E-Commerce in the Organic Food Sector: A Comparative Study of Ukraine and China

Abstract

Whether digital market access translates into retained farm income remains insufficiently resolved by studies of adoption and purchase intentions. Drawing on the resource-based view and transaction cost economics, this study examines revenue and profitability in a non-probability telephone survey sample of 186 farms: 74 in Ukraine and 112 in China, including 175 farms reporting organic certification and 11 uncertified comparison farms. Welch’s tests, country-specific OLS regressions with HC3 standard errors, and moderation analysis distinguish unadjusted sample contrasts from within-country associations. Mean monthly online revenue is higher in the Chinese sample (USD 12,878.30 versus USD 4237.30), whereas reported return on sales (ROS) is higher in the Ukrainian sample (30.85% versus 17.72%). The ROS gap is 13.13 percentage points (95% confidence interval: 11.43–14.83). Marketing estimates lack robust positive support, and primary tests do not establish video moderation or differences in ROS between platform groups at the 5% level. Outcomes cover all online activity, not platform-specific accounts. The contribution is an empirical distinction between sales scale and margin retention across contrasting producer settings. Instagram and WeChat cases translate this distinction into proposals for channel-cost accounting and controlled marketing trials. Findings concern participating farms rather than national populations or causal country effects.

1. Introduction

Digital selling redistributes commercial work along agri-food supply chains. Producers can reach customers and negotiate orders themselves, while assuming responsibility for promotion and sales support. The constraints faced by short food supply chains are closely connected to the organization of successive operational tasks [1]. Their considerable diversity makes a single model of direct exchange an inadequate basis for evaluating performance [2]. Digital platforms add value when their functions align with the farm’s needs rather than merely expanding its technological options [3]. The economic question is therefore one of commercial organization: who performs each function, which resources are committed, and how the resulting value is distributed.

Direct contact with final consumers can retain part of the intermediary’s margin within the producer–customer relationship, but it also transfers the intermediary’s work to the farm. The often-under-recognised labor contribution of local producers is central to this trade-off [4]. Digital marketing performance similarly depends on strategy and organizational capabilities, rather than expenditure alone [5], while the choice of agricultural marketing arrangements reflects asset specificity and the conditions of exchange [6]. The resource-based view (RBV) and transaction cost economics (TCE) offer complementary explanations: RBV addresses the firm’s ability to organize sales, and TCE addresses the relative costs of performing functions internally or delegating them.

Organic products add an information problem to these commercial tasks. Buyers rely on provenance and certification because product appearance does not reveal production practices. Verifying such credence attributes involves several supply chain participants and mechanisms to address information asymmetry [7]. Evidence on the joint use of product certification and e-commerce shifts the focus from quality assurance to farm-level economic outcomes [8]. Certification provides a basis for differentiating an offer, while a digital channel communicates that differentiation to the customer. Financial performance depends on how effectively the two are combined.

The presentation of the product also matters. Consumer responses to information about organic food depend on trust [9], whereas producers evaluate digital tools in relation to their own commercial needs [3]. This creates a potential complementarity between marketing, finance, and communication: expenditure secures access to an audience, but the content of the offer shapes the customer’s response. The relevant farm-level question is whether a given marketing budget is associated with greater online revenue when video is used to present the product.

The unresolved issue is not simply whether farmers enter e-commerce, but what distinguishes commercially viable use after entry. Adoption, consumer intention, sales receipts and the profit share retained by a producer represent different stages of the exchange process. Connecting these stages requires an account of both customer acquisition and the work needed to complete an order. Recent research on market access likewise distinguishes participation decisions from channel choice and the distribution of economic gains [10]. The present study addresses the narrower, observable relationship between marketing inputs, selling practices, and farm-level financial outcomes among existing online sellers.

Comparing Ukraine and China places this question in contrasting production and marketing environments. Maintaining access to customers during wartime is particularly important in Ukraine. Research on disruptions to production and trade establishes the wider economic context [11], while work on marketing in the bioeconomy [12] and the local identity of artisanal foods [13] highlights two related priorities: sustaining sales and differentiating the farm’s offer. Artisanal positioning concerns the brand’s identity; certification concerns the product’s production status. These can coexist within the same business model.

The Chinese literature places greater emphasis on the organization of market access. Cooperative membership is associated with participation in e-commerce [14], and rural online trading is examined in terms of market integration and household income [15]. Studies of rural livestreaming add the acquisition of commercial skills and dependence on platforms to this account [16], as well as knowledge sharing within agricultural clusters [17]. Small landholdings can thus coexist with access to commercial resources beyond the individual farm. This provides a useful starting point for comparison: the capacity to sell digitally develops both within the business and through its organizational relationships.

The choice of Ukraine and China follows this organizational question rather than an assumption that their farms are interchangeable. The Ukrainian context makes continuity of customer contact and flexible order coordination economically salient, while Chinese research foregrounds cooperative services, platform access and the development of selling skills [11,14,15,16,17]. Both settings concern producers seeking customers beyond the farm gate, but they expose different ways of combining internal resources with external commercial services. The analytical gain comes from examining the same financial questions under contrasting coordination conditions, not from attributing the observed margin difference to nationality. Country-specific regressions retain a common economic specification while allowing the resource distributions and channel categories to differ.

The present study covers 186 farms, including 74 in Ukraine and 112 in China. Ukrainian landholdings range from 8 to 2800 ha, with a median of 31.5 ha; Chinese holdings range from 0.3 to 20 ha, with a median of 0.8 ha. Owners are aged 32–45 and 35–66 years, respectively. All Ukrainian farms and 101 Chinese farms report certification; the remaining 11 Chinese observations form an uncertified comparison group. The main platforms are Instagram, Viber/Telegram, TikTok, and Prom.ua in Ukraine, and Taobao Live, WeChat, and Pinduoduo in China. All participants already sell online, so the analysis focuses on the organization and performance of existing digital sales rather than on initial adoption.

The objective is to establish how marketing expenditure, video use, and the main platform are associated with online revenue and profitability. Combining RBV and TCE connects the resources committed to selling with the allocation of exchange costs. The contribution is a theory-informed economic assessment of digital practices in two contrasting, non-probability farm samples, distinguishing the scale of online receipts from the share retained after costs. Environmental outcomes require a separate set of measures [18]. Local-market development motivates the research, but buyer locations and delivery distances were not recorded. The empirical scope is therefore the farms’ total online activity, not a geographically delimited local market.

2. Literature Review and Hypotheses Development

2.1. Digital Capability, Resource Complementarity, and Coordination

Access to a digital channel and the ability to earn a return from it are distinct. A platform may be available to many sellers, but each seller brings a different assortment, customer service experience, and order-fulfillment capacity. RBV explains this heterogeneity through the combination of technology and organizational capabilities [5]. What matters is the suitability of digital functions for the commercial task. A tool that broadens reach may also complicate the coordination of small consignments or demand skills the producer has yet to acquire. Its expected value consequently depends on its fit with established farm practices [3]. Finance enables access; the organization of use determines how productive that access becomes.

The foundational RBV distinction is between access to a resource and the advantage arising from a heterogeneous combination of resources [19]. In this setting, a marketing budget is an observable input; the ability to target customers, maintain a credible offer, and fulfil orders is an organizational property. Recent farm-level studies operationalise digital capability through multiple practices and identify different configurations of transformation [20,21]. Expenditure alone is consequently not treated as a validated capability measure.

Digital capability develops through combinations of resources rather than within a single application. Multidimensional measures used in agricultural marketing research capture a wider range of practices than online presence alone; even direct selling may encompass both electronic transactions and physical markets [20]. Configurational studies of digital transformation and research on value creation in farmers’ markets place similar emphasis on organizational conditions and resource complementarity [21,22]. Land, customer knowledge, and communication practices do not operate independently. An additional opportunity to present a product becomes commercially useful when the farm can deliver the promised quality and fulfil the resulting order. The coordination of these activities distinguishes capability from a collection of technical tools.

Resource productivity also depends on the organization of exchange. Customer search, negotiation, monitoring, and the protection of specialized investments are central concerns of TCE. Cooperative marketing can distribute these functions among participants, and its association with access to e-commerce makes institutional coordination relevant to digital selling [6,14]. Market access helps explain the attraction of participation, but the financial benefit to an individual farm also depends on the price of that access. An intermediary’s service should be assessed against the work it replaces. Performing the same tasks internally requires resources of its own.

