Customer Behaviour in Response to Disaster Announcements: A Big Data Analysis of Digital Marketing in Hospitality

Abstract

In today’s hyperconnected world, disaster announcements—regardless of actual impact—can significantly shape consumer behaviour and brand perception in the hospitality sector. This study investigates how customers respond online to disaster-related signals, focusing on digital marketing activities by luxury hotels in Santorini, Greece. Drawing on a case study of the Santorini Earthquake in February 2025—during which the Greek government declared a state of emergency—we use big data analytics, including web traffic metrics, social media interaction and fuzzy cognitive mapping, to analyse behavioural shifts across platforms. The findings indicate that disaster signals trigger increased engagement, altered sentiment and changes in advertising efficiency. This study provides actionable recommendations for tourism destinations and hospitality brands on how to adapt digital strategies during crisis periods.

1. Introduction

Tourism is a sector deeply exposed to disruption. Natural disasters—such as earthquakes, floods and wildfires—can have sudden and profound effects on destination image, consumer confidence and operational continuity (T. Wang et al., 2022). The hospitality industry, in particular, is acutely vulnerable, given its reliance on physical infrastructure, perception of safety and consumer trust (Alvarez et al., 2022). In this evolving risk environment, the digital environment has become a frontline space for both customer decision-making and brand response, as travellers increasingly interact with hotels and destinations through search engines, booking platforms and social media (Angeloni & Rossi, 2020; Kumar, 2024).

Despite the rising importance of digital platforms, relatively little is known about how consumer behaviour evolves online in response to real-time disaster-related signals (Ritchie & Jiang, 2021). Much of the existing literature on tourism crisis and disaster in tourism focuses on macro-level outcomes such as destination recovery, resilience planning or government response (Singh et al., 2022; Zhang et al., 2024; Rahmafitria et al., 2021). However, these studies often overlook the granular, real-time behavioural adjustments consumers make in digital environments—adjustments that may precede physical travel decisions. As highlighted by Ritchie and Jiang (2021) and Song et al. (2025), there is an urgent need to examine the temporal and behavioural dimensions of tourist responses—particularly the real-time micro-level shifts in engagement that occur across digital platforms during disaster periods. Motivating this study is the growing urgency for real-time, data-informed decision-making in tourism crisis contexts. As disasters become more frequent and digitally mediated, hospitality providers are under pressure to interpret behavioural signals quickly and respond with targeted messaging across fragmented platforms. This digitization of crisis response has made real-time behavioural data not only measurable but strategically indispensable. Despite this need, few studies offer frameworks that connect platform-specific metrics—such as bounce rate, session depth or direct traffic—to emotional and cognitive states triggered by risk perception. By analysing how digital engagement unfolds before and after a disaster announcement, this study aims to equip decision-makers with actionable insights that go beyond post-crisis reflection and enable proactive digital strategy design during disruption.

Although distinctions between crises and disasters (Faulkner, 2001) are useful in framing systemic responses, most tourism disruptions—such as pandemics and earthquakes—combine natural and human causes (Bhati et al., 2016; Walch, 2014; Wisner et al., 2004), with the true impact reflected in the behavioural shifts they provoke (Neef & Grayman, 2018). Increasingly, these shifts are manifested in digital spaces. Revealing how consumer trust and engagement evolve across platforms during disaster periods offers a novel lens for understanding consumer-brand dynamics (Liu et al., 2024). Research from adjacent sectors supports this behavioural focus. Sakas et al. (2022b) observed that during the COVID-19 pandemic, consumers initially turned to well-known, trusted brands, later explored alternatives as risk familiarity grew and eventually returned to established providers during renewed uncertainty. Similarly, Sakas et al. (2022a) demonstrated how centralized firms strategically adapted their social media strategies following war-related disruptions by emphasizing emotional tone, consistency and message agility.

Building on these insights, this study argues that hospitality consumers display similar patterns of temporal, platform-specific behaviour in response to disasters, as suggested by Matiza and Kruger (2021). These patterns are often visible through real-time changes in digital behaviour—such as fluctuations in website traffic, bounce rate, paid search responsiveness and social media engagement. Duan et al. (2021), in their comprehensive review of the tourism crisis literature, highlighted that most existing studies focus on broad destination-level impacts and long-term recovery strategies, rather than on immediate behavioural shifts at the individual level. Their findings underscored a critical research gap in micro-level, time-sensitive responses, particularly as they unfold across digital platforms. Filippou et al. (2024), meanwhile, demonstrated how different digital marketing channels—such as organic search, paid campaigns and social media—interact to influence direct traffic and conversion patterns. However, their work did not specifically address crisis contexts or behavioural volatility during disaster periods. Building on these insights, this study introduces a temporal framework to capture how hospitality consumers recalibrate digital engagement in response to disaster announcements—focusing on behavioural immediacy, platform interplay and the role of risk perception in shaping online decision-making.

To address these gaps, the present study proposes four research questions that explore how behavioural signals evolve across digital channels in the wake of a disaster announcement. Each question targets a distinct yet interconnected aspect of consumer engagement and marketing strategy in crisis conditions. This paper uses the Santorini Earthquake of February 2025 as a real-time case study to investigate the following research questions, drawing from user data on luxury hotel websites and related digital channels.

RQ1: 

How do different types of online traffic and engagement patterns influence future digital advertising effectiveness?

RQ1 investigates how different types of web traffic influence subsequent advertising effectiveness. This question stems from the need to understand how early-stage information-seeking (e.g., organic traffic) predicts future marketing performance—an underexplored link in disaster tourism literature.

RQ2: 

What is the relationship between user interaction quality and behavioural outcomes on hospitality websites during the period preceding a disaster announcement?

RQ2 focuses on the relationship between user interaction quality (e.g., bounce rate, pages per visit) and observable behavioural outcomes in the period preceding a disaster announcement. While prior studies emphasize broad recovery phases, few have examined how pre-crisis engagement patterns reflect early signals of anxiety, trust-seeking or cognitive dissonance—despite their importance in predicting subsequent consumer response.

RQ3: 

How do historical and concurrent marketing activities across various digital channels contribute to sustained customer interest and loyalty?

RQ3 explores the cumulative role of historical and current marketing activities in fostering brand loyalty during crises. This question addresses the lack of temporal models that connect pre-existing brand relationships with post-disaster digital resilience.

RQ4: 

In what ways do cross-channel interactions shape the performance of targeted advertising strategies in dynamic, crisis-affected environments?

RQ4 examines how cross-channel exposure shapes targeted advertising performance during volatile conditions. It reflects the growing need to understand how users navigate across platforms (e.g., from paid search to social media) in emotionally charged contexts like the Santorini Earthquake.

The overarching objective of this study is to explore how disaster-related announcements affect short-term consumer behaviour across digital marketing channels in the hospitality sector. Specifically, it aims to (1) identify how users interact with digital platforms before and after a crisis signal, (2) understand the role of different traffic sources (e.g., organic, paid, direct, social) in shaping trust and conversion, (3) examine how engagement metrics like bounce rate and session depth reflect emotional and cognitive responses to risk and (4) develop a data-driven framework for optimizing digital strategies during disaster periods, grounded in the case of the 2025 Santorini Earthquake. The remainder of this paper is organized as follows: Section 2 offers a comprehensive review of the relevant literature. Section 3 outlines the hypotheses and conceptual framework. Section 4 details the methodology employed. Section 5 presents the empirical results. Section 6 discusses the insights gleaned from the findings, while Section 7 concludes with implications for theory, practice and future research.

2. Theoretical Background

2.1. Disaster Announcements and Digital Behavioural Response

The tourism and hospitality sectors are highly exposed to disaster-related disruptions—not only due to their operational fragility, but also because consumer behaviour can rapidly shift in response to perceived threats (Verrucci et al., 2016; Pretto et al., 2023). Traditionally, tourism disaster research has focused on the aftermath of events—such as economic impact, destination recovery efforts and communication strategies (Çakar & Aykol, 2023). However, more recent scholarship emphasizes the anticipatory effects of disaster announcements, including early media coverage, government alerts and social media discourse, as critical triggers of behavioural change (Pahrudin et al., 2023).

