Digital retailing practices for triggering physical retailers' bounce-back and bounce-forward performance against a great shock: evidence from the COVID-19 pandemic

The study investigates whether and how pre-shock digital retailing practices enable physical retailers to achieve resilient performance against a great shock, specifically COVID-19. Motivated by concerns that prior research often assesses composite digitalization and general resilience capabilities (absorptive, adaptive, transformative) without linking concrete practices to observable performance under specific shocks, the paper proposes two shock-specific performance constructs: bounce-back (containing unpredictable crises during the shock) and bounce-forward (continuous adjustment to achieve a new normalcy). It adapts the structure–conduct–performance (S-C-P) framework to directly map digital retailing practices to these performance types. The research focuses on four pre-shock practices centered on online-to-offline food delivery services (O2OFDS) and social media marketing services (SMMS), each operated via proprietary platforms (PP) or large third-party platforms (TP): O2OFDS-PP, SMMS-PP, O2OFDS-TP, and SMMS-TP. The hypotheses posit: (H1) PP-based practices will not trigger bounce-back, while TP-based practices may; (H2) O2OFDS-TP will trigger bounce-back throughout the shock; (H3) SMMS-TP will trigger bounce-back in the later stage; and (H4) all four practices will trigger bounce-forward performance. The paper aims to address temporality concerns of digital resilience by examining the generality vs. individuality of these practice–performance links within the COVID-19 context.

Prior work largely evaluates digitalization-enabled resilience via composite strategies or capabilities (e.g., enterprise systems, IoT, big data analytics, AI, Industry 4.0, supply chain digitalization) and often reports positive links to resilience and performance. However, it seldom isolates specific digitalization practices or examines shock-specific performance under great shocks. The review (summarized in Fig. 1) highlights a gap: studies emphasize capabilities and sometimes resilient outcomes but rarely map concrete digital practices to resilient performance in a singular shock. Some evidence during COVID-19 shows mixed success for e-commerce adoption, underscoring that composite digitalization may not equate to practice-level effectiveness. The paper thus advances by focusing on specific digital retailing practices (O2OFDS, SMMS; PP vs. TP) and linking them to bounce-back and bounce-forward performance using an adapted S-C-P lens, recognizing the individuality of great shocks and the potential temporality of digital resilience.

Design: Quantitative difference-in-differences (DID) supplemented with qualitative executive interviews and an event-study (ESA) dynamics analysis. Data: Panel of 50 Chinese listed general merchandise retailers, 549 firm-quarter observations from 2018Q1–2020Q3 (post indicator after 2019Q4). Firms are brick-and-mortar retailers; some established digital practices pre-2020. Treatments: Four pre-shock digital retailing practices—O2OFDS-PP (15 firms, 105 obs.), SMMS-PP (8 firms, 56 obs.), O2OFDS-TP (9 firms, 63 obs.), SMMS-TP (3 firms, 21 obs.). Placebo treatments: owning PP, participating in TP, proprietary delivery (PD), supermarket main business (SuperM), diversified. Outcomes: Asset turnover (AT) and components measured as growth rates—Sales growth (SalesG), Current assets growth (CAG), Illiquid assets growth (FAG). Financials from Choice SSE/SZSE Stock Database. Controls and fixed effects: Firm-level (total assets, inventory), regional GDP and express volume; firm and quarter fixed effects; cluster SEs at firm level. Model: DID with interaction terms of each treatment and Post to estimate average treatment effects on treated (ATTs), controlling for placebo interactions and covariates. Event study specifications decompose dynamics and test pre-trends (parallel trends assumption). No triple-difference terms due to mutually exclusive practice advantages. Identification hinges on stable parallel trends pre-shock. Qualitative: Semi-structured interviews with 6 executives from 6 retailers (3 with O2OFDS-related, 5 with SMMS-related businesses during COVID-19). Only insights corroborated by at least two interviewees used as supplementary evidence. Mechanism check: Annual data on newly established branches (2019, 2020) to assess whether TP-based practices correlate with a slowdown in branch expansion during COVID-19, conditional on controls.

