Advertisers commonly need multiple versions of the same advertisement (ad) at varying durations for a single campaign. The traditional approach involves manually selecting and re-editing shots from longer video ads to create shorter versions, which is labor-intensive and time-consuming. In this paper, we introduce a framework for automated video ad clipping using video summarization techniques. We are the first to frame video clipping as a shot selection problem, tailored specifically for advertising. Unlike existing general video summarization methods that primarily focus on visual content, our approach emphasizes the critical role of audio in advertising. To achieve this, we develop a two-stream audio-visual fusion model that predicts the importance of video frames, where importance is defined as the likelihood of a frame being selected in the firm-produced short ad. To address the lack of ad-specific datasets, we present AdSum204, a novel dataset comprising 102 pairs of 30-second and 15-second ads from real advertising campaigns. Extensive experiments demonstrate that our model outperforms state-of-the-art methods across various metrics, including Average Precision, Area Under Curve, Spearman, and Kendall. The dataset and code are available ( https://github.com/ostadabbas/AdSum204 ).

As generative AI becomes more prevalent in advertising, firms increasingly face requirements to disclose AI involvement. Although prior research shows that such disclosures may generate negative consumer responses, it remains unclear whether explanatory disclosures can mitigate these effects. This research examines beneficiary-framed AI disclosures, explanations communicating why AI was used and who benefits, across three studies. In a controlled text-based experiment (Study 1), more specific explanations improve evaluations relative to minimal and less specific AI labels. However, beneficiary framing effects are not systematic, and practically negligible. Moreover, these benefits do not generalize to more realistic advertising contexts. Across two Instagram-style ad studies (Studies 2A and 2B), explanatory disclosures fail to improve consumer responses and, in some cases, lead to more negative evaluations with effects that are either statistically equivalent to zero or significantly negative. Across studies, AI aversion emerges as a robust predictor of negative responses, suggesting that disclosure effects are driven more by consumers’ prior beliefs than by the specific framing of explanations. The findings suggest caution in adding explanatory disclosures, as default inferences of firm-serving motives are difficult to override.

Household economic conditions are a fundamental determinant of consumer demand, yet we lack causal evidence on how they affect advertising responsiveness. We address this gap by exploiting local housing cost shocks as a source of exogenous variation in household budget pressure, instrumenting local housing price growth in a shift-share IV design, and estimating how these conditions moderate TV advertising effectiveness for 78 restaurant brands across 13 million store-weeks. We find that a one standard deviation increase in local housing cost growth nearly eliminates the baseline advertising elasticity for restaurants, with the IV estimate roughly eight times larger than the naive observational estimate. This negative housing cost-interaction effect is amplified in areas with higher renter shares, consistent with housing cost pressure operating as a negative income shock through rent increases, and effects are also significantly larger for high ad-spend brands, which show greater exposure to local economic conditions. Applying our estimates to observed housing cost trends, our projections imply that advertising elasticities swing by 0.15 or more as local conditions change, suggesting that optimal ad allocation levels may vary substantially with economic conditions. Smaller-sample evidence from 22 supermarket brands recovers a significant positive effect of housing cost shocks on grocery store ad elasticity, indicating that the direction of the effect is category-specific.

Marketing strategy research increasingly relies on multi-source datasets that must be linked across entities such as firms, brands, and products. Yet entity matching is often treated as an implicit data construction step, even though mismatches and missed links can change samples, attenuate estimates, and hinder replication. This research note synthesizes the identifier landscape most relevant to marketing strategy work, clarifying what common identifiers represent, where coverage breaks down, and when crosswalks are reliable. Building on established record linkage practice, we propose a cascading workflow comprising three approaches: deterministic joins on shared identifiers, probabilistic name-based linkage with explicit similarity and threshold rules, and ML or LLM-assisted screening for residual ambiguous cases. We then translate this workflow into an audit checklist that specifies what to record, report, and archive. The overall goal is to provide auditable standards that improve precision, transparency, and replicability in merged dataset studies.

