How do sparkling wine producers adopt a sub-appellation? Evidence from an exploratory study on heroic Prosecco Superiore Rive

This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as: Barisan L., Galletto L. (2021), How do sparkling wine producers adopt a sub-appellation? Evidence from an exploratory study on heroic Prosecco Superiore Rive, Wine Economics and Policy, Just Accepted


Introduction
Over the last few decades, the international wine market has dramatically changed by, inter alia, increasing the number of appellations and sub-appellations. Under the pressure of several market and socioeconomic forces, the Italian wine supply has followed these developments towards a reference model of higher quality wines [1]. As evidenced by Scozzafava et al. [2], Italian wine legislation has designed a model that differs from the French model based on the hierarchical territorial classification, where vineyards or groups of vineyards are typically recognized for quality. Therefore, a new appellation (or sub-appellation) faces difficulties in attempting to become successful mainly for institutional, wine labelling, and market differentiation reasons [3][4][5].
However, some of these new designations may build a reputation of excellence as an acclaimed and successful collective brand [6].
In this context, Conegliano Valdobbiadene Prosecco Protected Designation of Origin (CVPP) appellations are encompassed within the most prominent case studies deserving of attention, with a supply of over 91 million bottles in 2018 [7][8][9]. The CVPP is made up of 15 districts (municipal areas), and it takes its name from the two main towns of the zone. It represents the top PDO for Prosecco, as it is rooted in a specific tradition, rural heritage, terroir, and landscape. Prosecco wine is made from the Glera grape variety and is obtained using the Martinotti method after secondary fermentation in pressure tanks. The Prosecco Reform (2009) provided by the Italian Ministry of Agriculture, upgraded the former CVPP's from DOC (Controlled Designation of Origin) to DOCG (Controlled and Guaranteed Designation of Origin), and instituted the new Prosecco DOC, which covers nine provinces belonging to Veneto and Friuli-Venezia Giulia regions [10]. It is the largest Italian wine PDO for the number of bottles (500 million bottles in 2020), most of which are exported worldwide [11].
The CVPP hierarchical quality model is in turn classified into three geographical levels. These levels are Superiore di Cartizze sub-appellation (SA), which is locally considered in the same manner as the first-growth or grand cru and represents 1.7% of the CVPP supply; Rive SA, a second-growth wine or small SA, which accounts for a proportion of 2.9%; and the basic Prosecco 4 Superiore (PS), (i.e., the classical CVPP sparkling wine), which is the most widespread (95,4%) [12].
The first sub-appellation (1.5 million bottles sold) indicates absolute top-quality wines within the CVPP appellation. It comes from a subzone that covers just 107 hectares of vineyards in the borough of Valdobbiadene, where the maximum yield allowed by the disciplinary system is 12 tons of grapes per hectare. It has not changed its territorial boundaries since the CVPP was set up in 1969. Its supply has stabilized since the 1980s. Hence, the alternative decision to produce Superiore di Cartizze has substantially no chance of being developed by the CVPP's sparkling wine houses. Therefore, it can be excluded from the choice set due to the current CVPP's disciplinary rules. In contrast, a major choice in adding value to CVPP supply chains is through Rive SA's adoption strategy by replacing PS production. In July 2009, the Prosecco Reform introduced the Prosecco Superiore and Rive SAs, both stemming from the CVPP's Spumante (sparkling wine) DOC [10]. Production for the former is approximately 83.8 million bottles, and the maximum yield allowed is 13.5 tons of grapes per hectare. Rive SA production and yield lie between those already described for Superiore di Cartizze and Prosecco Superiore. In 2018, the Rive SA covered an area of approximately 249 hectares with a supply equal to 2.7 million bottles sold. Interestingly, over the 2010-2018 period, the number of bottles claimed under the Rive SA increased at a double-digit annual growth rate (15%), which is almost three times higher than that of the PS.
The term "Rive" indicates, in the patois of the local inhabitants, small parcels of steeply sloped vineyards that are characteristic of the area and where the best-quality grapes are produced (sensu stricto). This category of wine highlights the different expressions of the CVPP. Rive wines are often obtained from grapes grown in the steepest, highest-quality vineyards in a single borough or hamlet, thus emphasizing the characteristics that a terroir gives to the wine. The concept of the Rive SA as a brand arose as an answer to the awareness of the need to link the image of a wine to its terroir to highlight the synergies between soil, weather, grape and winegrower [13].
Within the CVPP appellation, 43 Rive wines are now allowed, of which 12 get their name from their borough and 31 from their borough's hamlet. Each terroir expresses a different and specific combination of soil, exposure, microclimate, and human factors. In the Rive, yields are limited to 13 tons of grapes per hectare, the grapes are picked exclusively by hand, and the vintage must be shown on the label. The Rive SA represents a viticultural potential of 83.7% of the CVPP area, of 5 which 48.4% falls within the UNESCO "Core Zone," and the rest falls within the so-called "Buffer Zone" (Figure 1). The marketing literature on geographical indications is rather vast and has mostly focused on the consumer side [2,14,15]. However, less attention on the supply side has been paid to subappellations or small appellations within larger ones. The soundness of SAs proposed for the Niagara Region of Canada has been investigated from both a consumer viewpoint [5,16] and a terroir perspective [17]. In Europe, Gergaud and Ginsburgh [18]   Rive SA's disadvantages over PS. The price differential between Rive SA and PS in the grape and base wine markets should be greater to properly sustain both higher labour intensity and its expanding effect among producers. Yet, compared to PS, the spread of Rive SA is more linked to the domestic market than to exports; given its recent creation, the promotion of Rive SA has not been established as a PS. Ultimately, consumers' knowledge of PS in a broader sense is stronger than that of Rive SA [27].
The paper is organized as follows. Section two presents the theoretical approach. Section three relates to the methodology and the data employed. Section four addresses the results, and section five discusses those results. Final considerations conclude the work.