Main DID results (Table 4): - Sales growth (SalesG): - O2OFDS-TP × Post: +31.028, p<0.01 (significant sales lift vs. non-adopters). - SMMS-TP × Post: +17.140, p>0.10 (not significant). - O2OFDS-PP × Post: +5.515 (ns); SMMS-PP × Post: +4.760 (ns). - Current assets growth (CAG): No strong significant effects for any practice. - Illiquid assets growth (FAG): - O2OFDS-TP × Post: −113.993, p<0.10 (slower illiquid asset growth). - SMMS-TP × Post: −153.175, p<0.05 (slower illiquid asset growth). - O2OFDS-PP × Post: −86.475, p<0.01, but event study shows a pre-existing downward trend (likely not a causal effect of the shock exposure). - SMMS-PP × Post: +117.647, p<0.01 (short-term investment jump consistent with added costs to upgrade proprietary infrastructures). - Asset turnover (AT): - O2OFDS-TP × Post: +0.183, p<0.01; SMMS-TP × Post: +0.132, p<0.05 (overall operational performance improved for TP-based adopters). - O2OFDS-PP and SMMS-PP effects on AT not significantly positive (SMMS-PP negative coefficient −0.129, p<0.05). Dynamics (event study, Fig. 7): - Parallel trends generally hold; exception: O2OFDS-PP on FAG shows a pre-shock declining trend. - Bounce-back timing: O2OFDS-TP boosts sales immediately after outbreak (2020Q1); SMMS-TP boosts sales in 2020Q2 (later stage). Effects fade over time, consistent with temporary during-shock market structure. - Inventory/current assets: Little evidence of inventory reductions; current assets effects limited. - Bounce-forward: TP-based practices (O2OFDS-TP, SMMS-TP) consistently associated with reduced illiquid asset growth post-shock, indicating reconfiguration toward asset-light models. Mechanism (Supplementary): Retailers on large TP platforms reduced branch expansion more than non-adopters during COVID-19, with no significant differences in total assets changes, suggesting intentional strategic reconfiguration rather than forced contraction. Hypotheses: H1 supported (PP-based did not trigger bounce-back; TP-based offered bounce-back opportunity). H2 supported (O2OFDS-TP triggered bounce-back across the shock). H3 supported (SMMS-TP triggered bounce-back in the second stage). H4 partly supported (bounce-forward triggered, especially via TP-based practices; PP-based evidence weaker/mixed).

Findings validate the adapted S-C-P framework’s distinction: bounce-back requires coincidental alignment with temporary, shock-induced market structures, while bounce-forward reflects firms’ capacity to reconcile operations with long-term market evolution. TP-based practices matched shifted consumer preferences for high-visibility, high-capacity channels during COVID-19, enabling immediate sales recovery (O2OFDS-TP) and later-stage demand capture (SMMS-TP). Conversely, PP-based practices lacked visibility/scale during the shock, failing to trigger bounce-back. Across practices, general involvement in digital retailing supported bounce-forward performance by enabling quicker sensing of post-shock trends and responsive reconfiguration (e.g., slowing illiquid asset growth, reduced branch expansion). This suggests a form of generality: different practices can converge to similar bounce-forward outcomes even if their bounce-back efficacy differs. The results contribute to IS/OM debates by directly mapping specific digital practices to observable resilient performance, addressing temporality concerns and highlighting the individuality of practice–shock matching for bounce-back. Managerially, reliance on digital practices to achieve bounce-back poses cost-effectiveness risks when misaligned with shock-specific conditions; firms may be better served by temporary measures leveraging existing physical strengths for survival, while using digital channels to build long-term adaptability. To fully realize bounce-forward benefits, firms should integrate digital initiatives across operations rather than isolating them in standalone segments.

The paper demonstrates that pre-shock digital retailing practices affect physical retailers’ resilience during COVID-19 in distinct ways. TP-based practices (O2OFDS-TP, SMMS-TP) triggered bounce-back performance aligned with evolving consumer behavior during the shock and improved asset turnover, while PP-based practices did not. Multiple practices—particularly TP-based—also facilitated bounce-forward performance by enabling strategic reconfiguration toward an asset-lighter posture (e.g., slower illiquid asset growth, reduced branch expansion). Contributions include: (1) defining and operationalizing bounce-back and bounce-forward as shock-specific resilient performance constructs; (2) providing a direct practice-to-performance mapping via an adapted S-C-P framework; and (3) supplying causal evidence (DID with ESA) on how specific digital practices perform under a great shock. Future research should test these mappings across other industries, countries, and shocks to evaluate generality of bounce-forward effects, and further unpack mechanisms and contexts where PP-based practices can effectively enable longer-term resilience.

- External validity: Sample restricted to Chinese listed general merchandise retailers; industry and country specificity may limit generalization. - Measurement and identification: For SMMS-PP, explanations are non-exclusive and evidence on bounce-forward is mixed; O2OFDS-PP illiquid assets effect likely reflects pre-trends rather than treatment. - Data constraints: Some mechanisms (branch expansion) available only annually, limiting granularity; interviews are few and provide supplementary—not definitive—evidence. - Scope: Focus on the first COVID-19 wave and four specific practices; other digital practices and subsequent shocks warrant examination to assess the robustness of bounce-forward generality.