The authors estimate the heterogeneity of TV advertising effectiveness across store characteristics and advertising levels using a large-scale panel of 135 U.S. retail and restaurant brands, and then use these estimates to assess strategies for improving TV advertising performance. The findings show significant heterogeneity in TV advertising elasticity across characteristics for over 93% of brands, but also show that firms’ observed allocations generally fail to fully exploit this estimated heterogeneity and instead covary much more closely with simple heuristics. For example, firms tend to advertise in areas where they already have high revenue, rather than in the areas estimated to have the highest incremental revenue from advertising. In particular, brands tend to overinvest in dense, high-income markets and underinvest in markets with high concentrations of college-educated residents. The authors project that brands could improve ad lift by a median 2.35 percentage points (relative to <.5% median baseline ad lift) and earn tens of millions in additional revenue under identical-budget reallocations that better leverage this heterogeneity, and that such reallocations could increase the proportion of brands achieving positive return on investment from TV advertising by 14–16 percentage points.

Marketing strategy research increasingly relies on merged multi-source datasets spanning firms, brands, and products. Yet such entity matching remains a largely undocumented research design choice, despite its potential to alter samples, estimates, and replicability. This research note develops a cascading entity-matching workflow for marketing strategy research. We outline three progressively intensive matching approaches: deterministic joins using shared identifiers and crosswalks, probabilistic name-based matching with transparent similarity rules, and screening assisted by machine learning (ML) or large language models (LLMs). Finally, we translate this workflow into a practical audit framework specifying what researchers should record, report, and archive. The goal is to establish transparent and auditable matching standards that improve precision, replicability, and cumulative knowledge development in merged-dataset research.

Marketing is widely considered one of the job functions most exposed to generative AI, but large-scale evidence on marketing-specific employment effects remains limited. We study how the public release of ChatGPT in November 2022 affected marketing employment using a panel of over 110 million employment records from Revelio Labs covering 164,780 firms. Comparing marketing to non-marketing employees within the same firm in a difference-in-differences design, we find that marketing headcount declines by approximately 0.92% relative to non-marketing headcount after ChatGPT's release. We demonstrate that this relative decline is strongest among both junior and senior employees, while mid-level marketing employees are least affected relative to same-seniority non-marketing peers; and that within fields, the contraction is concentrated in non-digital marketing and customer service employees in particular, while sales employment has enjoyed a modest relative increase. Lastly, we show in triple-differences analyses that the relative decline in marketing headcount is significantly larger in firms and industries with higher pre-period AI exposure.

PurposeThis paper aims to advance understanding of consumer vulnerability in digital service ecosystems by extending existing typologies to include a third conceptually distinct form - unobserved vulnerability - alongside experienced vulnerability and observed vulnerability. Drawing on service-dominant logic, it proposes a framework that integrates the perspectives of platform companies, regulators and consumers to examine how institutional arrangements and consumer factors shape vulnerability in human-to-human and human-to-algorithm service interactions.Design/methodology/approachThis conceptual paper synthesizes marketing and legal literatures on consumer vulnerability, platform governance and regulation. It refines the visibility-based typology and integrates it into a service-dominant logic-informed framework, supported by illustrative cases and a future research agenda.FindingsVulnerability in digital service ecosystems emerges from the interplay of platform governance (openness and transparency), regulatory frameworks and consumer factors (stigma and digital literacy), mediated through human-to-human and human-to-algorithm interactions. The proposed typology distinguishes between experienced vulnerability (recognized by the consumer), observed vulnerability (identified by third parties) and unobserved vulnerability (structural disadvantage that remains invisible to both consumers and external observers).Originality/valueBy introducing the concept of unobserved vulnerability and situating it within service-dominant logic, this paper extends prior typologies of consumer vulnerability. The framework demonstrates how visibility functions as a critical lens for understanding consumer disadvantage in digital service ecosystems. It contributes to marketing scholarship by clarifying the structural conditions under which vulnerability emerges and by outlining implications for theory, managerial practice and regulatory design.

Products people praise should be products people choose. Yet ratings and popularity can disagree. What predicts when they do? We propose that one answer lies in a cultural product's emotional variance-the degree to which distinct emotions (e.g., joy, fear, sadness) fluctuate over its duration. A multimethod investigation combines computational linguistics, machine learning, and experiments. We validate and apply a linguistic emotion classifier to more than 50,000 cultural products-over 3,000 movies, 12,000 TV episodes, and 36,000 songs. We find that products with greater emotional variance consistently receive higher ratings but are less popular, even after accounting for a broad set of alternative explanations. Two controlled experiments then manipulate emotional variance in written narratives. Before consumption, emotionally variable content seems more challenging to consume, discouraging consumption. During consumption, the same content is more engaging, leading to higher ratings. Together, the findings identify a predictable source of divergence between ratings and popularity and show how a single product feature can pull two common measures of success in opposite directions.