A compelling example is the February 2025 earthquake swarm in Santorini, where over 80 seismic tremors occurred between 1 and 10 February, intensifying media and public attention. As EMSC (2025) reported, the seismicity prompted heightened monitoring, while Euronews (2025) documented evacuations and official advisories from the Ministry for Climate Crisis and Civil Protection (2025). Communication efforts spanned multiple media: early information was provided via national portals such as MySafetyPlan (2025), while real-time alerts during the swarm were delivered through the 112 SMS system, hotels’ websites and social media. Seismic updates were shared by the Ministry for Climate Crisis and Civil Protection (2025), while traditional media and tourism platforms amplified messaging. Hotels and travel providers used mobile apps, booking engines and newsletters to manage guest communication and logistics, following guidance from national preparedness plans (Travel + Leisure, 2025). In the aftermath, continuous digital communication played a key role in reassuring tourists, providing updates on safety inspections, restoring service and facilitating gradual normalization of travel plans.

Disaster announcements serve dual functions as psychological triggers and informational cues that shape real-time user behaviour across digital platforms, prompting consumers to reassess travel plans based on perceived safety, destination stability and service reliability (Lee & Yu, 2020; Jiang et al., 2021). These anticipatory responses are particularly evident in digital spaces, where the immediacy of information amplifies uncertainty and speeds up decision-making (Khan et al., 2023). Heightened threat perception—often influenced by media narratives—can lead to cancellations, delayed bookings or increased demand for reassurance and clarity from service providers.

These behavioural adaptations align with risk perception theory, which posits that individuals assess threats through a combination of emotional response and perceived control (Cheng et al., 2022). In tourism, perceived risk has been shown to influence not only destination choice and travel intention but also how and where consumers seek information online (Carvalho, 2022). Today’s travellers rely heavily on digital platforms—search engines, hotel websites, online reviews and social media—to form perceptions of safety and credibility (Angeloni & Rossi, 2020; Ho et al., 2022; Pocchiari et al., 2024). Consequently, observable changes in digital behaviour during disaster warning periods—such as increased search activity or shifts in traffic sources—can serve as early indicators of market sentiment and brand trust.

It is important to distinguish between the behaviour of tourists planning a trip and those already present in the destination. For example, tourists physically in Santorini during the earthquake swarm faced a heightened need for real-time information and immediate decision-making support. Their digital behaviour reflected urgency and emotional intensity: many relied on mobile devices to access evacuation notices, locate safe zones and verify hotel safety measures (Travel + Leisure, 2025). Unlike potential travellers browsing future options, these users engaged heavily with localized digital (emergency services, hotel announcements) and sought reassurance through peer commentary and updates on social platforms. For these on-site tourists, digital communication was not merely informative—it was central to safety, control and psychological coping during the unfolding crisis. In this context, the effectiveness of crisis communication becomes critical in shaping consumer perceptions. As Acikara et al. (2023) note, perceived risk significantly affects destination image and travel intention, but well-timed and transparent communication can moderate these effects. Clear messaging not only reduces uncertainty but also helps determine whether travellers perceive a destination as resilient or unprepared. Similarly, Seneviratne et al. (2024) emphasize the dual role of social media in disaster situations: while it facilitates real-time communication, it can also amplify misinformation and public anxiety. Increasingly, consumers turn to these platforms not only for updates but to assess the preparedness and responsiveness of hospitality brands and tourism authorities.

The Tourism Disaster Vulnerability Framework adds a geographic layer to these behavioural insights. As Becken et al. (2013) highlight, small island destinations are particularly exposed to disaster risk due to their heavy reliance on tourism, limited infrastructure and geographic isolation. In such contexts, disaster announcements often provoke swift and emotionally charged consumer responses, shaped by perceptions of evacuation feasibility, threat severity and the reliability of available information—much of which is accessed digitally. These evolving consumer sensitivities, shaped by disaster signals, must be interpreted through a behavioural lens—particularly through how individuals perceive risk and how trust is formed, eroded or regained during times of uncertainty.

2.2. Risk Familiarity and the Evolution of Digital Trust in Crisis Contexts

Understanding how consumers respond to disasters requires a temporal lens that captures how behavioural and psychological patterns evolve before, during and after a crisis. In the pre-disaster phase, travellers often enter a state of anticipatory uncertainty, marked by increased search activity, risk scanning and emotional sensitivity to emerging signals. During the acute phase, when disaster signals are confirmed and emergency communication peaks, user behaviour becomes urgent, trust-seeking and emotionally charged—as reflected in real-time digital activity like mobile searches, safety verification and decision-making. In the post-disaster phase, consumer engagement typically shifts toward re-evaluation and normalization, where users recalibrate their trust and may re-engage with familiar brands or explore safer alternatives (Mert & Koksal, 2025; Ministry for Climate Crisis and Civil Protection, 2025). This process aligns closely with the risk familiarity effect (Sakas et al., 2022b), which conceptualizes these stages as part of a cyclical trust recalibration, shaped by emotional processing, digital exposure and environmental stability.

The risk familiarity effect refers to the evolving way consumers perceive and respond to risk based on their prior exposure to similar situations. As individuals become more familiar with a threat—whether it be a public health crisis, natural disaster or other form of disruption—their behaviour tends to shift from avoidance and caution to more normalized patterns. In the digital and consumer behaviour literature, this concept is particularly relevant in explaining loyalty, trust and decision-making during crises. In tourism and hospitality contexts, this temporal cycle influences how travellers engage with digital content, from initial reassurance-seeking to revalidation of brand loyalty. As users move through each phase, their behaviour—channel choice, engagement depth and conversion intent—shifts accordingly. These dynamics provide the foundation for interpreting digitally observed behaviour in disaster settings.

In the early stages of a disaster announcement, consumers tend to gravitate toward trusted brands as a form of emotional reassurance (Sakas et al., 2022b). As familiarity with the risk environment increases, they may explore alternative providers or prioritize value over safety—only to return to established brands if uncertainty resurfaces or media narratives intensify. These cyclical shifts in brand trust are influenced not only by external conditions but also by consumers’ prior experience, information exposure and the consistency of a brand’s digital presence (Czarnecki et al., 2023).

In digital environments, shifts in consumer trust and perception are often reflected through channel-specific behavioural patterns. For example, an increase in direct traffic typically signals brand familiarity and confidence (Filippou et al., 2024), while surges in organic search may indicate heightened uncertainty and information-seeking behaviour (Spyridou & Danezis, 2022). These platform-specific signals—such as bounce rate, session depth and traffic source—function as indirect indicators of trust, uncertainty and loyalty during crisis periods (Celestin et al., 2024). While they may not capture emotional states explicitly, they reflect how users cognitively and emotionally process safety cues (Moreschini et al., 2025; Czarnecki et al., 2023).

Effective crisis communication plays a central role in shaping digital behaviour during uncertainty. Šerić et al. (2025) emphasize that timely, emotionally intelligent messaging can strengthen brand relationships. Çakar (2021) similarly argues that digital familiarity reduces fear-based responses, such as tourophobia. Extending this, Chen et al. (2025) highlight that tourist decisions depend not only on personal risk tolerance, but also on the clarity and consistency of information across digital channels.

As users move through stages of risk evaluation (before, during and after a disaster), their engagement with digital channels shifts accordingly. Branded search and direct traffic often increase during high-risk periods, as consumers seek reassurance from familiar providers (Erdmann et al., 2022), while social media serves as a space for emotional validation and peer reassurance (Steinert, 2021). These patterns underscore that digital interaction is inherently adaptive, shaped by perceptions of environmental stability.

Although this study does not measure consumer sentiment directly, it draws on the risk familiarity effect as a conceptual framework for interpreting behavioural signals observed across digital platforms. These digitally observable behaviours—such as changes in direct traffic or bounce rates—reflect the emotional and cognitive recalibration that occurs before, during and after crisis signals, as users move between uncertainty, reaction and recovery. By analysing cross-channel engagement, this study offers strategic insights into how hospitality brands can monitor and respond to shifting trust and attention dynamics throughout crisis phases.

2.3. Multichannel Digital Marketing in Hospitality: Behavioural Patterns in Crisis Response

Digital marketing has become indispensable to the hospitality sector, shaping how consumers discover, evaluate and engage with tourism-related services (Kurdi et al., 2022). In an industry defined by intangible offerings, high emotional investment and the importance of trust, digital channels serve not only as transactional gateways but also as experiential and psychological touchpoints (Ciasullo et al., 2024). Especially during periods of disruption, these channels become central to how consumers process uncertainty, assess safety and choose whether to engage with a destination or brand (Ketter & Avraham, 2021). Recent literature has emphasized the need for multichannel integration that goes beyond visibility to strategically build brand trust across platforms. Vaishnav and Ray (2022) show that digital marketing success during uncertainty depends on the firm’s ability to coordinate customer engagement touchpoints across owned, earned and paid media. Their thematic review of 25 years of multichannel marketing research highlights how emotionally responsive communication is especially crucial during periods of heightened consumer anxiety—like in disaster contexts.

