as measured by what received the Practice Prizes and what was highlighted in editorials asking for more relevance.

With the benefit of decades of hindsight, it is now clear that many of these technical papers proved highly relevant. The authors of several of these technical papers would go on to lead aspects of the marketing function in the technology industry, with many former marketing professors now at Google, Amazon, and elsewhere. Had we focused entirely on the existing audience in consumer packaged goods and offline retail, this eventual success in contributing to marketing practice may never have happened.

Thus, as the internet diffused, marketing scholarship faced an Innovator’s Dilemma. Our core audience did not seem to want papers on auction design, causal inference, and ever-more-complicated structural demand estimation. Nevertheless, enough editors and reviewers allowed these research areas to grow and eventually flourish. Being too beholden to the immediate needs of our 1990s practitioner audience would have led our scholarship to practical irrelevance.

Today, we face a similar challenge. AI is already changing the practice of marketing. We do need to help our current practitioner audience of consumer packaged goods companies, online and offline retailers, ad tech companies, financial service providers, and others learn how to address the challenges ahead. We also need to continue to help consumers make better decisions and to help governments develop policy that generates benefits of new technologies while limiting the downsides.

We also need to be open to the potential for new audiences to arise. In the coming years, the most important audience for our work may no longer be managers at today’s technology companies nor those in consumer packaged goods or retail. One possibility is that marketing management will increasingly be embedded in autonomous systems. If that happens, our audience will include the designers and evaluators of AI systems, and eventually the AI models themselves (Cowen 2025). Another possibility is a new set of organizations will arise that use some new rigorous toolkit to make current practices obsolete.

By focusing on what our current and historical audiences can understand and immediately embrace, we trade apparent relevance today for potential relevance tomorrow. Part of our job is to lead: to be ahead of the practitioners and to anticipate what information, analysis, and tools marketers will demand in a decade or more.

Christensen emphasizes that small markets do not solve the growth needs of large companies, and so “many large companies adopt a strategy of waiting until new markets are ‘large enough to be interesting’” (Christensen 2005, p. xxv). In academia, one essential purpose of our scholarship is to develop a body of knowledge for when an opportunity becomes large enough to be interesting. Focusing too narrowly on the needs of senior managers at existing companies means that we risk missing the next big thing. Not every paper will change how people think, and not every new audience will turn out to matter. Nevertheless, trying new audiences is essential if we are to have an impact during periods of disruption, including the one we are experiencing today, and to remain relevant well into the future.

---

Avi Goldfarb ([email protected]) is a 2026 ISMS Fellow, the Rotman Chair in Artificial Intelligence and Healthcare and a professor of marketing at the Rotman School of Management, University of Toronto, as well as Chief Data Scientist at the Creative Destruction Lab, a research lead at the Acceleration Consortium, and a Research Associate at the National Bureau of Economic Research. He served as Senior Editor at Marketing Science from 2016-2021.