Theoretical approach and hypotheses
As argued by Rogers [28], the adoption of Rive SA can be contextualized as the process of deciding on the introduction of an innovation. The entrepreneur, to start with, goes through a phase of acquiring knowledge about the rules of the product specification and then the assessments of conditions for its claims and of market demand, which leads to the formation of a positive or negative attitude towards the Rive SA when ultimately deciding whether to adopt the new SA.
The review of the literature has widely investigated the relevance of factors affecting the introduction of new geographical indications for wines. For instance, the process can be influenced by a wide variety of patterns [29][30][31][32], including cultural and psychological factors (e.g., belonging to a community of heroic winegrowers, personality, empathy); structural factors (e.g., land under cultivation, human capital, production size); marketing factors (firms' entries into new channels and markets, price positioning, branding); factors related to the development of wine tourism (e.g., visitor reception, wine events); and political factors (e.g., differentiation strategies implied by the Tutelary Consortium at the territorial level).
Thus, a broader framework can be applied to study the model for the adoption of SA by firms in the CVPP territory [33,34]. The framework can identify group variables derived from the structure-conduct-performance model [35] that can influence the process by which firms adopt the Rive SA as a relevant innovation, i.e., the structural, marketing, and wine tourism contexts.
This design is in line with similar research and strategies implemented in the wine industry [19,20,36,37].
This study considers that the development of the conceptual adoption model assumes that the differentials in the vineyard yield per hectare, average selling prices and production costs between the PS and Rive SA in current and future years are known with certainty to winegrowers. Therefore, one can assume that the decision has low risk and uncertainty to properly portray the Rive SA adoption decision process; this process is aligned with the CVPP Consortium's aim to support the added value of the heroic viticulture wines of the Designation as a UNESCO World Heritage Site.