The 2023 Türkiye earthquakes showed how hospitality brands and civic actors jointly mobilized digital platforms for emotional reassurance and public coordination—reinforcing the trust-building function of integrated media. (Ayyıldız & Tümbek Tekeoğlu, 2024). This aligns with findings by Zhang et al. (2024), who emphasize that digital platforms play a vital role in disaster recovery and resilience by enabling stakeholder collaboration, community engagement and adaptive communication planning in tourism settings. Gkeredakis et al. (2021) further argue that crises act as both disruptions and catalysts—prompting the rapid reconfiguration of digital technologies to maintain continuity, foster innovation and surface new ethical and operational challenges in service delivery. Collectively, these insights illustrate the broader role of digital channels during disasters as instruments of trust-building and public reassurance, especially when traditional infrastructures are strained.

Hospitality brands, particularly those operating in disaster-prone environments, must navigate these platforms strategically. As Belias et al. (2023) highlight, social media reviews can significantly influence a hotel’s brand image in times of crisis. Platforms like TripAdvisor or Instagram serve as dynamic arenas of public sentiment, where trust can be reinforced or damaged in real time. Manningham et al. (2024) further emphasize that the consistent and strategic use of social media, rather than follower count alone, is what drives engagement and brand resilience. Similarly, Sakas et al. (2022c) demonstrate that multichannel alignment—particularly between organic search and social media channels—can enhance crisis resilience by creating more predictable patterns of direct user return. Their findings suggest that higher organic engagement before a crisis leads to lower reliance on high-cost, paid traffic afterwards, as user trust carries forward.

Importantly, digital channels in hospitality are not isolated touchpoints but part of a nonlinear, emotionally driven consumer journey (Hu & Olivieri, 2020). While branded platforms shape loyalty, third-party OTAs also act as trust mediators, especially during crises, when travellers seek platform-level reassurance over brand-specific promises (Kumar, 2024). A potential guest may begin their search on an OTA, move to Google for price comparison or booking incentives and return later via branded search after engaging with user-generated reviews. During disasters, this multistep journey becomes even more emotionally charged, as users oscillate between rational evaluation and the need for reassurance (Yada et al., 2024). Each platform corresponds to a different phase of consumer trust—discovery, validation or commitment—requiring targeted, emotionally attuned messaging across each touchpoint. Singgalen (2023) reinforces this model through a sentiment analysis of YouTube travel vlogs, showing that emotionally positive content during crises drives engagement and re-engagement across platforms. The study also confirms that sentiment in user-generated content becomes a behavioural driver for search behaviour and paid ad responsiveness, providing practical evidence for integrated marketing under stress.

Understanding the impact of past behaviours on future engagement is critical in post-crisis recovery strategies. Previous studies have shown that brand familiarity and organic search activity play a pivotal role in driving direct traffic in crisis conditions (Sakas et al., 2022c). Given the emotional impact of a disaster and its effect on consumer trust, it is important to understand how previous organic search behaviours influence post-crisis engagement. Building on this behavioural logic, the next hypothesis evaluates how cumulative channel exposure predicts brand re-engagement via direct traffic. Hence, the following hypothesis is proposed:

H₃: 

Direct traffic (3 months after) is significantly influenced by prior organic search activity and concurrent paid search efforts.

This hypothesis tests RQ3, which explores how historical and concurrent digital marketing activities contribute to sustained customer interest and brand loyalty in the aftermath of a disruption. Direct traffic is commonly used as a behavioural proxy for brand familiarity and user loyalty (Sakas et al., 2022b; Önder & Berbekova, 2022). Prior organic search activity can imprint brand associations, while concurrent paid search reinforces visibility, particularly when trust needs to be re-established. If a significant relationship is found, it would suggest a synergistic effect across time and channels—supporting the theory that sustained exposure through earned and paid media contributes to digital loyalty (Filippou et al., 2024; Halkiopoulos & Papadopoulos, 2022).

Hospitality marketing is thus inherently multichannel, with different platforms corresponding to varying levels of user trust, engagement and emotional readiness (Ngo et al., 2023). Organic search typically reflects early-stage information-seeking; paid search often captures users with stronger booking intent, especially when paired with risk-reduction incentives (Sakas et al., 2022c). Direct traffic signals brand familiarity and consumer trust, particularly under conditions of uncertainty, while social media facilitates emotional validation and peer feedback on brand tone and responsiveness (Seneviratne et al., 2024).

While the importance of organic traffic has been established, its role in predicting paid traffic costs in the context of a crisis remains underexplored. Previous literature has established the predictive power of organic traffic for paid traffic effectiveness in stable conditions (Erdmann et al., 2022). However, this dynamic has not been adequately examined in post-crisis contexts. Given the potential volatility of user behaviour during and after a disaster, understanding how pre-crisis organic signals influence future marketing expenditures is crucial for refining crisis recovery strategies. Building on this evidence, we hypothesize that paid traffic cost after a disaster will be significantly predicted by organic traffic levels both before and after the crisis announcement. Hence, the following hypothesis is proposed:

H₁: 

Paid traffic cost (3 months after) is significantly predicted by organic traffic levels from both 3 months before and 3 months after the announcement.

This hypothesis directly addresses RQ1, which explores how different types of online traffic influence future advertising effectiveness. It builds on performance marketing literature that highlights the predictive role of organic traffic as a signal of user trust and brand familiarity (Erdmann et al., 2022). As W. Wang et al. (2019) and Sakas et al. (2022c) suggest, robust organic engagement may reduce reliance on paid advertising by sustaining consumer interest without direct financial input. By testing this relationship across pre- and post-disaster periods, this hypothesis captures how early organic signals influence future digital marketing expenditures during crisis recovery.

Halkiopoulos and Papadopoulos (2022) argue that digital platforms differ not only in function but in the emotional and cognitive processing they demand. Social media amplifies emotional cues and peer influence (Nesi et al., 2023), while branded search and direct visits reflect deeper trust and loyalty consolidation (Simonov & Hill, 2021). These insights underscore that digital channels function as behavioural proxies—offering clues into how hospitality consumers process risk, trust and decision-making in volatile environments. Thus, successful digital strategies in crisis contexts rely not just on maintaining visibility, but on delivering consistent, emotionally intelligent messaging tailored to each platform’s role in the consumer journey.

Building on the evidence from the discussion on multichannel engagement, we hypothesize that paid social traffic after a crisis will be significantly predicted by both direct traffic and prior paid search activity. Hence, the following hypothesis is proposed:

H₄: 

Paid social traffic (3 months after) is significantly predicted by current direct traffic and previous paid search activity.

This hypothesis directly tests RQ4, which explores how cross-channel interactions shape targeted advertising outcomes in crisis contexts. The underlying assumption is that digital users exposed to paid search ads may initially engage via direct visits and later be captured by paid social campaigns, especially when familiarity and trust are already established (Seneviratne et al., 2024; Pai et al., 2025). This reflects the behavioural sequence common in multichannel marketing ecosystems—where exposure timing and channel transitions influence performance. Paid social performance is particularly sensitive to prior engagement, and this hypothesis evaluates whether the combination of search and direct traffic creates a retargetable, conversion-prone audience during the post-disaster recovery period (Halkiopoulos & Papadopoulos, 2022; Chan et al., 2021).

While these patterns clarify how users arrive at digital platforms, they do not fully capture how emotional alignment and trust manifest during user interactions. To assess that behavioural depth, the next section turns to web analytics as diagnostic tools for measuring engagement quality and real-time sentiment shifts.

2.4. Web Analytics and Digital Engagement in Disaster Contexts

In the hospitality industry—where experiences are subjective and decisions are emotionally charged—understanding how users behave on digital platforms is as critical as knowing how they arrive there. Web analytics tools offer a robust framework for interpreting these behaviours, providing real-time visibility into consumer sentiment, intent and trust levels during periods of both stability and disruption (Park, 2021). Metrics such as bounce rate, session duration and pages per session do more than track performance; they serve as behavioural proxies that reflect how users psychologically and emotionally interact with hospitality brands—particularly in the context of risk (Celestin et al., 2024). As shown by Sakas et al. (2022c), bounce rate and session depth during crises reflect real-time user emotion and brand alignment, particularly when interpreted alongside organic and paid traffic sources. Their study highlights how brands can use these indicators not only for performance measurement but as early-warning signals of disengagement due to trust erosion or message fatigue.

