Community-based management for low-digitalized communities using cross-cutting purchasing behavior

The paper addresses the decline of traditional commercial areas due to online shopping and global retail chains, highlighting increased vacancies and reduced footfall, as seen in the UK and Japan. It emphasizes the importance of retaining and circulating money within local communities for revitalization. Community-based management is proposed as a solution, extending beyond individual store strategies to coordinated efforts across commercial areas. The authors argue that transaction-focused consumer behavior analysis is valuable, but low-digitalized shopping communities lack POS and other digital tools to capture cross-cutting purchasing data. Therefore, technology-free methods are needed. The study’s research questions are: (1) How can low-digitalized shopping communities realize community-based management? (2) How can conventional retail management methods using consumer behavior analysis be expanded to community-based management? The proposed approach uses paper-based community currencies to collect transaction-linked, cross-cutting consumer behavior data and applies extended ABC and association analyses at the community level. Two field experiments in Japanese shotengai test feasibility and effectiveness.

The literature indicates transaction data are central to retail management, with prior work using video tracking with transactions, smart carts, AR, and POS/ID-POS data to understand purchasing behavior (Zhang et al., Van Ittersum et al., Tan; Sano & Yada; Williams et al.; Amemiya et al.; Ishimaru et al.). However, most studies focus on single-store management. Efforts to collect cross-cutting consumer behavior data in commercial areas have relied on questionnaires, GPS, observations, and image processing (Sisiopiku & Akin; Flamm & Kaufmann; Denzler & Niemann; Sexton et al.), and agent-based simulations (Kaneda et al.). These methods are not directly tied to transactions, limiting their suitability for transaction-focused analyses. Sharing POS across stores (Matsumoto et al., Machi-POS) improves operations but requires digital adoption and is limited to participating stores, making it unsuitable for low-digitalized communities. Community currencies have been used to promote local consumption and cohesion (Seyfang; Graugaard; Williams et al.; Lietaer & Hallsmith), with analyses typically using questionnaires and network analysis (Kichiji & Nishibe; Kurita et al.; Mattsson et al.). Prior work has not applied consumer behavior analysis to community currency data to understand cross-cutting purchasing at the community level. This study fills that gap by leveraging paper-based currencies to gather transaction-linked data across stores and customer segments.

Data collection used two types of paper-based community currencies that do not require store or customer technology: FT (from-to) and CA (customer attribute). FT-type currencies are distributed by stores; each note carries an ID tied to the distributing store. When redeemed in any participating store, the pair (distributing store i, destination store j) is recorded, enabling analysis of co-occurrence relationships among stores (Cij). CA-type currencies are distributed at a central reception desk, with IDs linked to customer attributes (e.g., students, faculty). When redeemed, the attribute–store pairs (CA, store k) are recorded, enabling analysis of store preferences by customer segments. Both are single-use, one note per transaction, allowing direct estimation of visit frequency. Electronic currencies could capture richer data but have higher implementation costs; combining FT and CA paper types approximates some electronic advantages with low barriers. Analysis methods: Extended ABC analysis identifies frequently visited stores. For FT: compute Sk as total uses of store k; compute Rk by dividing by twice the number of circulated FT notes; top 20% stores are classified as high-frequency. For CA: compute CCA,k (uses at store k by attribute CA), then RCA,k = CCA,k / total circulated for that attribute; top 20% per attribute are high-frequency for that segment. Association analysis (FT only) examines store–store relationships using Cij, Di (total distributed at i), and Uj (total used at j), with support SUPij = (Cij + Cji)/Σi Di, confidence CONFij = (Cij + Cji)/(Di + Ui), and lift LIFTij = CONFij / ((Dj + Uj)/Σi Di). Field Experiments: Study 1 (FT-type): Location: shopping districts around Shin-Okubo and Okubo stations, Tokyo. Period: Nov 29–Dec 13, 2020. Participants: 17 stores (restaurants, general goods, services) out of 55 identified customer-facing stores in the association (~31%). Currency: 200-yen FT note with 4-digit distributor ID; one note per transaction; notes not redistributed after use. Customers received an FT note upon purchases >200 yen; usable at any participating store. Post-period, all notes were collected and logged. Study 2 (CA-type): Location: shopping district near Waseda University, Tokyo. Periods: Nov 12–23, 2018 and Jan 14–25, 2019. Participants: 33 stores (19 restaurants, 9 general goods, 5 services). Currency: 500-yen CA note with slight design differences denoting customer attributes; one note per transaction. Distribution: at a university reception desk; customers earned points via the Machi-Navi walk-rally app and exchanged points for CA notes; receptionists issued notes according to customer attribute (students st; faculty fa). Post-period, redeemed notes were collected and logged. Data governance: store participation voluntary; no direct incentives; data shared only within research group under strict control; customers informed via posters/flyers.