Firms increasingly face pressure to communicate environmental initiatives, yet public claims can also invite scrutiny, political backlash, and credibility risk. We study greenhushing, defined as the strategic downplay of substantive environmental efforts, and develop a measure to capture the extent of a firm's greenhushing and greenwashing as the gap between firms' external environmental conduct and managerial communication in earnings calls. Using 30,786,703 sentences from quarterly earnings call transcripts, a high-salience and manager-led disclosure channel, for 3,727 U.S. firms (2005-2021), we measure environmental communication with the ESG-BERT deep learning model across six environmental dimensions and pair it with TruValue Labs' environmental Pulse Scores, which capture the valence of publicly available information about firms' environmental policies, actions, and outcomes. We show that greenhushing is higher in the post-2017 period associated with increased politicization, and varies systematically with firm and industry conditions. Specifically, the extent of greenhushing is greater for firms with more financial leverage and firms operating in competitive industries. We then examine the relationship between greenhushing and quarterly cumulative abnormal returns (CAR) post-earnings call. A one SD increase in Greenhushing Tendency (SD = 1.27) is associated with a 2.29 percentage point increase in three-month CAR (0.018 x 1.27 = 0.023), translating to about $238 million in market value for the average sample firm (mean market cap = $10.4 billion). The findings suggest that silence may carry informational value in environmental settings.

The integration of generative AI into web search delivers synthesized answers to user queries, changing how people navigate and assess information, while raising concerns about the downstream impacts on publishers who supply the underlying content. We conduct a preregistered field experiment (N=1,100) on Google Search, the dominant online search platform, to estimate the causal effects of AI Overviews and AI Mode on user behavior, perceptions, and publisher traffic. We show that removing AI Overviews and AI Mode increases click-through rates to publishers, while an AI Mode-only experience reduces click-through rates and erodes user experience and trust in information found on Google. These findings show that integrating generative AI into web search reshapes online attention, with economic consequences for the online publishers that sustain both search platforms and the overall information ecosystem.

We study the potential risks and benefits of using large-language model (LLM) annotations in video ad creative research. Using a custom-built, large-scale dataset of over 10,000 human-labeled video ads, we demonstrate that off-the-shelf multimodal LLMs perform poorly when encoding certain types of features. We then show, using ad quality ratings from a large (500+) consumer panel provided by iSpot.tv, that such misaligned measurement may lead to downstream effect estimates that are significant in the opposite direction to those inferred with human-labeled data. However, we demonstrate that such bias can be largely mitigated by fine-tuning a model using our large-scale human annotations. This fine-tuned model exceeds average pairwise human agreement on many features, realigns downstream estimates with those based on human annotations, and substantially improves the explanatory power of labeled content features for ad performance, allowing for the recovery of significant effects that are otherwise missed when using human-labeled data due to inter-annotator noise.

Firms increasingly operate across multiple social media platforms, yet it remains unclear whether diversifying engagement across platforms enhances performance or simply fragments marketing efforts. We examine how the allocation of user engagement across platforms affects e commerce sales performance. Using panel data on approximately 2,000 leading U.S. online retailers from 2012 to 2019, combined with detailed engagement measures across five major social media platforms, we construct firm year indicators of engagement diversification and explore how they relate to sales performance. We find that greater diversification in engagement allocation is associated with significantly higher web sales. Importantly, this effect is not driven by platform adoption breadth or overall engagement volume. Rather than increasing traffic quantity, diversification improves conversion rates and enhances traffic quality. Mechanism analyses reveal that these performance gains stem from cross modality complementarities: engagement distributed across heterogeneous content modalities (image, video, and mixed) generates reinforcing brand exposure, whereas diversification across platforms within the same modality yields limited benefits. Furthermore, the positive effects of diversification arise only when there is sufficient overlap in audiences across platforms, providing additional evidence for a memory-reinforcement mechanism. Taken together, these findings document the importance of engagement allocation structure and highlight the role of cross-modality complementarities in multi platform digital marketing strategies.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