These engagement metrics apply across a range of digital media types within the hospitality sector. On websites and booking engines, indicators such as bounce rate, session duration, pages per session and conversion rate offer insights into user confidence and content clarity, especially under stress (Önder & Berbekova, 2022; Celestin et al., 2024; Muralidhar & Lakkanna, 2024). In social media environments, behavioural signals like likes, shares, comment volume and click-through rates reflect emotional resonance, peer influence and the virality of safety-related messaging (Ho et al., 2022; Pocchiari et al., 2024; Acikara et al., 2023). Mobile apps and loyalty platforms provide tap-through rates and session frequency data, which reflect habitual use, brand affinity and consumer anxiety when navigating disruption (Seneviratne et al., 2024; W. Wang et al., 2019). For paid search and display ads, metrics such as cost-per-click (CPC), impression share and conversion rate serve as indicators of message–target fit and disaster-period responsiveness (Filippou et al., 2024; Vitkauskaitė & Mayeur, 2025). When viewed across these media types, such indicators help decode how trust, intent and user engagement fluctuate during a crisis—enabling a more precise understanding of platform-specific dynamics (Gour et al., 2021; Mariani & Baggio, 2022; Mousavian et al., 2023). Karimiziarani (2023) extends this logic to social media analytics during disasters, arguing that digital response frameworks must account for real-time user discourse as a sentiment barometer, not merely for marketing but for strategic communication. Bounce rate, for instance, is often interpreted as a signal of dissonance between user expectations and perceived value (Muralidhar & Lakkanna, 2024). In disaster contexts, an elevated bounce rate may indicate that users are searching for reassurance but are confronted with unclear, inconsistent or unconvincing messaging. Conversely, lower bounce rates during a disruption can reflect perceived reliability or emotional alignment with the brand’s digital tone, which in hospitality is often conveyed through images, guest policies and crisis communications. These interpretations align with tourism analytics studies that link consumer disengagement to emotionally dissonant digital environments during recovery efforts (Chan et al., 2021). The combination of real-time behavioural data and emotional resonance analysis forms the core of a crisis-aware analytics strategy—allowing hospitality firms to shift from reactive responses to pre-emptive engagement. Future studies may benefit from extending this framework with AI-based sentiment extraction, as suggested by Mousavian et al. (2023), to detect behavioural anomalies linked to disaster perception.

Building on this interpretation, we hypothesize that pre-crisis bounce rates can be explained by user engagement behaviours that reflect either cognitive overload or clarity in communication. Specifically, deeper engagement, as measured by pages per visit, may reduce bounce rates, while poorly targeted paid search activity could elevate them due to message mismatch. Hence, the following hypothesis is proposed:

H₂: 

Bounce rate in the period preceding the disaster announcement is significantly affected by the number of pages per visit and paid search activity during the same period.

While RQ2 emphasizes behaviour during disruption, this hypothesis tests how pre-crisis engagement patterns reflect baseline user expectations. According to dual-process theory, bounce rate captures cognitive–emotional tension (Blömker & Albrecht, 2024), and pages per visit reflects effortful, reassurance-seeking behaviour (Önder & Berbekova, 2022). Paid search traffic can either improve engagement—if messages align with user intent—or drive disengagement if poorly targeted (Chan et al., 2021). Testing this relationship before the crisis allows for benchmarking how engagement shifts during the crisis, addressing RQ2 by implication.

Similarly, session depth and average duration can serve as indicators of cognitive and emotional engagement. A user navigating multiple booking-related pages, reading reviews and revisiting health and safety protocols likely reflects deeper information processing and trust development (Önder & Berbekova, 2022). These metrics align with dual-process theories in consumer psychology, which suggest that during periods of elevated risk, users shift from intuitive browsing to systematic information gathering (Blömker & Albrecht, 2024). Several studies emphasize that hospitality users, under uncertainty, tend to spend more time online, seeking reassurance through detailed exploration of policies, guest experiences and crisis responses (Sann et al., 2022).

These engagement metrics also serve as inputs for predictive and adaptive strategy (Mathur et al., 2024; Upadhyaya, 2024). When monitored over time—especially around disaster announcements—they enable hospitality providers to anticipate changes in consumer behaviour and adjust campaigns, offers and messaging in near real time. For instance, a sudden spike in organic traffic coupled with high bounce rates and low session depth may suggest anxiety-driven browsing with low conversion readiness. On the other hand, rising direct traffic alongside longer session duration might indicate renewed consumer confidence and readiness to book, particularly for trusted providers (Sakas et al., 2022b; Moreschini et al., 2025). Such patterns are increasingly embedded in data-driven marketing models designed to guide dynamic decision-making in volatile tourism environments (Gour et al., 2021; Mousavian et al., 2023).

From a strategic perspective, web analytics offer a pathway to shift from reactive brand response to proactive crisis engagement. Rather than viewing user metrics as static outcomes, firms can interpret them as feedback loops that reflect shifts in consumer trust, attention and sentiment (Mariani & Baggio, 2022). When integrated with sentiment analysis and customer segmentation tools, these insights can inform not only communication tone and timing but also content structure and cross-platform consistency (Gour et al., 2021). Advanced analytics systems are now capable of detecting early warning signs of customer disengagement or message fatigue, empowering firms to course-correct before reputational or financial damage occurs (Garg et al., 2025).

This study incorporates these behavioural signals into a broader modelling framework to investigate how digital interaction quality evolves in disaster settings. While a broad range of digital engagement indicators exists across various media types, the present study focuses specifically on a targeted subset of metrics—namely organic traffic, paid search, direct traffic, paid social, bounce rate and pages per visit. These were chosen not only for their strategic relevance to digital behaviour during crises, but also because they were the only indicators with consistent and reliable data across all observed hotel websites during the crisis period.

Each of these indicators reflects a distinct behavioural response:

• Organic traffic signals early-stage interest and information-seeking, often under conditions of uncertainty.
• Paid search traffic captures user behaviour influenced by targeted messaging, which is particularly relevant during efforts to reinforce trust and visibility.
• Direct traffic is commonly interpreted as an indicator of brand familiarity, loyalty and trust—which is especially critical in post-disaster scenarios.
• Paid social traffic reflects emotionally responsive engagement, which is often shaped by prior exposure through other channels.
• Bounce rate serves as a proxy for dissonance or unmet expectations, being potentially linked to anxiety or a lack of reassurance in the digital environment.
• Pages per visit reflects deeper information processing and higher engagement, particularly when users seek clarity or validation during unstable periods.

By narrowing the analysis to these indicators, this study provides a focused, empirically grounded perspective on how consumer engagement shifts during a disaster event. These metrics serve as both diagnostic tools and behavioural proxies—allowing hospitality brands to monitor shifts in sentiment and decision-making in real-time, and to optimize digital strategies accordingly.

2.5. Conceptual Framework: Digital Consumer Behaviour in Disaster Tourism

In response to growing calls for integrative frameworks in tourism crisis research (Ritchie & Jiang, 2021; Duan et al., 2021), this study proposes a conceptual model that synthesizes insights from behavioural science, digital marketing and tourism risk theory to explain how consumers interact with hospitality brands across digital channels during disaster contexts. This framework draws from three foundational concepts:

• Risk familiarity effect: As outlined by Sakas et al. (2022b), consumer behaviour evolves over time as users become accustomed to disaster-related uncertainty, initially seeking trust signals, then exploring alternatives, and ultimately returning to familiar brands when threat perception resurfaces.
• Dual-process theory: During high-stress periods, users shift from intuitive, fast processing (e.g., bounce behaviour) to deliberate, deeper engagement (e.g., increased session depth), as described by Blömker and Albrecht (2024).
• Multichannel digital engagement: Consumers interact with multiple digital touchpoints—organic search, paid ads, social media and direct traffic—each reflecting different stages of trust, anxiety and decision-making (Filippou et al., 2024; Vaishnav & Ray, 2022).

The model includes two primary layers:

• Traffic sources (inputs): Organic search, paid search, direct visits and paid social.
• Engagement signals (outputs): Bounce rate and pages per visit, which function as proxies for emotional–cognitive alignment and trust evolution.

The temporal dimension is central to the framework, with disaster-related signals dividing behaviour into three stages:

• Pre-disaster: High organic interest, early uncertainty.
• Crisis period: Shifts in bounce rate, urgent searches, high social engagement.
• Post-disaster: Re-engagement via direct and branded traffic; bounce rates stabilize.