Rive SA adoption
Structural variables. According to the existing literature and empirical evidence, structural resources have been consistently identified as important factors for the adoption of a SA [13,26,38]. Winery size is supposed to affect the choice of introducing the Rive SA. However, the decision 8 is challenging. On the one hand, large wineries have more possibilities to diversify their portfolios than small ones do. On the other hand, small wineries are more focused on local consumers who may be more interested in terroir features. Given this ambiguity about the size effect, we include various size indicators in the model: the number of bottles, PDO surface, pressure tank capacity and three categories of employees. Considering the human factor, a younger entrepreneur is thought to be more likely to push for the adoption of new SAs. The amount of CVP bottled by third parties should imply a lower incentive to use SAs, as more third-party bottling indicates a weaker tie to the winegrowing area. A similar consideration may apply to the quantity of purchased grapes to be crushed in the winery, given that this can be an obstacle to a strong supply identity based on a winery's own grapes. Therefore, it is reasonable to hypothesize the following: H1: Firms with higher levels of human factors, owned structural endowments, and commercial size are more likely to adopt Rive SA than firms with lower levels.
Market variables. The company's decision to adopt Rive SA may also be influenced by market or distribution channel conditions [22,39,40]. A possible impact on Rive SA adoption may derive from a higher tendency to favour selling to the domestic market in comparison with the export market, as Rive SA is arguably more acknowledged and appreciated by Italian consumers. The degree of use of some marketing channels may also favour the Rive SA; e.g., a large share of Prosecco sold by winery shops or in the Horeca channels, where quality and reputation affect the outcome more than in other outlets, would imply higher interest for the Rive SA, while the opposite would happen for a large share sold to the mass market, i.e. Large scale retail (LSR). Hence, we can hypothesize the following: H2: Firms with higher Italian market shares and higher penetration rates in the wine shops or in the Horeca channels are more likely to adopt the Rive SA than firms with lower shares and rates.
Tourism variables. Since "Rive" sparkling Prosecco is linked more strongly to terroir, viticulture tradition and landscape than Superior Prosecco is, we can expect that a strong involvement of wineries in wine tourism may increase the probability of including this sub-brand in their portfolio [41]. To verify and clarify this aspect, we test the number of visitors and propose the hypothesis 9 that the greater the involvement in certain events, the greater the likelihood of adopting the Rive SA [42,43]. Therefore, we hypothesize the following: H3: Firms with higher levels of involvement in wine tourism and particularly in events organized on-site are more likely to adopt Rive SA than firms with lower levels.
Although this paper's main issue concerns Rive SA adoption by CVPP wineries, a secondary field of investigation is the intensity of this adoption, i.e., its share of total CVPP sparkling wine sales.
In particular, we are interested in verifying the effect of two variables on it: the share of direct sales on total sales and winery size. Given previous considerations, the former is expected to affect Rive SA market share, while the latter should reduce it.
The lack of investigation on SA choice makes our analysis explorative. Therefore, in achieving the two goals of our research, we have not estimated models based on specific sets of variables; rather, we have attempted to select sets of variables among those suggested by the previous theoretical considerations.

Quasi-census study and data collection
The study is based on 2017 data, and the sample is made up of 158 wineries that produce sparkling CVPP, 38 of which have chosen the Rive SA. The data were collected through an ad hoc survey using the listed CVPP's sparkling wine house members. The survey showed a very high response rate (over 89%), which was yielded by face-to-face interviews with the business owner and/or the firm's management representatives while ensuring that gathered data were treated anonymously with confidence and sensitivity. Although some firms were initially less inclined to participate in the survey, they were persuaded to take part after some telephone reminders.
The population distribution was analysed using data coming from the certification bodies in charge of Prosecco production control [11] by using strata based on bottled production. The variable distribution from this database was compared with our dataset by considering the size of the bottled production sold annually (standard = 0.75 litres) and avoiding omitting the largest companies.
Hence, our data represent a quasi-census study having specific representativeness of the surveyed population (Mooi et al., 2018). The remaining share of sparkling wine houses (11%) did not 10 respond to the research because they were either too busy to participate or not available to provide the requested information.