TCE locates governance choices in the characteristics of exchange, including asset specificity, uncertainty and transaction frequency [23]. Its application here concerns the allocation of customer search, negotiation, and fulfillment between the producer and commercial intermediaries. A platform fee can purchase coordination that would otherwise require on-farm labour; a direct relationship can avoid that payment while leaving the underlying work with the producer. These alternatives motivate a profitability comparison without presuming that either channel structure is invariably less costly [4,6,24].

For organic products, establishing a shared understanding of quality can be particularly demanding. Buyers assess both the product and the credibility of information about its production. Exchange, therefore, relies on verification mechanisms involving producers, retailers, and certification bodies [7]. Price barriers intertwine with information barriers: acceptance of a more expensive offer depends on whether its distinctive attributes are intelligible and credible. The literature on organic consumption in China, synthesised from studies conducted between 2011 and 2021, reflects this interdependence [25]. Digital channels provide alternative ways to address it by placing certification information, provenance, and access to the seller alongside the product. Transaction costs may fall when this arrangement makes relevant answers easier to obtain and agreements easier to reach.

The boundary between work performed by the farm and services supplied by a platform is shaped by both resource capacity and coordination costs. Managing an offer internally preserves control over its content while assigning the producer responsibility for sales support. Delegation changes both workload and payment obligations. This distinction separates the capacity to expand sales from the capacity to retain profit. Revenue records the scale of demand served, whereas return on sales (ROS) expresses the profit remaining after the costs of serving it.

2.2. Marketing Expenditure and the Scale of Online Revenue

Marketing expenditure purchases access to customer attention. Its relationship with revenue runs through exposure to the offer, visits to the digital channel, and the eventual purchase, with different conditions affecting each stage. Agri-food web analytics treats traffic composition and acquisition expenditure as distinct elements of marketing activity [26]. Revenue can also change without an equivalent change in customer numbers: access to digital markets is associated with selling prices and the timing of transactions, as shown in research on apple producers [27]. Sales expansion, therefore, has both quantity and price dimensions. Identical promotional budgets may support product flows that differ substantially in volume and value.

Accumulated reputation lies between current spending and customer response. Price differentiation in the organic segment is linked to certification, advertising support, production costs, and consumers’ evaluations of the brand. The classification of premiums for nine green brands, drawing on assessments by 250 buyers, illustrates the multiple factors involved [28]. Loyalty, calculated from satisfaction and the importance of product attributes, reflects an evaluation of the offer; a repeat order is a realised behaviour. The object of trust matters as well. Information about an organic product is interpreted through trust in its source, and confidence in the producer and retailer may reinforce or weaken one another [9,29]. A monthly budget thus captures current communication expenditure, while the return on that expenditure also reflects relationships developed over time.

Expenditure and performance can influence each other. Promotion helps attract demand, but existing turnover also provides the funds for the next campaign. Comparisons across farms are subject to self-selection because resources and commercial opportunities differ before a particular tool is used. Instrumental-variable approaches to market integration and endogenous switching models of joint certification and e-commerce choices address this problem explicitly [8,15]. When budgets and revenue are observed simultaneously, a regression coefficient describes their conditional association across farms. Its interpretation depends on the resource differences included in the model, as well as on the sign of the estimated coefficient.

A positive association remains a reasonable theoretical expectation: promotional finance expands opportunities to present the offer and maintain contact with prospective buyers. Assessing that expectation requires controls on available landholding characteristics, owner age, video use, and the main platform, together with certification where status varies. The relevant outcome is current monthly online revenue, evaluated separately within each country.

The expectation is conditional on deployment. Evidence on digital-marketing performance links resources to strategic execution [5], while agricultural studies associate e-commerce with changes in selling prices, transaction timing and the returns to certification [8,27]. These mechanisms support a positive expenditure–revenue hypothesis without implying a uniform marginal return. At the same time, larger receipts can finance promotion and weaker demand can induce extra spending. A cross-sectional coefficient must therefore be read as an association within the observed resource configuration, rather than as a campaign-level return on investment.

H1. 

Controlling for observed farm characteristics and the main platform, higher monthly digital marketing expenditure is positively associated with current online revenue. This association is tested separately for Ukraine and China.

2.3. Video Content as a Complementary Marketing Practice

Audience reach acquires commercial value when customers find a reason to choose the offer. In social commerce, that reason is shaped by the information environment, presentation, and interaction with the seller as well as by price and product identity [30,31]. Video extends the scope of presentation through visual demonstration; live broadcasts also allow audience questions to become part of the presentation itself. Associations between livestreaming system characteristics and intentions to purchase agricultural products underscore the importance of this interactive component [32]. Interest must nevertheless be followed by agreement on price, availability, and fulfillment. Communication operates within the trading process rather than replacing it.

What the message emphasizes is equally consequential. An information experiment involving 1019 respondents identified a difference in willingness to make environmentally motivated purchases between health-focused and environment-focused messages, while the control and environmental groups did not differ significantly [33]. Persuasiveness, therefore, depends on the particular emphasis and comparison. Trust and the mixed reception of digitalized short supply chains help explain why a more convenient or visually engaging format can have different value for different buyers [9,34]. A provenance demonstration, a consultation, and a promotional offer may all use video, but each serves a different commercial purpose. Richer information is useful when it helps the buyer assess the offer.

Potential complementarity between expenditure and video follows from this distinction between attracting attention and converting contact into an order. Finance brings the offer to an audience; communication practices shape how that contact is handled. Their combination requires skills. Presentation, customer responses, and sales coordination develop through learning and knowledge exchange, while dependence on the platform can increase alongside commercial opportunity [16,17]. The costs of acquiring and applying these skills remain part of the channel’s economics. Models of agricultural livestreaming link profitability to the terms of interaction among participants, thereby distinguishing sales expansion and margin improvement as distinct outcomes [24].

H2 concerns whether the combination is more productive than the separate practices. Social-commerce research links purchasing to the information and interactions surrounding an offer [31], and studies of agricultural livestreaming examine the interactive features that support purchase intentions [32]. Under the resource-complementarity argument, video should strengthen the revenue association of marketing finance when the audience reached can more readily evaluate and order the product. The interaction tests that observable implication; content quality, presentation skill, and audience trust are not themselves measured by the binary video indicator.

Complementarity implies more than higher average revenue among video users: it implies a stronger association between revenue and marketing expenditure. Econometrically, this is a difference in elasticities, represented by the interaction ln(Marketing_USD) × Video_Stream in a model of ln(Sales_USD), with both main effects included. Variation in video use within Instagram and Taobao Live provides a basis for this comparison. Because Video_Stream covers both live and recorded material, the hypothesis concerns the use of video broadly. A positive interaction coefficient indicates a higher marketing–revenue elasticity among farms that use video.

H2. 

Video use positively moderates the association between marketing expenditure and current online revenue: holding other characteristics constant, the corresponding elasticity is higher for farms with Video_Stream = 1 than for farms with Video_Stream = 0.

2.4. Platform Differences in Profitability and the Allocation of Transaction Costs

Expanding digital demand raises the question of how much commercial work the producer will perform internally. Reducing intermediation in direct selling leaves more tasks on the farm: negotiation, small-order assembly, and customer relationship maintenance all require time and organizational effort [4]. The diversity of short supply chains reflects different allocations of this workload, while their resilience depends on coordination across successive stages [2,35]. The absence of a separate payment to an intermediary does not remove the cost of the intermediary’s functions. What matters is whether the retained trading margin exceeds the cost of the work taken on.

A platform offers a different allocation of responsibility. The producer delegates part of the coordination to a specialized participant and accepts its charges and rules. This can be financially worthwhile even with a lower profit share if access to buyers and the cost of serving them remain favourable. In models of agricultural video commerce, this balance depends on seller–platform arrangements and incentives for quality control involving logistics providers [24,36]. Intermediation is therefore a choice about how to perform commercial functions, rather than a contrast between an unnecessary commission and a costless direct relationship. Research on cooperative organizations similarly distinguishes the decision to participate from its financial consequences [6].

In organic production, these costs interact with opportunities for price differentiation. The economic performance associated with joint certification and e-commerce use varies by scale and production type, while the benefits of digital access can appear separately in selling prices and transaction timing [8,27]. Quality assurance adds further work: credible information must be prepared and communicated to the buyer, and the associated tasks are distributed across the exchange relationship [7]. A higher price creates scope for greater profit, but the retained share depends on production and marketing costs. ROS provides a common financial measure of these potentially offsetting changes.