Figure 1 visualizes this framework, showing how traffic inputs and behavioural outputs evolve in response to disaster signals. The flow from user uncertainty to engagement depth reflects the interaction of emotional, informational and temporal factors that influence consumer behaviour in digital hospitality environments. This model provides an integrated view of how consumer trust is built, challenged and restored across distinct traffic channels during disaster phases. It sets the stage for our empirical investigation, which follows in Section 4 through an integrated analysis of traffic sources, behavioural metrics and advertising performance.

Figure 2 illustrates the Google Trends data for the three months preceding and the three months succeeding the disclosure of the prospective disaster.

3. Case Description: The Santorini Earthquake

The Santorini Earthquake occurred in February 2025, registering a magnitude of 6.7 on the Richter scale and resulting in significant infrastructural damage and disruption to the island’s tourism sector. On 5 February, emergency services were deployed and on 6 February, the Greek government formally declared a state of emergency, which remained in effect until 3 March 2025. Communication escalated rapidly: early alerts were issued by the Irish National Seismic Network (INSN, 2025) and the European Mediterranean Seismological Centre (EMSC, 2025), while UNESCO (2025) raised alarms about potential threats to the island’s cultural heritage. National and international media outlets, including Euronews (2025), provided real-time coverage, broadcasting civil protection directives and safety advisories to both residents and tourists. These overlapping institutional and media-driven communications played a pivotal role in shaping consumer sentiment, digital engagement and tourism-related decision-making during the crisis. This event served as the primary crisis signal around which this study is centred.

Prior to the earthquake, seismic monitoring agencies and local government authorities issued precautionary alerts indicating elevated seismic activity. These communications were disseminated via official government websites, hospitality sector digital platforms and social media channels, emphasizing preparedness measures and flexible travel options.

During the disaster event, communication efforts intensified, with continuous updates provided through emergency management websites, local news media and social media platforms. Messaging focused on safety instructions, evacuation procedures and real-time status reports concerning infrastructure and service availability.

In the aftermath of the earthquake, communication emphasis transitioned toward recovery efforts. Hospitality providers updated digital touchpoints with reassurances regarding safety, flexible cancellation policies and promotional campaigns aimed at restoring consumer confidence and re-engaging visitors.

This phased communication strategy played a critical role in shaping observed digital consumer behaviours, influencing patterns of organic, paid, direct and social traffic throughout the study period. Understanding the temporal alignment of these communications with behavioural signals contributes to interpreting the evolution of digital engagement in crisis contexts.

4. Materials and Methods

The temporal and causal relationships among several critical digital marketing key performance indicators (KPIs) are the subject of this study, which employs a rigorous quantitative research methodology. The primary objective is to comprehend the extent to which fluctuations in specific variables—including organic search traffic, paid search traffic, return rate and direct traffic—can predict variations in paid traffic expenditure and the efficacy of paid social media campaigns (Singh, 2024). Data were collected over two discrete three-month periods, a pre-disaster phase from November 2024 to January 2025 and a post-disaster phase from February to April 2025, in order to capture the impact of an external event on these relationships (

Table 1. Web analytics KPIs.

Web Analytics KPIs	Description of the Web Analytics KPIs
Organic Traffic	The quantity of website visitors who access the site via organic search engine results. This KPI indicates the amount of traffic generated by search engine optimization (SEO) initiatives. It reflects a website’s ranking in search engine results pages (SERPs) for pertinent keywords without paid advertising (Jones, 2024).
Paid Search	Traffic derived from adverts featured in search engine results (e.g., Google Ads). This KPI monitors the quantity of visitors who engage with sponsored listings or pay-per-click (PPC) advertisements. It evaluates the efficacy of a brand’s paid search initiatives in drawing users (Blake et al., 2015).
Bounce Rate	The proportion of visitors who exit the website after seeing a single page. An elevated bounce rate may signify inadequate user engagement or material that lacks relevance. This KPI evaluates the quality of landing pages and the user experience (X. Wang et al., 2021).
Pages per Visit	The mean quantity of pages a user examines in one session on the website. This measure assesses visitor involvement. Elevated scores often indicate that consumers perceive the content as engaging and are navigating the site extensively (Sakas et al., 2022c).
Paid Traffic Cost	The whole financial outlay on paid digital advertising to generate website visitors. This KPI indicates the budget designated and expended on paid media initiatives, including search advertisements, display advertisements and social media advertising. It is essential for assessing return on advertising expenditure (Sakas & Reklitis, 2021).
Organic Search	The procedure by which consumers discover a website via organic search engine results. Frequently used synonymously with organic traffic, it more precisely denotes the channel or source within analytics platforms. It is propelled by SEO strategies designed to enhance visibility in search outcomes (Jones, 2024).
Direct Traffic	Traffic originating from visitors who enter a website’s URL directly into their browser or access it through bookmarks. This KPI reflects brand recognition and allegiance. It indicates that people possess familiarity with the brand or regularly revisit the site independently of intermediaries such as search engines or social media platforms (Go, 2024).
Paid Social Media Traffic	Website traffic derived from sponsored adverts on social media sites (e.g., Facebook Ads, Instagram Ads, LinkedIn Sponsored Content). This KPI assesses the reach and efficacy of social media advertising strategies in generating site visitors. It enables marketers to evaluate the effectiveness of their social advertising expenditures (Beveridge, 2024).

). This timeline corresponds with the Santorini Earthquake swarm, which commenced on 1 February 2025, and was subsequently followed by the declaration of a state of emergency on 6 February 2025. The research concentrated on the ten most prestigious luxury hotels in Santorini, documenting performance metrics that were pertinent to their digital marketing initiatives (Condé Nast Traveller, 2025).

The data collection process employed sophisticated digital analytics platforms, such as SEMRUSH, to monitor organic and paid search metrics. Additionally, paid advertising interfaces were implemented to provide detailed financial and campaign-specific information. The key performance indicators (KPIs) that were selected were strategically chosen based on their prominence in digital marketing performance assessment. These metrics included user engagement metrics (such as pages per visit and bounce rate), organic search, paid search and direct traffic, as well as cost-related financial indicators (cost of paid traffic). The data structure enabled a thorough temporal comparison between the two timeframes, which enabled the analysis of potential disruptions and adaptive responses that were precipitated by the earthquake event (Hameed et al., 2025).

In order to guarantee both statistical robustness and the capacity to capture intricate causal dynamics, the analytical approach was composed of three interconnected phases. Initially, correlation analysis was implemented to ascertain and quantify the intensity and direction of linear relationships between independent variables and their corresponding dependent variables. This preliminary stage was crucial for the identification of underlying patterns and associations, which in turn provided foundational insights into the dataset and guided subsequent modelling decisions. Before employing predictive modelling, the correlation coefficients served as a diagnostic tool by indicating which variable pairings were most likely to influence one another and at what level.

Secondly, each hypothesis was formally tested using multiple linear regression analysis in this study. One dependent variable and two independent variables were used to define each hypothesis, which reflected specific causal assumptions regarding digital marketing interactions. The regression models assessed the individual impact of each independent variable on the dependent variable while accounting for other variables. We evaluated the model’s performance and validity using a variety of critical metrics, including the coefficient of determination (R-squared), F-statistics for overall model fit and p-values to determine the statistical significance of predictors. This step enabled the conduct of rigorous hypothesis testing, which elucidated the degree to which each factor predicted marketing outcomes and whether these relationships varied between the pre- and post-disaster periods.

Third, fuzzy cognitive mapping (FCM) was incorporated into the analytical framework to capture the complexity of interdependent relationships and feedback cycles that are typical of digital marketing systems. FCM is capable of representing nonlinear interactions and the propagation of influences across multiple variables in a networked structure, in contrast to traditional linear models. FCM was developed by integrating empirical data with domain knowledge by utilizing expert-defined weights that were informed by correlation and regression findings. This method offered a dynamic visualization of the potential impact of changes in one metric, such as paid search traffic, on others, such as return rate or paid social media performance, over time. The model could provide more comprehensive interpretative insights than linear cause-and-effect assumptions by addressing the uncertainty and ambiguity that are inherent in marketing data through the use of fuzzy logic (Salmeron & Arévalo, 2025).

This study utilized correlation analysis, multiple regression modelling and fuzzy cognitive mapping (FCM) to investigate the interrelationships among the chosen digital marketing KPIs. Correlation and regression techniques facilitated the detection of significant linear associations and the strength of predictive linkages, whereas FCM was employed to model more intricate, feedback-driven and nonlinear dependencies that frequently characterize digital marketing environments. In contrast to conventional statistical models that presume unidirectional linear causality, FCM offers a semi-quantitative, expert-informed methodology for comprehending how alterations in one variable might disseminate throughout a system to affect others (Kosko, 1986; E. Papageorgiou & Kontogianni, 2012). This work involved the construction of FCM utilizing expert input, which was enhanced by empirical data patterns, including weights obtained from correlation and regression analyses.