As video game companies adopt business models that monetize post-purchase consumption, they commonly encourage and promote engagement activities amongst consumers as a collective. However, little is known about the effect of such collective activities, which could take away time from playing the game. We introduce a novel framework that distinguishes the breadth of collective engagement from the intensity of engagement. In a unified time series framework, we model both momentary and ongoing consumption effects and feedback loops. Our analysis of 34 video game brands reveals that consumption and engagement fuel each other. Whereas the breadth of collective momentary engagement boosts momentary consumption, its intensity has a negative effect for the first 15 min. However, both ongoing breadth and intensity positively influence ongoing consumption. Importantly, game developers can amplify these effects by adding social in-use features, which both directly stimulate consumption and strengthen the reciprocal relationship between engagement and consumption. In terms of marketing actions, featured price discounts and new content releases focusing on play functionality get people talking about and searching for the game, whereas cosmetic content changes immediately increase consumption. Overall, our research underscores the need to move from a narrow focus on instant, real-time metrics toward a more strategic approach that effectively fosters engagement and consumption over time. (c) 2025 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Because negative reviews have the potential to dissuade consumers, brands necessarily worry about them. Prior literature generally supports the notion that negative information offers greater value to and influences consumers more powerfully, yet in specific circumstances, negative online reviews might be less helpful and influential. As the current research establishes, when negative reviews exhibit near (vs. far) temporal proximity cues, relative to a reviewed experience with a product or service, and some degree of negative emotionality, consumers tend to discount them in their decision-making. Such outcomes seemingly arise because consumers identify a negative review that combines near temporal proximity (i.e., posted or written in a way that makes the experience feel temporally close) with negative emotionality as a form of "venting." They then ascribe the information in the negative review to reviewer-related attributes rather than to relevant product or service quality attributes. Consistent evidence for this venting discount effect emerges from analyses of actual hotel reviews from Tripadvisor as well as multiple experimental studies involving both products and services.

Both gambling and trading involve risk-taking in exchange for potential financial gains. In particular, speculative high-risk high-frequency trading closely resembles disordered gambling behaviour by attracting the same individuals who tend to be overconfident, sensation-seekers, and attracted to quick large potential payoffs. We build on these studies via an incentivised experiment, in which we examine how manipulated levels of market volatility affected trading frequency. Gamblers (N=604) were screened based on the existence of household investments and recruited across the four categories of the Problem Gambling Severity Index. The volatility of stocks was manipulated between-participants (high vs. low). Participants traded fictitious stocks and were provided bonuses based on the results of their trading activity (M=US$4.77, range=[0, 16.99]). Participants traded more often in the high-volatility market, and this finding remained robust after controlling for financial literacy, overconfidence, age, and gender. Many investors trade more frequently than personal finance guides advise, and these results suggest that individuals are more likely to commit this error in more volatile markets. Exploratory analyses suggest that the effect of the volatility manipulation was strongest amongst gamblers who were at low-risk of experiencing gambling harms. As they might be otherwise considered low-risk, these individuals could be overlooked by protective gambling interventions yet nonetheless suffer unmitigated financial harms due to unchecked excessive trading.

Ad designers often use sequences of shots in video ads, where frames are similar within a shot but vary across shots. These visual variations, along with changes in auditory and narrative cues, can interrupt viewers' attention. In this paper, we address the underexplored task of applying multimodal feature extraction techniques to marketing problems. We introduce the "AttInfaForAd" dataset, containing 111 baby product video ads with visual ground truth labels indicating points of interest in the first, middle, and last frames of each shot, identified by 75 shoppers. We propose attention interruption measures and use multimodal techniques to extract visual, auditory, and linguistic features from video ads. Our feature-infused model achieved the lowest mean absolute error and highest R-square among various machine learning algorithms in predicting shopper attention interruption. We highlight the significance of these features in driving attention interruption. By open-sourcing the dataset and model code, we aim to encourage further research in this crucial area. (Dataset and model code available at https://github.com/ostadabbas/Baby-Product-Video-Ads).

Storytelling is a powerful tool that connects us and shapes our understanding of the world. Theories of effective storytelling boast an intellectual history dating back millennia, highlighting the significance of narratives across civilizations. Yet, despite all this theorizing, empirically predicting what makes a story successful has remained elusive. We propose narrative reversals, key turning points in a story, as pivotal facets that predict story success. Drawing on narrative theory, we conceptualize reversals as plot: essential moments that push narratives forward and shape audience reception. Across 30,000 movies, TV shows, novels, and fundraising pitches, we use computational linguistics and trend detection analysis to develop a quantitative method for measuring narrative reversals via shifts in valence. We find that stories with more, and more dramatic, turning points are more successful. Our findings shed light on this age-old art form and provide a practical approach to understanding and predicting the impact of storytelling.