Comparing profitability requires separating the platform from the environment in which it operates. The Ukrainian and Chinese platform sets do not overlap in this sample. A direct comparison of Instagram and Taobao Live would consequently need to account for platform differences, production structures, and country-specific conditions. Within-country analysis addresses a more focused question: do ROS differences between platform groups persist after controlling for farm characteristics and video use? The platform identifies the principal digital channel, whereas direct-to-consumer (D2C) selling is defined by the transaction itself. Both internal coordination and specialized intermediation can generate advantages, leaving the direction of profitability differences open.

A non-directional H3 follows from these competing cost mechanisms. Direct selling retains commercial tasks within the farm [4], whereas platform participation changes the allocation of services, charges and quality-control responsibilities [24,36]. Different net margins are therefore plausible even when the channels all broaden market access. The hypothesis concerns differences between farms grouped by their principal channel, conditional on the available characteristics; it does not rank the intrinsic efficiency of applications or identify platform-specific profits.

H3. 

Controlling for observed resource and demographic characteristics and video use, mean online ROS differs between farms using different main digital platforms within the same country.

2.5. Economic Logic of the Empirical Model

Digital sales performance emerges from the relationship between demand acquisition and the cost of serving that demand. Marketing expenditure finances the presentation of the offer, video changes the mode of communication, and the platform provides the setting for the transaction. Resource complementarity explains how combining these elements may expand revenue; the allocation of transaction functions explains why margins may follow a different path. H1 and H2, therefore, address sales scale and the revenue association of marketing resources, while H3 concerns the share retained by the farm.

The empirical framework comprises eight substantive variables—Age, Area_ha, Sales_USD, Marketing_USD, ROS_%, Certificate, Platform, and Video_Stream—alongside ID and Country. Land area represents the physical resource base; owner age and certification capture demographic and production-status differences. The country defines the comparison setting. Revenue and ROS cover the farm’s overall online activity, while grouping by the main platform identifies its principal channel orientation. The precision of comparisons depends on group size, particularly for Prom.ua, which is represented by two Ukrainian farms. The platform hypothesis is therefore assessed through the joint test of platform coefficients.

RBV and TCE connect these observations through the economics of commercial decision-making: acquiring resources is valuable when their deployment yields an acceptable financial return after order fulfillment. The regressions estimate conditional relationships among the observed measures, preserving the distinction between current revenue and reported profitability. Direct measurement of transaction costs would extend the analysis from these outcomes to the underlying mechanism. As sales expand, so do the producers’ obligations to customers. Managing that relationship is central to the economic value of digital selling.

The two theories organize exploratory associations rather than tests of latent mechanisms. Table S1 maps resource deployment, complementarity, governance, and quality assurance to the observed variables and specifies what each measure leaves unresolved. The same distinction defines the paper’s sustainability contribution: it addresses the economic viability of producers’ sales, while environmental performance and social outcomes require independent measurement [18].

3. Materials and Methods

3.1. Data Collection and Sample Profile

A cross-sectional telephone survey conducted over a two-month April–May fieldwork period yielded 186 complete analytical records from farms engaged in online selling: 74 in Ukraine and 112 in China. Publicly accessible directories of organic producers in the two countries provided business contacts. Interviewers telephoned the listed numbers and interviewed representatives who answered, were able to discuss the farm’s online selling, and agreed to participate. Ukrainian co-authors conducted the interviews in Ukraine, while Chinese co-authors conducted the interviews in China. The farm is the unit of observation, and the outcomes are average monthly online revenue and reported return on sales. The analytical dataset contains ten fields: ID, Country, Age, Area_ha, Sales_USD, Marketing_USD, ROS_%, Certificate, Platform, and Video_Stream. All identifiers in the supplied analytical file are unique, and all ten fields are complete; 186 is the number of available complete records, not the number of attempted contacts. The corrected dataset supplied by the authors was analysed without imputation, winsorisation, or further alteration of the observations.

Recruitment is described as directory-based convenience sampling. The substantive population of interest comprises farm businesses in the organic-food sector that use online channels, with uncertified records retained as an explicitly identified comparison group. Neither random selection probabilities nor fixed country quotas governed inclusion. Directory coverage, telephone contactability and willingness to disclose information shape the resulting sample. Total calls, refusals, and any incomplete interviews were not retained in the analytical documentation, so a response rate and a population sampling fraction cannot be calculated. The results support an exploratory comparison of participating farms, not estimates for either national farm population.

All 74 Ukrainian farms and 101 Chinese farms report certification. These 175 producers form the certified subsample; the 11 Chinese farms without certification are retained as an uncertified comparison group. Primary estimates use the full sample, with additional analyses restricted to certified farms. Landholdings range from 8 to 2800 ha in Ukraine (median 31.5 ha) and from 0.3 to 20 ha in China (median 0.8 ha). Owners’ ages range from 32 to 45 and from 35 to 66 years, respectively. The Ukrainian platforms are Instagram, Viber/Telegram, TikTok, and Prom.ua; the Chinese platforms are Taobao Live, WeChat, and Pinduoduo. Differences in landholding structure and the non-overlapping platform sets motivate separate country models.

Sales_USD measures average monthly gross online revenue, R; Marketing_USD measures monthly digital marketing expenditure, M; Area_ha records total operated agricultural land, including rented land, A; and Age records the owner’s age, even when another representative answers. ROS is the respondent-reported percentage of online revenue remaining as profit after attributable production, logistics, marketing, and other costs. Certificate equals 1 for farms reporting organic certification and 0 otherwise; certificates were not independently checked. Video_Stream similarly records video use, including both live broadcasts and recorded material. Platform identifies the principal channel designated by the respondent, rather than a channel selected from verified revenue shares; revenue and ROS encompass the farm’s entire online activity. Respondents were asked to provide both monetary amounts directly in US dollars. These are respondent-reported USD equivalents, including their own conversion where needed. The research team did not impose a common exchange rate or reconvert the answers. Original-currency amounts, respondent exchange rates and conversion dates were not recorded, so the dollar figures are approximate reported equivalents rather than harmonised accounts.

Financial responses were not independently reconciled with accounting statements. The questionnaire defines a common reference period and online-sales scope, but completed reference-period dates and itemised cost schedules are absent from the analytical file. Consistent treatment of unpaid family labour, depreciation, taxes, land costs, subsidies, commissions, returns and shared overheads cannot therefore be verified. April–May describes when interviews took place, not necessarily the period over which each monthly average was calculated. Respondent gender and role, product categories, cooperative membership, online-selling tenure and the share of online activity in total sales were not recorded as analytical variables. Owner age must not be interpreted as the age or digital proficiency of the person managing sales.

The contact-based sample comprises active online sellers with completed telephone interviews. Statistical inference assumes independence across farms; cooperative identifiers were unavailable for modelling within-cooperative dependence. The commercial organization of the channels is examined through two illustrative business cases: INBERRI’s Instagram profile and a video-based product offer on WeChat. Author-supplied screenshots support the comparison of positioning, assortment navigation, seller contact, and ordering.

Supplementary Section S6 reproduces the author-supplied questionnaire and its English translation prepared for editorial review. It makes the item wording, response instructions, and consent script available without treating that later translation as evidence of historical cross-language validation. Ukrainian and Chinese co-authors conducted the respective national interviews; neither a contemporaneous Chinese instrument nor a translation-validation record is available in the documentation. Table S2 distinguishes author-confirmed procedures from unresolved fieldwork details, including the calendar year, exact directory editions, sampled regions, and interview-level financial periods. The reported ethics approval and verbal-consent procedure are retained in the back matter; dated field records are needed to complete the chronology check.

Sample adequacy depends on the inferential task rather than the sample’s percentage of an unspecified population [37]. The baseline models contain eight estimated parameters, leaving 66 residual degrees of freedom in Ukraine and 104 in China; the moderation models leave 65 and 103. These counts permit estimation but do not establish adequate power for every coefficient. The unadjusted ROS gap is estimated relatively precisely, whereas sparse platform groups, interaction terms and correlated resource measures carry greater uncertainty. No a priori power calculation is documented. Confidence intervals, distributional summaries and specification checks are therefore used to assess the information supplied by the sample, without converting a non-probability design into a representative one.