This hybrid methodology facilitates the amalgamation of data-driven evidence and domain-specific expertise to model system behaviour and evaluate the magnitude and direction of causal relationships across interconnected KPIs (Salmeron & Froelich, 2016; Stylios et al., 2008). Although FCM does not yield statistical causality in the strictest positivist sense, it is becoming acknowledged as a viable tool for causal inference in complex systems where conventional assumptions on variable independence and linearity are insufficient. Its efficacy resides in its ability to capture feedback loops, dynamic interactions and conditional correlations sometimes overlooked by traditional statistical methods (E. I. Papageorgiou & Salmeron, 2013). Consequently, FCM was considered a suitable and theoretically sound approach for discerning causal-like interactions within the complex framework of digital marketing success measures during crisis situations.

This study primarily utilized objective behavioural measures from the SEMrush web analytics platform—such as organic search volume, bounce rate and direct traffic—rendering typical psychometric validations, including convergent and discriminant validity, inapplicable. These indicators are not perceptual or latent variables obtained from respondent-based instruments; rather, they are system-generated, consistently operationalized metrics that reflect actual user behaviour. This methodology conforms to established norms in digital analytics research, wherein instrument dependability is presumed due to platform architecture and uniform automated tracking (Kitchin, 2017; Wedel & Kannan, 2016).

Additionally, correlation and regression analyses were utilized not as independent ways to assess causation, but as supplementary instruments to quantify linear relationships and facilitate the development of fuzzy cognitive maps (FCM). The use of FCM facilitates the modelling of nonlinear, bidirectional and dynamic interactions, providing a more sophisticated understanding of causal effect inside intricate digital systems (E. I. Papageorgiou & Salmeron, 2013; Salmeron & Arévalo, 2025).

In conclusion, the triangulated methodology, which employs correlation analysis, multiple linear regression and fuzzy cognitive mapping, establishes a robust framework for comprehending both direct linear dependencies and intricate systemic feedback mechanisms in digital marketing. This comprehensive approach allows for a nuanced analysis of the impact of historical and current performance metrics on future outcomes, providing marketers with actionable insights to optimize multichannel strategies in volatile environments (Nasiopoulos et al., 2023).

5. Results

5.1. Statistical Analysis

The results are derived from the extensive data of ten tourism company websites (Condé Nast Traveller, 2025).

Table 2. Coefficients between the examined metrics for H1.

Correlations	Organic Traffic (3 Months Before)	Organic Traffic (3 Months After)	Paid Traffic Cost (3 Months After)
Organic traffic (3 months before)	1
Organic traffic (3 months after)	0.727 **	1
Paid traffic cost (3 months after)	0.836 **	0.667 **	1

** Correlation is significant at the 0.01 level (1-tailed).

and

Table 3. Regression analysis for H1.

Variables	Standardized Coefficient	R2	F	p-Value
Constant (Paid traffic cost (3 months after))	-	0.707	53.041	0.075
Organic traffic (3 months before)	0.746			<0.001
Organic traffic (3 months after)	0.124			0.303

display the correlation and regression analyses for H1. (H1) shows a positive correlation with ρ = 0.727 ** between the organic traffic (3 months before) metric and the organic traffic (3 months after) metric. Furthermore, significant correlations were found between the paid traffic cost (3 months after) with and the organic traffic (3 months before) metric and the organic traffic (3 months after) metric, with ρ = 0.836 ** and ρ = 0.667 **, accordingly. The results indicate that increases in organic traffic prior to a campaign are strong indicators of future engagement, whereas deliberate investment in paid traffic enhances this momentum, revealing a dynamic synergy between organic growth and paid initiatives in fostering sustained interest from consumers. The regression for H1 is illustrated in Table 2. Regression analyses with p-values of less than 0.05 are identified as significant. In this nonsignificant regression, for every 1% rise in paid traffic cost, the organic traffic (3 months before) metric and the organic traffic (3 months after) metric increase by 74.6% and 12.4%, accordingly.

Table 4. Coefficients between the examined metrics for H2.

Correlations	Bounce Rate (3 Months Before)	Pages per Visit (3 Months After)	Paid Search (3 Months After)
Bounce rate (3 months before)	1
Pages per visit (3 months after)	0.357 *	1
Paid search (3 months after)	0.096	0.222	1

* Correlation is significant at the 0.05 level (2-tailed).

and

Table 5. Regression analysis for H2.

Variables	Standardized Coefficient	R2	F	p-Value
Constant (Bounce rate (3 months before))	-	0.128	3.216	<0.001
Pages per visit (3 months after)	0.353			0.019
Paid search (3 months after)	0.018			0.900

present the correlation analysis and regression analysis for H2. (H2) shows a positive correlation with ρ = 0.931 ** between the bounce rate metric and the pages per visit metric. Furthermore, nonsignificant correlations were found between the paid search with bounce rate and pages per visit metrics. The results reveal a significant correlation between bounce rate and pages per visit, indicating that user engagement patterns are closely intertwined following a disaster, although paid search initiatives seem to exert no impact on the depth of user interaction with site content during this timeframe. The regression model for H2 is presented in Table 4. Regression analyses with p-values of less than 0.05 are identified as significant. In this regression, for every 1% rise in paid search, the pages per visit increases by 35.3% and the bounce rate increases by 1.8%.

Table 6. Coefficients between the examined metrics for H3.

Correlations	Direct Traffic (3 Months After)	Organic Search (3 Months Before)	Paid Search (3 Months After)
Direct traffic (3 months after)	1
Organic search (3 months before)	−0.441 **	1
Paid search (3 months after)	−0.561 **	0.132	1

** Correlation is significant at the 0.01 level (1-tailed).

and

Table 7. Regression analysis for H3.

Variables	Standardized Coefficient	R2	F	p-Value
Constant (Direct traffic (3 months after))	-	0.452	18.143	<0.001
Organic search (3 months before)	−0.373			0.002
Paid search (3 months after)	−0.512			<0.001

present the correlation analysis and regression analysis for H3. H3 shows a positive correlation with ρ = 0.132 between the organic search (3 months before) activity and concurrent paid search efforts (3 months after). Furthermore, significant negative correlations were found between the direct traffic (3 months after) with the organic search (3 months before) activity and concurrent paid search efforts (3 months after). The findings indicate that robust organic search activity prior to a disaster can subsequently result in heightened paid search investment, while concurrently causing a decrease in direct traffic—underscoring a strategic transition in user acquisition channels and marketing emphasis in the post-disaster hospitality sector. The regression model for H3 is presented in Table 6. Regression analyses with p-values of less than 0.05 are identified as significant. In this significant regression, for every 1% rise in direct traffic, the organic search decreases by 37.3% and the paid search decreases by 51.2%.

Table 8. Coefficients between the examined metrics for H4.

Correlations	Paid Social Traffic (3 Months After)	Direct Traffic (3 Months After)	Paid Search (3 Months Before)
Paid social traffic (3 months after)	1
Direct traffic (3 months after)	0.401 **	1
Paid search (3 months before)	0.624 **	0.643 **	1

** Correlation is significant at the 0.01 level (1-tailed).

and

Table 9. Regression analysis for H1.

Variables	Standardized Coefficient	R2	F	p-Value
Constant (Paid social traffic (3 months after))	-	0.390	14.043	<0.001
Direct traffic (3 months after)	0.590			<0.001
Paid search (3 months before)	0.624			<0.001

present the correlation analysis and regression analysis for H4. H4 shows a positive correlation, with ρ = 0.931 ** between the paid social traffic (3 months after) and the current direct traffic (3 months after) metrics predicted to be significant. Furthermore, significant correlations were found between paid social traffic (3 months after) and direct traffic (3 months after) and previous paid search (3 months before). The results indicate that future paid social traffic is significantly influenced by current direct traffic levels, along with the impact of prior paid search initiatives—highlighting the cumulative effect of multichannel interaction on post-disaster digital marketing results in the hospitality industry. The regression model for H4 is illustrated in Table 8. Regression analyses with p-values of less than 0.05 are identified as significant. In this significant regression, for every 1% rise in paid social traffic, direct traffic (3 months after) and previous paid search (3 months before) increase by 59% and 62.4%, accordingly.