Study 1 (FT-type, 17 stores): • Transactions collected: 498 total; 300 Cij pairs after removing same-store distribute/use data; sufficient for community-level discussion by literature-based scaling. • ABC analysis (visit frequency Rk): Top 20% (3 stores): Store 7 (11.0%), Store 10 (9.5%), Store 12 (11.7%). Store 12 (a pharmacy) acted as a hub: ~74% of its tender uses were distributed from other stores, indicating centrality beyond spatial proximity. • Association analysis (top co-occurrences): - Stores (10,11): SUP=0.172; CONF=0.756 (i=10), 0.872 (i=11); LIFT=3.836. Interpretation: strong linkage likely due to thematic affinity (Indonesian restaurant and Indonesian goods). - Stores (12,14): SUP=0.081; CONF=0.232 (i=12), 0.667 (i=14); LIFT=1.913. - Stores (3,12): SUP=0.056; CONF=0.423 (i=3), 0.159 (i=12); LIFT=1.214. Spatial proximity sometimes aligns with co-occurrence (e.g., 10–11; 12–14) but is not the sole driver (3–12 distant yet linked). • Pareto principle: Not supported here; top 20% stores’ Rk sum to 32.2%. Study 2 (CA-type, 33 stores): • Transactions collected: 338 total; sufficient by literature-based scaling. • Overall ABC (Table 6): Frequently visited stores (top 20%, 6 stores): 8 (18.2%), 9 (10.7%), 10 (22.0%), 11 (8.9%), 14 (18.5%), 17 (4.2%). • Segment-specific ABC: - Students (st) top stores: 1, 9, 10, 11, 14, 16 (e.g., Rst,9=37.3%; Rst,10=30.1%; Rst,11=19.3%). - Faculty (fa) top stores: 8, 10, 11, 13, 14, 17 (e.g., Rfa,8=23.7%; Rfa,10=19.4%; Rfa,14=23.3%). • Behavioral insights: Purchases were concentrated in restaurants, consistent with the area’s proximity to work/school rather than residences. Notable differences: store 8 drew mainly students; store 9 drew mainly faculty; student preferences were more concentrated. • Pareto principle: Supported here; top 20% stores’ Rk sum to 82.5%. Practical insights: For university-adjacent districts, inviting restaurants can drive traffic; coupling food/beverage coupons with promotions for general goods/services may steer cross-category shopping.

The findings demonstrate that paper-based community currencies can collect transaction-linked, cross-cutting consumer behavior data in low-digitalized communities, addressing the first research question. By extending ABC and association analyses from product-level retail contexts to store- and community-level contexts, the study provides a viable analytical framework for community-based management, addressing the second research question. Identifying high-frequency stores helps allocate shared community resources (e.g., event materials, targeted discounts) to maximize community-wide impact, aligning with central place theory’s emphasis on service hubs. Association analysis reveals store–store linkages driven by proximity and thematic affinity, informing cooperative strategies such as joint promotions or targeted couponing between linked stores. Segment-based ABC using CA-type currencies enables customer segmentation at the community level, supporting tailored interventions (e.g., student- vs faculty-focused campaigns). The mixed support for the Pareto principle (absent in Study 1, present in Study 2) suggests context dependence and the need for further research across varied settings.

The study shows how low-digitalized shopping communities can realize community-based management by employing paper-based community currencies (FT and CA) to collect cross-cutting, transaction-linked behavioral data. It extends conventional retail consumer behavior analyses (ABC and association analysis) to the community level, demonstrating feasibility in two field experiments (498 FT and 338 CA transactions). The approach yields actionable insights: identification of community traffic generators, store–store associations for cooperative strategies, and segment-specific preferences for targeted management. The work contributes a practical methodology for practitioners (commercial area managers, local governments) and opens a pathway for researchers to adapt additional retail analytics (e.g., RFM, trend analysis) to community management. Future research should (1) implement and evaluate concrete management interventions informed by these analyses, (2) extend beyond ABC and association analyses to richer community analytics, and (3) improve data quality via multi-use or item-recording community currencies and/or digital currencies to capture multi-store and product-level relationships.

• Data granularity: Single-use paper currencies capture only pairwise relationships (FT: two stores; CA: one attribute–store) and lack product-level detail. Enhanced designs (multi-use notes, recording purchased items) or digital currencies could capture multi-store and multi-attribute relationships. • Analytical scope: Only ABC and association analyses were applied; broader methods (e.g., RFM, temporal trend analyses) could yield more comprehensive insights. • Generalizability: Findings derive from two Japanese shopping districts with specific contexts (university-adjacent and multicultural areas); broader studies across varied communities are needed to assess external validity and contextual drivers (e.g., why Pareto held in Study 2 but not Study 1).