3.2. Econometric Strategy

The analysis distinguishes sales scale, the association of marketing resources with revenue, and profitability across channel orientations. Before estimating the models, revenue, marketing expenditure, and ROS were checked for internal financial consistency. For the same accounting period and sales coverage, total costs implied by ROS must be at least as large as marketing expenditure. The implied balance of other costs must therefore satisfy

$$C_{\mathit{other}}=R\times \left(1-{\displaystyle \frac{\mathit{ROS}}{100}}\right)-M \ge 0$$

(1)

The analysis uses the corrected workbook already underlying the baseline manuscript. Comparison with its preceding version identifies seven changed entries: one Ukrainian revenue value and six Chinese marketing values (Table S3). Both values are reported for each change; the file comparison does not establish the original cause or provide respondent-level verification. Three Chinese records remain inconsistent with the common-period accounting identity (IDs 117, 170 and 186; Table S4). They are retained in the all-record baseline because the identity alone cannot identify which field should be corrected. Accounting-screened models omit all three without substituting values and are displayed beside the full-sample estimates in Tables S6 and S7.

Observations with a negative balance of other costs were flagged for sensitivity analysis. Full-sample estimates were compared with estimates excluding those records. Quantitative distributions were summarised using means, sample standard deviations, medians, and ranges. Between-country differences in age, area, revenue, ROS, and marketing expenditure were tested using two-sided Welch’s t-tests for independent samples. Mean differences were calculated as Ukraine minus China, with 95% confidence intervals (CIs). Holm adjustment was applied across the five comparisons. The significance level was set at 0.05.

The level model is retained as the reporting baseline because its coefficients have a direct monetary interpretation. Logarithmic models examine proportional associations, while the area-exclusion specification diagnoses dependence between the two resource measures. No archived analysis plan establishes that this ordering preceded inspection of outcomes; the comparison is exploratory. Area is not removed to obtain a preferred sign. Table S5 presents the marketing estimate and its confidence interval under each specification so that interpretation rests on their consistency rather than one selected coefficient.

H1 was assessed using country-specific multiple linear regressions estimated by ordinary least squares (OLS). The level specification preserves the monetary interpretation of coefficients and includes land area, owner age, video use, certification, and the main platform:

$$R = \alpha + {\beta }_{1}M + {\beta }_{2}A + {\beta }_3\mathit{Age} + {\beta }_{4}V + {\beta }_{5}C + {\sum }_k{\delta }_k P_k + \epsilon$$

(2)

Here, V denotes Video_Stream, C denotes Certificate, and P_k denotes platform indicators. The coefficient β₁ represents the difference in online revenue associated with a USD 1 increase in marketing expenditure, holding the other included predictors constant; β₂ represents the corresponding difference per hectare. Instagram and WeChat are the reference platforms for Ukraine and China, respectively. The certificate is omitted from the Ukrainian model because it is constant across all 74 farms. The design matrices have full column rank. Heteroskedasticity-consistent HC3 standard errors were used, with t-tests and CIs based on residual degrees of freedom. Tables report unstandardised coefficients, B. The predictor set was retained irrespective of individual coefficient significance.

HC3 corrects the estimated covariance matrix for heteroskedasticity; it does not resolve multicollinearity, omitted-variable confounding or simultaneous determination of budgets and revenue. The dollar interpretation refers to reported USD equivalents. It is an association between records conditional on the included covariates, not a causal advertising multiplier.

For H2, the log-revenue model includes an interaction between marketing expenditure and video use. The natural logarithm of Marketing_USD was centred on its mean within each country:

$$m = \mathit{ln}\left(M\right) - \overline{\mathit{ln}\left(M\right)}$$

Revenue, expenditure, and area are strictly positive, allowing the use of natural logarithms without added constants. The model includes the interaction together with both main effects and the control variables:

$$\mathit{ln}\left(R\right) = \alpha + {\beta }_{1}m + {\beta }_2\mathit{ln}\left(A\right) + {\beta }_3\mathit{Age} + {\beta }_{4}V + {\beta }_{5}C + {\sum }_k{\delta }_k P_k + \theta \left(m \times V\right) + \epsilon$$

(3)

The parameter θ measures the difference in marketing–revenue elasticities between video users and non-users. For V = 0, the elasticity is β₁; for V = 1, it is β₁ + θ. The standard error of the latter was calculated from the coefficient variances and covariance. H2 was assessed using the interaction’s sign, p-value, and 95% CI. For H3, ROS replaces revenue as the dependent variable in Equation (2), retaining the same controls. Platform indicators were assessed jointly using an HC3-robust Wald test reported in F form. Pairwise contrasts were Holm-adjusted across all six platform pairs in Ukraine and all three pairs in China. The joint test determined the assessment of H3.

Sparse channel groups are retained to preserve the submitted sample composition, with the two-farm Prom.ua contrast treated descriptively and its exclusion examined separately. Full column rank is a computational condition, not evidence of precise interaction estimation. Video combines recorded material and live broadcasting; frequency, duration, and third-party presentation were not separately measured. Table S8 reports the country-specific platform-by-video cells. Neither a non-significant interaction nor a joint platform test indicates equivalence.

Diagnostics examined the marketing–area correlation, variance inflation factors (VIFs), and observation leverage. H1 was additionally assessed in a log specification and in a level model excluding land area. Sensitivity analyses were conducted for certified farms only, after excluding financially inconsistent records, after omitting the two Ukrainian Prom.ua observations, and after omitting the Chinese farm with 20 ha. Individual regression tests are two-sided and unadjusted. Marketing intensity is total expenditure divided by total online revenue within each country. A hypothetical common revenue base of USD 1000 illustrates the monetary meaning of mean ROS; it is not an estimate of mean farm profit. Computations were performed in Python 3.13.5 using NumPy 2.3.5, SciPy 1.17.0, and statsmodels 0.14.6. Alternative specification coefficients, confidence intervals, sample restrictions, and diagnostic summaries are in the Supplementary Materials. The diagnostic extension additionally examines externally studentised residuals, Cook’s distance, and leave-one-out marketing estimates. Distribution plots are presented in Figures S1–S5, resource and revenue scatterplots in Figures S6–S9, and influence plots in Figures S10 and S11.

Between-country tests remain unadjusted sample comparisons; separate regressions do not make the two samples compositionally equivalent. In the shared land range of 8–20 ha, there are 27 Ukrainian records and only one Chinese record, the 20-ha holding. Matching on area would discard almost the entire Chinese sample without producing broad common support. No matching, reweighting, or pooled causal country coefficient is therefore reported. Differences in significance between the two national regressions are not used to test differences between their coefficients.

4. Results

4.1. Descriptive Statistics and the ROS Gap

The two samples exhibit contrasting financial profiles: farms in the Chinese sample have higher online revenue and marketing expenditure, whereas farms in the Ukrainian sample maintain a larger reported profit share despite lower turnover. The comparison also involves different landholding structures and owner ages. All five between-country differences remain significant after Holm adjustment (p < 0.001). Table 1 provides the descriptive baseline for these resource and financial contrasts.

Table 1. Descriptive statistics and between-country comparisons. Panel (A): means, standard deviations and Welch tests. Panel (B): medians, quartiles and ranges.

(A)
Variable	Ukraine, n = 74 Mean ± SD	China, n = 112 Mean ± SD	Difference Ukraine−China	Welch t	p
Owner age, years	40.39 ± 3.46	54.53 ± 8.73	−14.13	−15.408	<0.001
Land area, ha	660.66 ± 824.02	1.05 ± 1.85	659.61	6.886	<0.001
Online revenue, USD/month	4237.30 ± 3494.35	12,878.30 ± 8921.43	−8641.01	−9.234	<0.001
ROS, %	30.85 ± 6.43	17.72 ± 4.47	13.13	15.281	<0.001
Marketing expenditure, USD/month	283.65 ± 178.25	1209.64 ± 712.35	−925.99	−13.148	<0.001
(B)
Variable	Ukraine (n = 74) Median [Q1, Q3]	Ukraine Min–max	China (n = 112) Median [Q1, Q3]	China Min–max
Owner age, years	41.00 [38.00, 43.00]	32.00–45.00	57.00 [51.00, 61.00]	35.00–66.00
Operated land area, ha	31.50 [16.13, 1175.00]	8.00–2800.00	0.80 [0.60, 1.20]	0.30–20.00
Online revenue, USD/month	3000.00 [2125.00, 4025.00]	720.00–18,000.00	14,750.00 [2600.00, 19,000.00]	170.00–34,000.00
Marketing expenditure, USD/month	172.50 [140.00, 407.50]	90.00–700.00	1375.00 [265.00, 1750.00]	95.00–2600.00
ROS, %	32.00 [30.00, 35.00]	12.00–38.00	18.00 [16.00, 20.25]	6.00–26.00

Notes: SD is the sample standard deviation. Differences are Ukraine minus China; the ROS difference is in percentage points. Tests are two-sided Welch’s t-tests. Unadjusted p-values are reported; all five comparisons remain significant after Holm adjustment. Degrees of freedom, in row order, are 156.615, 73.000, 155.762, 119.119, and 131.244. Calculations use unrounded values. Source for all statistical tables: the corrected author-supplied dataset of 186 farms. Q1 and Q3 denote the 25th and 75th percentiles, calculated by linear interpolation. The records are identical to those in Panel A. Age refers to the owner; area is operated agricultural land. Figures S1–S5 show the distributions.