The overall analysis shows a dynamic interaction between digital marketing platforms and consumer behaviour in the hospitality industry subsequent to disaster announcements. Significant relationships were identified between pre- and post-disaster organic traffic, demonstrating enduring user interest over time. Pre-disaster organic search activity significantly influenced future sponsored search initiatives, while ironically corresponding with less direct traffic. Moreover, engagement measures like bounce rate and pages per visit exhibited a strong correlation, whereas paid search exerted less impact on these behaviours. Sponsored social traffic following a tragedy was highly influenced by both current direct traffic and prior sponsored search activity, underscoring the cumulative effect of integrated digital tactics over time.

5.2. Fuzzy Cognitive Map

The fuzzy cognitive map (FCM) was constructed based on the statistical studies presented in preceding sections, encapsulating the intricate dynamics between digital marketing initiatives and consumer behaviour in the context of disaster announcements. An FCM depicts the fundamental framework of a system and the intensity of directional interactions among variables (spanning from −1 to 1). This is particularly significant in the domain of big data, where unstructured information needs to be converted into useful insights. The FCM facilitates the development of data-driven marketing strategies by visualizing cause-and-effect links. In Figure 3, more robust correlations are depicted by thicker lines. The incorporation of standardized protocols and empirical evidence into the FCM was crucial for enhancing model precision and clarity. The process commenced with the “Clarification of objectives”, which, in this study, entailed the formulation of three strategic scenarios to aid hospitality enterprises in effectively addressing disaster situations—by optimizing paid advertising, augmenting social media engagement and fortifying digital brand presence. The following steps involved identifying data sources and adjusting the model to guarantee coherence and applicability to real-world scenarios (Salmeron & Arévalo, 2025).

The digital brand name optimization scenario was created to analyse anticipated changes in essential web analytics performance metrics after disaster announcements, emphasizing the improvement of user engagement and the restoration of confidence in the hospitality industry. This scenario endorses the FCM paradigm by emulating strategic initiatives designed to affect customer behaviour on digital channels. Figure 4 illustrates that to effectively enhance both organic and direct traffic following a disaster, hospitality websites must prioritize the development of compelling content that conveys safety and reassurance, resulting in a 1% increase in pages per visit. Moreover, a strategic investment in Google Ads, namely a 9% rise in sponsored search, bolsters traffic recovery. Nonetheless, social media advertising seems to be less effective in this setting, exhibiting a 7% decrease in paid social performance. These findings highlight the necessity of synchronizing digital communication and platform-specific tactics with changing customer sentiment during post-disaster recovery initiatives.

6. Discussion

This study investigated the impact of several online analytics metrics on digital consumer behaviour during a real-time crisis, utilizing the Santorini Earthquake as a contextual example. The results indicate that marketing performance in uncertain times is influenced not by singular indicators, but by the dynamic interaction of exposure, engagement and user familiarity across many digital platforms. The identified correlations among variables of H1 correspond with recognized theories in crisis communication and digital behaviour. The notable link between previous organic traffic and diminished paid traffic costs substantiates the risk familiarity effect (Sakas et al., 2022b). In periods of uncertainty, consumers typically depend on familiar or trusted sources. Our study indicates that robust organic traffic before the crisis diminished dependence on paid advertising afterward, supporting the idea that earlier exposure cultivates trust, hence decreasing abandonment and promoting re-engagement under heightened risk levels. This is further evidenced by the rise in direct traffic post-disaster, particularly among users who had previously encountered the brand through organic or sponsored search—clear validation of the familiarity-driven behaviour outlined in prior studies.

Furthermore, as for H2, the data corroborate essential aspects of dual-process decision theory (Blömker & Albrecht, 2024), which differentiates between intuitive (heuristic) and intentional (systematic) decision-making. Increased bounce rates before the crisis indicate emotional evasion or cursory browsing, aligning with heuristic processing during periods of stress. Conversely, the rise in pages per visit and session depth following the crisis indicates more intentional, goal-directed engagement, aligning with methodical processing. These developments illustrate the cognitive transformations suggested by Önder and Berbekova (2022), who associate psychological stress with heightened participation and intentionality in digital interactions. Consequently, measurements such as bounce rate and pages per visit, which are frequently seen as merely technical, function as behavioural markers of underlying cognitive conditions during crises. The results further corroborate the media synergy principle and theories of cognitive accessibility (Blömker & Albrecht, 2024; Spyridou & Danezis, 2022). The combined effect of organic and paid search on future direct traffic underscores how frequent exposure across channels enhances recall, diminishes cognitive load and increases brand trust.

This corroborates H3, suggesting that trust and memory-based recognition are augmented by coordinated, multi-touchpoint marketing efforts. Consumers exposed to both organic and paid content were more inclined to revisit the site thereafter, underscoring that regular cross-channel exposure is crucial for consumer loyalty and engagement during volatile periods. The efficacy of paid social media traffic, as indicated by previous direct and paid search traffic, emphasizes the significance of sequencing in advertising. These findings correspond with recent research by Halkiopoulos and Papadopoulos (2022) and Seneviratne et al. (2024), which underscores that social media success frequently relies on prior audience engagement through alternative channels. This interdependence demonstrates that user involvement with paid social campaigns is not arbitrary but adheres to a behavioural trajectory influenced by cumulative brand exposure.

H4 is addressed by demonstrating that paid social responsiveness escalates when it follows strategic organic and paid interaction. The findings enhance the current literature by offering micro-level, time-sensitive data that supplement earlier macro-level research on crisis recovery (Ritchie & Jiang, 2021; Zhang et al., 2024; Filippou et al., 2024). The Santorini scenario illustrates the capturing and interpretation of real-time user behaviour through a synthesis of digital KPIs and behavioural theory. This discourse connects empirical evidence with theoretical constructs, including the risk familiarity effect, dual-process cognition and media synergy, providing significant insight into the evolution of digital conduct amid perceived risk and uncertainty.

7. Conclusions

The Santorini Earthquake in 2025 served as a clear crisis signal, which divided the observation period into pre-disaster and post-disaster phases. This study confirms that digital consumer behaviour undergoes distinct, measurable transitions before and after disaster signals—transitions that align with risk perception theory and digital trust cycles.

• Organic traffic levels were higher before the disaster, consistent with risk scanning and proactive search for reassurance, as theorized in crisis psychology (Liu et al., 2024).
• Paid social traffic and direct traffic saw notable increases after the event, indicating heightened brand reliance and successful retargeting among already-engaged users (H3, H4).
• Bounce rate slightly increased pre-crisis, particularly on informational pages, suggesting emotional dissonance and uncertainty.
• Pages per visit deepened marginally after the crisis, especially among returning users, reflecting trust re-engagement and more focused interaction.

These before/after patterns support this study’s premise that behavioural signals evolve meaningfully in response to disaster triggers and that this evolution is detectable through platform-level engagement metrics.

This study makes four distinct contributions. First, it bridges the gap between behavioural crisis response and digital marketing performance by examining how micro-level digital engagement evolves around disaster signals, linking user behaviour under uncertainty to measurable advertising outcomes. Second, it offers a novel temporal framework, linking lagged user interactions with forward-looking advertising outcomes and offering insight into how digital trust and responsiveness are built incrementally during crises. Third, it pioneers the application of fuzzy cognitive mapping (FCM) within the hospitality sector, enabling the visual representation of nonlinear behavioural relationships and providing a decision support framework that captures the complexity of multichannel dynamics in unstable environments. Fourth, this study deepens the theoretical understanding of engagement metrics as cognitive–affective signals, reframing traditional KPIs like bounce rate and session duration as proxies for psychological states such as anxiety, trust and reassurance, particularly in volatile tourism contexts.

7.1. Theoretical Implications

This study contributes novel insights at the intersection of disaster communication, digital consumer behaviour and hospitality marketing. It contributes a micro-level, time-sensitive analysis of how user engagement metrics and traffic sources evolve in response to a real-time disaster signal. In contrast to prior studies that focus on macro-level recovery dynamics (Ritchie & Jiang, 2021; Zhang et al., 2024; Filippou et al., 2024) or platform-specific strategies in isolation (Seneviratne et al., 2024), this research bridges the behavioural layer between digital exposure and consumer decision-making in crisis conditions.

This study builds upon the risk familiarity effect (Sakas et al., 2022b) by applying it to a sudden-onset tourism crisis. As demonstrated by Sakas et al. (2022b) during the COVID-19 pandemic, consumers initially gravitated toward well-known, trusted brands to manage uncertainty. As their familiarity with the crisis increased, they became more open to exploring less-established alternatives, only to return to established providers during renewed periods of anxiety. This cyclical behavioural pattern reflects the psychological mechanism by which familiarity reduces perceived risk, enabling a shift in digital consumption and loyalty. In the context of our study, the Santorini Earthquake functioned as a real-time disaster signal, allowing us to observe how digital consumer behaviour shifted before and after the disruption. Notably, direct traffic increased after the crisis, particularly among users who had previously engaged via organic or paid search—a pattern consistent with the risk familiarity effect, where users re-engage with familiar or trusted digital brands during periods of renewed risk. Our findings support this effect by demonstrating how familiarity—expressed through previous digital touchpoints—buffers against abandonment and facilitates re-engagement during crisis recovery.

The data also affirm the relevance of dual-process decision theory (Blömker & Albrecht, 2024), particularly under conditions of heightened uncertainty. Metrics such as bounce rate and session depth emerge not merely as technical KPIs but as proxies for underlying cognitive mechanisms. During the pre-crisis period, increased bounce rates correspond with emotional dissonance, while deeper post-crisis session engagement suggests more deliberate, systematic processing. These patterns are consistent with the shift from heuristic to systematic information evaluation theorized by Önder and Berbekova (2022), reinforcing the link between psychological stress and digital interaction depth.

The findings of the present study also affirm a set of theoretically grounded assumptions regarding digital behaviour under conditions of perceived threat and uncertainty. First, the interaction between earned and paid media validates the longstanding concept of media synergy and message reinforcement. As users become familiar through organic channels, they exhibit increased responsiveness to paid promotions—enhancing cost-efficiency and confirming that organic visibility lays the groundwork for investment returns (Spyridou & Danezis, 2022). This reinforces the idea that engagement patterns influence the effectiveness of digital advertising (RQ1), particularly in a crisis context where consumers are more selective and emotionally driven.

Second, user engagement metrics such as bounce rate are revealed to be more than technical KPIs; they reflect psychological alignment and emotional resonance. The inverse relationship between bounce rate and page depth supports theories that users in uncertain environments seek trustworthy, emotionally consistent content (Önder & Berbekova, 2022). This insight directly answers RQ2, as it connects user interaction quality with behavioural outcomes, showing how psychological alignment influences user engagement on hospitality websites during times of crisis or disruption.

Third, the role of prior organic and concurrent paid exposure in shaping direct traffic underlines the memory-based and familiarity-driven foundations of brand trust (Sakas et al., 2022b). These findings align with cognitive accessibility theory (Blömker & Albrecht, 2024), which links repeated exposure to reduced cognitive effort and increased user certainty. By validating the dual influence of exposure timing and frequency, this contributes to understanding how historical and concurrent marketing activities across channels contribute to sustained customer interest and loyalty (RQ3), highlighting the importance of repeated engagement in building long-term trust.

Fourth, social media responsiveness is shown not as an isolated output of creativity or tone, but as the culmination of layered, multichannel exposure. Users are more likely to engage with social campaigns after encountering the brand via paid or direct channels—demonstrating the value of retargeting logic and sequenced behavioural engagement (Halkiopoulos & Papadopoulos, 2022; Seneviratne et al., 2024). This finding answers RQ4, showing that cross-channel interactions—specifically the sequence of organic, paid and social media exposure—shape the performance of targeted advertising strategies, especially in dynamic, crisis-affected environments. These outcomes underscore the novelty of this study: it combines web analytics and fuzzy cognitive mapping to construct a temporally layered, cross-channel behavioural model grounded in real-world disaster dynamics. The contribution is not only theoretical—offering new lenses through which to understand digital trust cycles—but also methodological, advancing how behavioural data can be used to trace and predict consumer responses in disaster tourism settings. The Santorini case illustrates this dynamic vividly. Hotels that responded promptly to alerts issued between 1 February and 6 February—such as by publishing safety notifications or offering flexible cancellation options—saw higher direct traffic and longer session durations compared to those that did not. This real-time response behaviour confirms the importance of early-stage digital alignment with evolving media and government messaging, supporting both the risk familiarity effect and broader crisis communication theory (UNESCO, 2025; Euronews, 2025; EMSC, 2025). As digital ecosystems grow more integrated and disaster communication more data-driven, frameworks like the one proposed here offer a foundation for real-time adaptive strategy in hospitality management.

7.2. Practical Implications

The findings offer practical insights for digital marketers, tourism authorities and hospitality managers, reinforcing the behavioural interpretation of engagement signals during disaster periods, as emphasized in prior literature (Celestin et al., 2024).

Interpreting Engagement Metrics as Behavioural Signals (RQ2): The observed relationships between bounce rate, pages per visit and user engagement demonstrate that metrics traditionally treated as raw performance indicators actually reflect underlying cognitive and emotional states during crises. For example, the significant correlation between increased bounce rates and emotional dissonance before the disaster, alongside deeper session durations post-disaster, directly responds to RQ2 regarding how behavioural engagement metrics signify psychological alignment. Marketers should therefore contextualize these KPIs during disaster periods as signals of user reassurance or distress, adapting content accordingly to improve message clarity and trust-building.

Strategic emphasis on Organic Search (RQ1): The strong inverse relationship between organic traffic levels and paid traffic cost (confirmed in hypothesis 1) demonstrates the practical importance of organic search performance in managing marketing budgets and building brand trust before crises. This directly answers RQ1, which investigates how organic and paid traffic interact to affect marketing outcomes. Investing in SEO-optimized, crisis-sensitive content such as safety messaging and FAQs allows brands to anchor user trust and reduce costly reliance on paid ads during volatile periods, making organic visibility a critical asset for crisis resilience.

Importance of Channel Sequencing and Media Synergy (RQ3 and RQ4): The findings on direct traffic’s role in amplifying paid social media effectiveness (hypotheses 3 and 4) highlight the nonlinear, multichannel consumer journey. This directly supports RQ3 and RQ4 by demonstrating how prior organic and paid exposures cumulatively influence subsequent behavioural responses, including direct revisits and social media engagement. For practical marketing, this underscores the need to design campaigns that reinforce organic and branded messages before deploying paid social or retargeting ads, ensuring consistency in narrative and emotional tone across channels for maximum synergy and cost-efficiency.

Maintaining Stable Branded Assets During Crises (RQ3): The increase in direct traffic following the disaster, particularly among users previously exposed to paid or organic search, confirms the critical role of direct traffic as an indicator of brand familiarity and digital loyalty, which is a core concern of RQ3. Hospitality providers should prioritize maintaining updated, emotionally aligned messaging on websites, email communications and branded search assets even amid operational interruptions to capitalize on this loyalty and facilitate seamless user engagement during unstable periods.

Social Media as an Emotional Barometer (RQ4): This study’s empirical evidence shows that social media engagement is strongly conditioned by prior exposures through paid and direct channels, reinforcing RQ4’s focus on the effectiveness of cross-channel sequencing. This indicates that social campaigns not only serve promotional purposes but also function as gauges of brand sentiment and emotional connection during crises. Consequently, marketers should invest in empathetic, credible social media content that reflects ongoing user sentiment and aligns with earlier touchpoints to optimize engagement and reputation management.

Utility of Fuzzy Cognitive Mapping (FCM) for Scenario Planning: While FCM’s application is methodological, it is firmly supported by this study’s real-world data, illustrating complex, nonlinear channel interdependencies during disaster recovery. This practical implication encourages marketers to adopt FCM tools for dynamic scenario simulation and proactive planning, moving beyond reactive adjustments. This recommendation derives from this study’s novel integration of FCM with digital engagement metrics, offering a replicable framework for anticipating user behaviour shifts in unstable environments.

Each of the above implications is based directly on statistically supported findings derived from the tested hypotheses, ensuring that managerial recommendations are empirically grounded rather than speculative.

7.3. Limitations and Future Research

While this study contributes a unique behavioural model, several limitations warrant consideration. First, the scope was restricted to a specific case (the 2024 earthquake in Santorini) and a defined six-month window. While this offered precise temporal insights, broader generalizability may require replication across different types of crises—political, climate-based or technological—and diverse regional contexts.

Second, this study captures mid-term behavioural shifts but not long-term post-crisis recovery. Future research could adopt longitudinal modelling to examine whether trust, engagement and loyalty revert to pre-crisis baselines or evolve into new behavioural norms.

Third, while this study relied on quantitative metrics, it did not include qualitative data such as sentiment in social comments, user reviews or visual content perception. Integrating web analytics with sentiment analysis or user interviews would enrich the interpretive depth and capture more nuanced emotional responses.

Finally, the increasing integration of AI tools and algorithmic curation in user journeys suggests that platform architecture itself is a variable worth studying. Future research should explore how platform-driven personalization during crises interacts with trust formation, consumer psychology and perceived authenticity.