The Ukrainian median of 31.5 ha conceals considerable dispersion: relatively small farms coexist with much larger producers, raising the mean above the median. Chinese holdings are concentrated at a much smaller scale, with a median of 0.8 ha. Yet the mean Chinese online revenue is 3.04 times the Ukrainian value, and the mean marketing expenditure is 4.26 times as large. Landholding size and digital sales scale, therefore, capture different aspects of the business. Market access and the organization of selling merit attention alongside the production base.

The distributional detail in Table 1B and Figures S1–S5 makes the asymmetry explicit: the middle half of Ukrainian-operated holdings spans 16.13–1175 ha, compared with 0.6–1.2 ha in China. Median online receipts are USD 3000 and USD 14,750, respectively. The comparison, therefore, concerns different production and trading profiles; product specialisation, resale, and the proportion of output sold online cannot be separated with the recorded variables.

The principal profitability result is a 13.13-percentage-point ROS gap in favour of the Ukrainian group (95% CI: 11.43–14.83; p < 0.001). On a hypothetical common USD 1000 revenue base, the unrounded mean ratios correspond to USD 308.51 and USD 177.23 of reported profit, respectively: a difference of USD 131.28. This standardised comparison expresses margin retention on a per-turnover basis, not mean farm profit; the latter depends on the joint distribution of individual revenues and margins. Farms scaling their digital operations must manage the value of incoming orders and the share retained after fulfillment together.

Marketing expenditure accounts for 6.69% of aggregate Ukrainian online revenue and 9.39% of Chinese online revenue. The mean Chinese budget of USD 1209.64 accompanies higher turnover but also requires a larger share of revenue to finance promotion. This measured cost contrast is consistent with different demands on the sales organization. Direct negotiation keeps more tasks on-farm, whereas platform-based selling combines access to organized demand with investment in visibility and service provision. Separating promotional expenditure, commissions, logistics, and the producer’s own labor would establish the full cost of each arrangement.

The ROS difference is stable across the accounting and certification checks. Three Chinese records have a negative implied balance of other costs under Equation (1). Excluding these observations yields a difference of 13.15 percentage points (95% CI: 11.44–14.87); restricting the analysis to the 175 certified producers yields 13.09 percentage points (95% CI: 11.36–14.82). Both comparisons remain significant at p < 0.001. The unadjusted between-sample contrast is therefore not driven by these particular records or by the uncertified comparison group. These checks do not adjust for production specialisation, location, seasonality, customer mix, or cost attribution (Table S7).

Video use is reported by 48.65% of Ukrainian farms and 35.71% of Chinese farms. It is concentrated on Instagram in Ukraine and on Taobao Live and WeChat in China. The presence of both users and non-users within these channels permits assessment of the expenditure–video interaction while accounting for the main platform.

The cross-tabulation also reveals limited within-channel variation: all 36 Ukrainian video users are on Instagram, whereas seven TikTok and two Prom.ua records report no video use. In China, 36 of 56 Taobao Live records report video use, and 20 do not; WeChat contains four video users among 28 farms (Table S8). The available records cannot distinguish whether Taobao non-use reflects the wider platform environment, third-party presentation, or inconsistent interpretation. Submitted codes are retained rather than inferred from platform names.

4.2. Regression Results

The financial contrast between countries is clearer than the explanatory pattern within each country. Both level models yield positive marketing coefficients, but only the Ukrainian estimate is statistically significant. The regressions retain the same economic controls: land area, owner age, video use, and the main platform, with certification additionally estimated in China (Table 2 and Table 3).

Table 2. OLS estimates for Ukraine. Dependent variable: Sales_USD.

Predictor	B	SE (HC3)	t	p
Intercept	−2707.889	3712.966	−0.729	0.468
Marketing expenditure, USD 1	57.786	17.563	3.290	0.002
Land area, 1 ha	−11.829	3.701	−3.196	0.002
Owner age, 1 year	11.064	125.510	0.088	0.930
Video use: 1 vs. 0	−2262.691	1782.314	−1.270	0.209
Certification: 1 vs. 0	NE	NE	NE	NE
Viber/Telegram vs. Instagram	−2876.903	1889.606	−1.522	0.133
TikTok vs. Instagram	−2338.491	2043.259	−1.144	0.257
Prom.ua vs. Instagram	−3509.452	3889.465	−0.902	0.370

Notes: N = 74; R² = 0.2692; adjusted R² = 0.1917; residual df = 66. B is an unstandardised coefficient; SE is the HC3 robust standard error. NE: not estimated because all Ukrainian farms report certification. Instagram is the reference platform.

Table 3. OLS estimates for China. Dependent variable: Sales_USD.

Predictor	B	SE (HC3)	t	p
Intercept	4744.332	7090.625	0.669	0.505
Marketing expenditure, USD 1	1.190	0.973	1.224	0.224
Land area, 1 ha	281.714	225.494	1.249	0.214
Owner age, 1 year	124.039	114.996	1.079	0.283
Video use: 1 vs. 0	1101.057	2164.057	0.509	0.612
Certification: 1 vs. 0	−1553.931	2912.984	−0.533	0.595
Taobao Live vs. WeChat	439.531	2196.043	0.200	0.842
Pinduoduo vs. WeChat	1693.895	2497.839	0.678	0.499

Notes: N = 112; R² = 0.0400; adjusted R² = −0.0247; residual df = 104. B is an unstandardised coefficient; SE is the HC3 robust standard error. WeChat is the reference platform.

In the Ukrainian-level model, USD 1 higher marketing expenditure is associated with USD 57.786 higher online revenue, holding the other included predictors constant (95% CI: 22.719–92.852; p = 0.0016). The result changes sharply with specification. Excluding area reduces the coefficient to USD 1.739 (95% CI: −5.127–8.605; p = 0.615), while the log model produces a negative marketing elasticity of −2.068 (95% CI: −3.781 to −0.355; p = 0.0187). Expenditure and land area are almost linearly related (r = 0.990; VIFs of 61.59 and 56.33). The positive budget coefficient consequently sits alongside a negative area coefficient in a model with little independent variation in resources to separate them. Across specifications, H1 lacks robust positive support: the independent marketing association cannot be reliably separated from the observed resource structure (Table S5).

The Chinese coefficient is USD 1.190 of revenue per USD 1 of marketing expenditure (p = 0.224; 95% CI: −0.739–3.119). The log estimate is also positive and non-significant (0.080; p = 0.462). Resource multicollinearity is limited, with VIFs close to 1, but the level model explains only 4.00% of the variation in revenue. Removing the three financially flagged observations increases the marketing estimate to USD 1.650 (p = 0.086), but it does not reach the adopted significance level. The available expenditure measure and farm characteristics provide no stable statistical support for H1 in the Chinese sample.

Adjusted R² is −0.0247 for the Chinese-level model, indicating little explanatory gain after accounting for model size, rather than the economic irrelevance of the included practices. Influence diagnostics identify maximum leverage of 0.518 in Ukraine and 0.975 for the 20-ha Chinese holding. Leave-one-out marketing coefficients range from 51.483 to 66.018 in Ukraine and 0.937 to 1.509 in China; deleting one record does not remove the Ukrainian instability across functional forms. An auxiliary Ukrainian budget–area regression explains 98.03% of the variation in budget. This is a descriptive fit, not evidence of a budgeting rule or of how the values were generated (Table S9; Figures S6–S11).

The main platform does not jointly distinguish revenue after controlling for the other characteristics (p = 0.485 in Ukraine and p = 0.789 in China). Profitability remains a separate economic outcome: similar turnover can involve different amounts of service work and external payments. Table 4, therefore, compares platform groups using ROS.

Table 4. Return on sales by the main platform and adjusted contrasts with the reference platform.

Country/Platform	N	ROS, % Mean ± SD	B, Percentage Points	95% CI for B	p (Holm)
Ukraine/Instagram	48	30.88 ± 6.79	Reference	—	—
Ukraine/Viber/Telegram	17	30.18 ± 6.95	1.03	[−5.15, 7.22]	1.000
Ukraine/TikTok	7	31.86 ± 2.79	2.29	[−2.84, 7.43]	1.000
Ukraine/Prom.ua	2	32.50 ± 3.54	2.98	[−4.83, 10.79]	1.000
China/WeChat	28	19.21 ± 2.96	Reference	—	—
China/Taobao Live	56	16.80 ± 5.13	−2.58	[−4.79, −0.38]	0.066
China/Pinduoduo	28	18.07 ± 3.97	−0.10	[−2.20, 1.99]	0.924

Notes: Means and SDs describe the observed platform groups. B and CIs are estimated from the ROS regression controlling for Marketing_USD, Area_ha, Age, Video_Stream, and Certificate where it varies. B is expressed in percentage points. CIs are individual, unadjusted 95% intervals; p-values are Holm-adjusted across all within-country platform pairs. An individual CI can therefore exclude zero while the adjusted p-value exceeds 0.05. Financial outcomes cover each farm’s overall online activity, grouped by its main platform.

Ukrainian mean ROS values are close across platforms, and the joint test of adjusted differences gives p = 0.765. In China, farms identifying Taobao Live as their principal channel have an adjusted ROS 2.58 percentage points lower than farms identifying WeChat (individual 95% CI: −4.79 to −0.38). At a hypothetical identical USD 1000 revenue base, the point estimate represents USD 25.84 in profit. Its unadjusted p-value is 0.0221, rising to 0.0664 after Holm adjustment; the joint Chinese platform test gives p = 0.0628. H3 is not supported by either of the primary country models at the 5% level.

The Chinese platform test remains above the significance threshold after excluding the three flagged financial records (p = 0.0664) and within the certified subsample (p = 0.1074). Excluding the 20-ha farm gives p = 0.0579. In Ukraine, removing the two Prom.ua observations leaves platform differences non-significant (p = 0.595). These checks preserve the distinction between the persistent between-sample ROS gap and the weaker evidence for differences among platforms within countries.

The primary interaction tests do not establish a difference in marketing–revenue elasticity by video use (Table 5). The interaction estimate is 0.059 in Ukraine (p = 0.853), and 0.331 in China (p = 0.166); both confidence intervals include zero. Adding the interaction increases R² by 0.0006 and 0.0189, respectively. In the Ukrainian moderation model, the estimated elasticities are negative for both non-users (−2.059; p = 0.021) and users (−2.000; p = 0.045), reinforcing the sensitivity of the budget relationship to functional form. Their difference is non-significant. Chinese simple slopes are −0.043 for non-users and 0.287 for users, neither of which is statistically significant. H2 is unsupported in both primary models.

Table 5. Moderation of the marketing–revenue association by video use.

Country	Interaction θ	SE (HC3)	t	p	95% CI
Ukraine	0.059	0.320	0.186	0.853	[−0.580, 0.699]
China	0.331	0.237	1.395	0.166	[−0.139, 0.800]

Notes: The dependent variable is ln(Sales_USD). The interaction is country-centred ln(Marketing_USD) × Video_Stream. Models include both main effects, ln(Area_ha), Age, platform indicators, and Certificate for China. R² is 0.1629 for Ukraine and 0.0695 for China; residual df are 65 and 103. SEs and CIs use HC3.

The Chinese interaction becomes larger after the three flagged records are excluded (0.407; p = 0.0506), indicating sensitivity to accounting consistency while remaining above the 5% threshold. Restricting the sample to certified farms or omitting the 20-ha holding also leaves H2 unsupported. For commercial decisions, the useful distinction is between adopting video and organizing its contribution to an order. Provenance demonstrations, consultations, and product promotions are combined in one binary indicator, although their demands on content preparation and fulfillment differ.

The interaction intervals, −0.580 to 0.699 in Ukraine and −0.139 to 0.800 in China, span both zero and economically meaningful differences in elasticity. The results leave the direction and magnitude unresolved; they do not establish that video practices are equivalent. Table S6 places the accounting-screened, certified-only and small-group exclusion results side by side, and Table 6 summarises the evidence for all three hypotheses.

Table 6. Summary of the hypothesis assessments and robustness qualifications.

Hypothesis	Ukraine	China
H1: positive marketing association	No robust support. Level B = 57.786; log elasticity = −2.068. Signs and magnitude depend on specification.	Not established. Level 95% CI: −0.739–3.119; log interval also includes zero.
H2: positive video moderation	Not established. θ = 0.059; 95% CI: −0.580–0.699. Sparse within-channel variation.	Not established. θ = 0.331; 95% CI: −0.139–0.800. Accounting-screened p = 0.0506.
H3: platform-group ROS differences	Not established. Joint p = 0.765; p = 0.595 without Prom.ua.	Not established. Joint p = 0.0628; Holm-adjusted Taobao Live–WeChat p = 0.0664.

Notes: H1 concerns a positive marketing–revenue association, H2 a positive video interaction, and H3 differences between principal-platform groups. Non-significant tests are not evidence of equivalence. Tables S5–S7 present specification and sample-restriction results.

4.3. Applied Business Cases

The Instagram and WeChat cases illustrate two ways of organizing commercial contact. INBERRI’s profile directs prospective buyers towards a personal exchange with the producer. WeChat offers the product demonstration, price, and ordering functions within the same interface (Figure 1 and Figure 2). These arrangements allocate work differently between the seller and the digital system.

Figure 1. INBERRI on Instagram: regional branding, product assortment, and direct customer contact. Source: author-supplied image; provenance details are recorded in Table S2.

Figure 2. A video-based product offer on WeChat: purple carrots, price, and purchasing functions. Source: author-supplied image; provenance details are recorded in Table S2.

The cases were selected from the author-supplied material to contrast a producer profile with an integrated product listing. The same four criteria organize their interpretation: producer identity, assortment presentation, seller contact, and order initiation. They are analysed separately from the survey regressions because no respondent-ID linkage or case-level accounts accompany the images. Their analytical role is to illustrate commercial arrangements and generate testable managerial propositions, rather than attribute the sample means to either business.

INBERRI uses blueberry_sumy as a recognisable storefront for vegetables and berries from the Sumy region. The farm name, regional identity, and themed categories give the assortment a clear origin and help buyers locate relevant products. Images of harvesting, cultivation, and packaging connect the offer to the production story. The message and contact buttons then provide a route for enquiries about availability, quantities, and dispatch. The same seller and contact route remain available for repeat enquiries. This arrangement creates an opportunity to maintain customer relationships and negotiate orders around seasonal availability; realised retention and service-time effects would require transaction-linked observation. Where the transaction bypasses a trading intermediary, the trading margin remains within the producer–buyer relationship, while the time spent on correspondence and order assembly becomes part of the cost of selling.

WeChat provides a different purchasing route. A video of purple carrots appears alongside a starting price of CNY 23.9, a shopping cart, and a purchase button. The crop setting and the seller’s emphasis on minimal post-harvest handling position the offer around proximity to production. Commercial action is available directly next to the presentation: customers can review the product and its price before placing an order in the same interface. This arrangement moves part of the information needed for ordering from personal correspondence into the product listing. To turn video attention into revenue, the seller must align the displayed consignment with stock availability, price, and fulfillment. Preparation, promotion, and servicing all enter the economics of the video-based offer.

The contrast concerns commercial organization as much as content format. The Ukrainian profile leaves greater scope for individual agreement and repeated contact within a small customer community; WeChat provides an integrated route to a more standardised purchase. Under a transaction-cost interpretation, the choice depends on the resources required to acquire and serve a completed order. A personalised profile can incorporate explicit prices and delivery terms, while a video-based product listing can give the farm a clearer identity. Combining a distinctive offer with straightforward ordering is a practical option to evaluate through repeat purchases, labor time, and channel profit.

5. Discussion

5.1. Sales Scale, Margins, and Channel Organization

The central economic contrast is between the scale of demand served and the profitability of that demand. Farms in the Chinese sample report a higher mean monthly online revenue, USD 12,878.30, compared with USD 4237.30, but a lower ROS, 17.72%, compared with 30.85%. The 13.13-percentage-point difference (95% CI: 11.43–14.83) persists across the specified accounting and certification sensitivity analyses. For the farm, these represent distinct management objectives: increasing the value of orders and retaining a sufficient share after fulfilling them. A higher margin raises profit per unit of revenue, whereas the absolute result also depends on turnover and the distribution of individual farm outcomes.

This contrast is not an estimate for otherwise equivalent producers. The landholding, age and platform distributions differ, and the available controls do not harmonise production mix, buyer composition or accounting practice. Its scientific value lies in exposing an economic configuration that warrants explanation: substantially larger reported receipts coexist with a smaller retained share. Treating that configuration as an outcome to investigate, rather than as proof of a national model, gives the subsequent comparison a clearer interpretive basis.

Previous research describes the benefits of digital access through several outcomes: higher prices and more favourable selling times, income associated with market integration, and complementarity between e-commerce and certification [8,15,27]. Designs accounting for endogenous participation compare different adoption states; this study compares producers who already trade online. Lower ROS in the Chinese sample is compatible with gains from market access. It identifies a tension between turnover and the resources required to sustain it, extending the adoption discussion to the channel’s ongoing economics.

The distinction in estimands matters. Research in Food Policy evaluates certification and e-commerce choices using a multinomial endogenous switching design and reports heterogeneous gains across producer scales and crop groups [8]. A market-access synthesis in Economic Analysis and Policy similarly separates participation, channel choice, and welfare outcomes [10]. Those findings concern adoption choices or broader market integration, whereas the present sample contains existing online sellers. A lower observed ROS in one participating group can coexist with benefits from e-commerce relative to non-participation; it does not contradict those studies or reproduce their causal comparisons.

The Ukrainian configuration combines universal reported certification, a median area of 31.5 ha, and the predominance of Instagram, the main channel for 48 of 74 farms. The mean area of 660.66 ha reflects the presence of much larger producers, making the resource profile heterogeneous. The mean ROS of Instagram users (30.88%) is close to the means for the other Ukrainian channels. Higher margins consequently characterise the broader sample rather than an isolated advantage on Instagram. The INBERRI case shows one commercial arrangement within this setting: recognisable provenance, assortment navigation, and direct contact are combined around a named farm. This organization is consistent with the emphasis on technological fit and value creation in short supply chains [3,22].

For a producer negotiating available products and dispatch dates directly with customers, flexibility means being able to change an order while maintaining the relationship. War-related disruption makes this control particularly relevant in Ukraine [11]. Its economic content includes labour; however, individual responses, small-order assembly, and repeated coordination remain farm responsibilities. Research on short supply chains connects their diversity to differences in the allocation of precisely these operations and workloads [2,4]. Retained trading margins support adaptation when they cover this work as well as production costs.

Farms in the Chinese sample, with a median area of 0.8 ha, operate through different channel structures: Taobao Live, WeChat, and Pinduoduo account for 56, 28, and 28 observations, respectively. High online revenue alongside small holdings draws attention to commercial resources beyond the individual plot. Cooperative services and knowledge exchange within agricultural clusters are discussed as ways to broaden access to those resources [14,17]. They provide a possible organizational explanation of platform integration; the variables observed here are platform choice, expenditure, and financial outcomes. The WeChat case makes the commercial arrangement concrete by placing video within the ordering process.

Market access requires finance. Chinese marketing expenditure is 4.26 times the Ukrainian mean, compared with a revenue ratio of 3.04, and accounts for a larger share of aggregate turnover: 9.39% versus 6.69%. The distribution of value between the producer, the digital channel, and order fulfillment becomes a central business question. Research on rural livestreaming links skill acquisition to platform dependence, while models of digital supply chains make profitability sensitive to intermediation arrangements [16,24]. Platform services can replace the seller’s own work; direct interaction leaves more tasks in-house. TCE locates the comparison in the cost and organization of these functions [6]. Separating advertising, commissions, delivery, and labor would establish where scaling increases profit and where it primarily expands turnover.

Country context also enters through the buyer side. Evidence from China associates organic-food choices with information credibility and trust, while the literature review identifies both informational and price barriers [9,25]. The Ukrainian context adds disrupted production and trade to the organization of market access [11]. These sources explain why identical digital functions may face different commercial conditions, but the survey does not observe consumers’ willingness to pay, basket composition, or repeat purchasing. Such mechanisms remain propositions for linked producer–customer research rather than inferred national behavioural traits.

5.2. Marketing Resources, Video Content, and the Limits of Platform Advantages

The Ukrainian results place a sharp qualification on the expected positive relationship between marketing resources and revenue. The level coefficient is large and positive (B = 57.786; p = 0.0016), but it shrinks after area is excluded and becomes negative in the log specification (elasticity = −2.068; p = 0.0187). This sign reversal, together with pronounced expenditure–area multicollinearity, makes specification sensitivity the substantive finding. Resource-based research connects performance to the fit between strategy, organizational capability, and farm needs [3,5]. The monthly budget captures financing, while the records leave campaign design, customer targeting, and fulfillment capability unmeasured. H1 lacks robust positive support across Ukrainian specifications, despite the significant positive coefficient in the level model.

The Chinese budget coefficient is positive (B = 1.190) but imprecisely estimated (p = 0.224). Its log counterpart and the sensitivity analyses likewise do not establish the positive association proposed in H1. Web analytics studies distinguish traffic composition and visitor behaviour as separate elements of marketing performance [26]; aggregate expenditure may finance both new audiences and existing demand. The sequence can also run in the opposite direction: revenue funds advertising, while weak sales can prompt additional promotion. Approaches that address selection and endogeneity explicitly separate such relationships [8,15]. These differences in measurement and identification explain why a monthly farm budget does not carry the same interpretive content as a campaign-level response estimate.

Neither primary moderation model supports H2. The expenditure–video interaction is 0.059 in Ukraine (p = 0.853) and 0.331 in China (p = 0.166), with 95% CIs of [−0.580, 0.699] and [−0.139, 0.800]. The larger Chinese coefficient after removal of the accounting flags (p = 0.0506) merits attention as a sensitivity result, not a change in the primary hypothesis decision. Consumer research captures information responses, purchase intentions, and social-commerce interactions [31,32]; farm revenue aggregates realized transactions. Stock availability, pricing, order acceptance, and fulfillment intervene between these outcomes. A favorable audience response can therefore coexist with an uncertain difference in revenue elasticities.

The cases illustrate this organizational transition. INBERRI centers the offer on familiarity with the producer and direct dialogue, whereas WeChat places purchasing functions beside the video. Provenance information, confidence in its source, and an accessible ordering route perform different tasks [9,32]. Their coordination, rather than video presence alone, offers a basis for learning across the two settings. Video is already reported by 48.65% of Ukrainian and 35.71% of Chinese farms. Seller preparation, presentation structure, responses to buyers, and fulfillment discipline are therefore more relevant areas for exchange than simply increasing broadcast frequency. Research on knowledge sharing and livestreaming skills supports this organizational focus [16,17]. Trials should record completed sales alongside content production and labor costs.

Within-country platform differences in ROS are less persistent than the between-sample gap. The primary joint tests yield p = 0.765 in Ukraine and p = 0.0628 in China, while the Taobao Live–WeChat contrast has p = 0.0664 after Holm adjustment. The additional sample checks retain the non-significant joint result. H3 remains unsupported at the adopted threshold. A channel’s contribution depends on contractual arrangements, the allocation of work, and audience access. Intermediaries can perform functions that are costly for individual farms, while direct channels retain more of the trading margin alongside the producer’s own labor [4,6,24]. Comparing the full cost of completed orders is therefore more informative for management than ranking applications by a single sample mean.

Certification occupies a distinct position in this account. Every Ukrainian farm reports certification, whereas the Chinese revenue estimate relies on a comparison with 11 uncertified farms and is not significant (p = 0.595). Published evidence of certification–e-commerce complementarity concerns broader combinations of practices [8], and quality assurance involves several supply-chain participants [7]. The commercial value of certification depends on how the relevant information enters the offer, customer choice, and transaction terms. Measuring these activities would complement the binary status indicator.

The comparison clarifies how RBV and TCE connect access to a resource, its commercial deployment, and the retention of value. Expenditure finances contact, video supports presentation, and the platform provides the transaction setting; profitability depends on their fit with the farm’s operations [3,5,6]. A persistent ROS gap alongside unstable digital-practice coefficients makes order fulfillment and customer management important targets for further explanation. Financial viability remains one dimension of sustainability. Environmental effects require their own measures [18].

For RBV, the implication is to distinguish the quantity of marketing finance from the organizational capability that deploys it [19]. For TCE, the focus is on distinguishing channel access from the costs and governance of completing exchange [23]. Neither distinction can be collapsed into a platform label. The findings thus refine the measurement requirements of a combined resource-and-governance account: campaign execution, customer-service labor and contractual charges should be observed alongside expenditure and receipts. This is a theoretical implication for subsequent testing, not a direct validation of unmeasured mechanisms.

The managerial priority is a comparable contribution account for each selling route: receipts net of promotion, commissions, packing, delivery, returns and service time. Cooperatives can use the same account to compare a shared service with the work it replaces for members. For support agencies, the corresponding evaluation concerns completed orders and retained income rather than registrations on a platform or the number of broadcasts. These proposals preserve the possibility that a lower-margin channel remains commercially worthwhile at higher turnover; they require firm-level testing rather than a universal recommendation to migrate to one application.

5.3. Limitations and Future Research

The comparison covers 186 active online sellers reached through publicly listed organic-producer contacts during the April–May fieldwork window. This recruitment route can favour businesses with accessible contact details and those willing and able to complete an interview. Selection probabilities are unknown because recruitment depended on directory coverage, contactability, and voluntary response, limiting population-level inference. Differences in landholding, owner age, and platform composition also constrain comparability, particularly with only two Ukrainian Prom.ua observations. A broader design should use comparable production specialisations and farm sizes and include producers who have left digital channels or have yet to adopt them. Recording the number of attempted contacts and non-responses would make telephone-survey coverage and participation more transparent.

Selection can operate before and during the call. Public directories may favour visible, formalised businesses, while willingness to disclose financial figures can select more organized or more confident respondents. Restriction to current online sellers also excludes ventures that stopped selling, a potentially important survivorship filter under wartime disruption. The unrecorded contact denominator prevents estimation of non-response bias, and the narrow owner-age range in Ukraine should be read as a feature of the recruited group, not as the national producer profile. Calendar year, directory editions, regional coverage, and completed financial reference periods remain details requiring documentary confirmation.

Farm-level aggregates reveal overall performance but provide limited detail on individual transactions. The ten fields do not separate product assortment, physical quantities, revenue by channel, commissions, delivery costs, and labor time. Recording these components would distinguish price, volume, and organizational explanations of turnover. Buyer locations and transaction types would also establish the geographical scope of local sales and the share of D2C activity. Cooperative identifiers would enable the modelling of shared conditions and clustered observations. Linking the business cases to survey and transaction records would give the qualitative account of the interface a farm-specific financial basis.

Reported USD equivalents improve the common presentation of answers but do not standardise the exchange rates used by respondents. Differences in rounding and conversion dates can affect monetary comparisons, and common interview months do not establish common financial averaging periods. ROS is unit-free, but its comparability still depends on consistent attribution of labour, land, overheads, taxes and other expenses to online activity. Neither a common currency label nor a satisfied accounting identity resolves these measurement questions.

Behavioural mechanisms require a different level of measurement. Trust, perceived quality, digital self-efficacy, and risk attitudes were not measured psychometrically, while Video_Stream combines several communication practices. Matching farm accounts with seller and buyer responses for the same period and channel would help trace how content and interaction lead to repeat orders. Distinguishing among provenance videos, consultations, and product promotions is particularly relevant because their costs and roles in the purchasing process differ.

Separate country teams introduce possible interviewer and language effects alongside recall and disclosure reluctance. The available questionnaire clarifies intended wording; no archived cross-language validation establishes identical implementation. Gender, respondent responsibility, and online-selling experience are also missing, so owner demographics cannot substitute for characteristics of the person answering. Future fieldwork should retain the language versions, common accounting instructions and contact-outcome records, alongside firm and cooperative identifiers, to separate measurement differences from substantive market differences.

Accounting comparability remains important. Three Chinese records require reconciliation because reported marketing expenditure exceeds the total costs implied by ROS for a common period and sales coverage. Excluding them preserves the between-sample margin gap but alters some Chinese coefficients, particularly those for video interaction. Certification was reported by respondents rather than independently verified. Strong expenditure–area multicollinearity constrains the Ukrainian model, in which the marketing coefficient changes sign across specifications. More budget variation among comparable farms, larger platform groups, and separate measures of video frequency and format would strengthen subsequent estimates.

Panel data would bring these extensions together. Monthly observations of the same farms, including budgets, revenue, returns, commissions, and service time, would distinguish persistent producer characteristics from seasonality and the sequence of marketing decisions. Dates of platform changes, the introduction of video selling, and logistics disruptions would support analysis of adaptation over time. A panel, combined with well-defined comparison groups and field experiments, could then examine the causal effects of specific changes in product presentation and ordering.

6. Conclusions

Telephone interviews with 186 farms indicate higher mean monthly online revenue in the Chinese sample—USD 12,878.30 versus USD 4237.30—and a higher reported ROS in the Ukrainian sample—30.85% versus 17.72%. The 13.13-percentage-point margin gap persists across sensitivity checks. The contrasting profiles underline the need to manage the expansion of demand and the costs of serving it together.

The comparison is explanatory in purpose and exploratory in design: it identifies associations and financially distinct sample profiles, without claiming national representativeness or a causal effect of country. Its empirical contribution is to place online revenue and retained profit share within the same farm-level assessment, showing why greater sales scale need not coincide with a higher margin. This extends the focus from participation in e-commerce to the financial conditions of continued participation and directs attention to the work and expenditure required to sustain sales.

Bringing RBV and TCE into a common financial assessment provides the study’s theoretical contribution. It connects the resources committed to attracting buyers with the organization of the work required to complete sales. The distinction between resource availability, resource deployment, and value retention helps explain why marketing expenditure and platform labels should not substitute for measures of digital capability or governance efficiency. This perspective defines a focused agenda for testing how campaign execution and channel costs shape retained income, alongside the observed financial outcomes.

The proposed positive marketing–revenue association is not robust across Ukrainian specifications and is not statistically supported in China. Neither primary model supports video moderation or within-country platform differences in ROS at the 5% level. Changes in budgets, video practices, and platform choice should consequently be tested at the farm level against the additional profit from completed orders.

Because revenue and ROS cover all online channels, the platform estimates describe principal-channel groups of farms. The findings support more detailed measurement of how marketing practices are executed, while the wide interaction intervals leave practically meaningful differences open. The following transfers are proposals for evaluation rather than estimated improvements.

Chinese cooperatives could trial producer-specific product lines within their existing platform sales. The INBERRI case offers a practical basis: a recognisable producer, regional provenance, a clearly organized assortment, and direct contact. A farm profile, batch information, and documented production attributes could strengthen the offer’s identity. A trial should compare the branded line with the existing offer by considering profit after all costs, repeat orders, and customer service time.

Ukrainian producers could adapt the organization of video selling described in Chinese research on skill formation and knowledge exchange [16,17]. Product demonstrations, responses to questions, stated availability, and clear ordering terms should operate as a single process. Comparing video-, photographic-, and text-based offers within comparable budgets would establish their relationship to completed purchases, returns, and labor requirements. Scaling is justified when the additional financial return exceeds the costs of content and fulfillment.

Both approaches require full channel-cost accounting, covering customer acquisition, packaging, delivery, commissions, returns, and labour. This provides a basis for adopting specific practices that fit the farm’s needs, combining a distinctive offer with convenient ordering and reliable fulfillment. The practical contribution is a decision criterion for farms, cooperatives and support agencies: assess digital expansion through the additional profit from completed orders after all channel costs, rather than through turnover or online activity alone. This focuses support for digitalization on producer economic viability, the dimension of sustainability examined here. Effective digital selling culminates in a profitable completed order, not simply a product view.