Variables
The following is the list of dependent and explanatory variables that were used in the analysis for this paper ( Table 1).
The dependent variable is a discrete binary variable that is assigned a '1' if the firm has already adopted the Rive SA; otherwise, the firm has adopted the PS (Prosecco Superior SA firm) and is thus assigned a '0'.
The explanatory variables were grouped into three components. The first group concerns structural variables, the second group of variables relies on the conduct of firms in the industry by markets and sales channels, and the third group considers the performance of wine-and tourism-related events [44].
Regarding structural assets, marketing skills (i.e., young commercial employees, young CEO, oenologist, young owner), physical and technological resources (i.e., Glera Docg surface, purchased grapes, own wine production, storage capacity of pressure tanks, under-contract bottling,) and firm size (i.e., small-, medium-, large-and very large-) have been analysed and implemented following a criterion that focuses on the importance of human capital and firm resources as crucial dimensions in Rive SA's innovation processes. All the variables are numerical, except for both owner and size, which were transformed into factors with two and four levels, respectively.
In the framework of market conduct, the model used tries to capture, in terms of the competitive strategy, the effects of Rive SA on market share of CVPP firms in Italy and abroad and on the domestic market through distribution channels (i.e., wine shops, Horeca, large-scale retail, wholesalers, e-commerce and others) and major export markets (i.e., Germany, the United Kingdom, Switzerland and the United States).
Wine tourism events were assessed considering: a) actions aimed at promoting CVPP wines and b) their performance in pursuing wine tourism policies. Drivers have been defined to capture the effects of how important the economic return of major wine events was (Likert scale from 1 to 5 points) and the number of visits per winery [45]. The major wine events, ranging from domestic to international, with a significant impact on CVPP firms, were as follows. Conegliano  Bottles produced (in thousand) Bottles sold (in thousand) 1=small-sized (less than 150,000 bottles sold), 0=otherwise 1=medium-sized (150,001-500.000 bottles sold), 0=otherwise 1=large-sized (500.001-1.000.000 bottles sold), 0=otherwise 1=very large-sized (more than 1,000,000 bottles sold), 0=otherwise [33] [33] Self-developed Self-developed Self-developed Self-developed [33] Self-developed Self-developed Self-developed Self-developed [7,46]

The model specification
To find the determinants of Rive SA choice, we estimated a logistic regression, which is a model broadly employed to examine the factors that affect a binary outcome such as undertaking an action [49,50].
The logistic regression model allows us to identify the variables that have the most impact on the choice to use the Rive SA for branding. We modelled the probabilities of the outcome based on producer characteristics and marketing behaviours. More specifically, this technique determines the significant drivers for classifying a winery as belonging to the Rive group or to the other group.
Therefore, it offers a "prognosis" (or propensity) relative to adopting the Rive SA. where βk is the estimate of the k th parameter and Xki is the k-th characteristic of the i th winery.
Although, as we discussed earlier, economic theory may suggest some reasons in favour of the Rive SA choice, we have not made any a priori assumptions about which variables should be included in the model.
Furthermore, a generalized linear model approach was used to determine the main drivers for Rive SA market share by employing the same candidate variables as those of the logistic regression [51,52]. Therefore, the estimations of both models rely on an exploratory stepwise procedure.

The data analysis
In our initial table of descriptive statistics, we present t-tests of differences in variables according to Rive SA or non-Rive SA.
In the stepwise logistic regression model-building procedures, backward selection, rather than forward selection, has been used to avoid the so-called suppressor effect [50,53]. As suggested by Snipes & Taylor [54], to discover the best logistic regression model, Akaike's criterion was used to support the model choice. As argued by Bendel & Afifi [55], to ensure less risk of failure when trying to find a relationship between explanatory and dependent variables when one exists, the usual p < 0.05 statistical significance criterion has been relaxed to 0.10. In the tables of results, we have reported not only the estimated coefficients and associated odds ratios but also the marginal effects of each variable. The marginal effects are the change in the probability that a winery chooses to produce Rive SA sparkling wine due to a unit change of a specific independent variable. To estimate the marginal effects, we have followed the approach where marginal effects on the binary dependent variable are computed by using the command 'margins' in Stata, as suggested by Cameron and Trivedi [56].
A generalized linear model (GLM) was fit using the maximum pseudolikelihood algorithm to estimate Rive SA's market share, which is its proportion on total wines sold in the firms' portfolio of Prosecco wines. This variable was rescaled into a range between 0 and 1 1 . We used a binomial family GLM, link to logit, following the method proposed by Papke and Wooldridge [57], which was subsequently and particularly enhanced in Stata by Baum to handle with fractional response data [58].
To model the data, we jointly considered the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) [54] [59].

Results
The results that examine the hypotheses of the study are presented considering a) the direction and significance of the differences between the average value of the variables for adopting and not adopting Rive SA and b) the contribution of the various independent variables to explain the adoption of Rive SA using logistic regression. Findings about the main drivers of Rive SA market share conclude the paragraph.

Differences between Rive SA adopters and non-adopters
The results of the t-test for the homogeneity of the means are shown in Table 2.
First, the differences between the average value of the group of Rive SA adopters and that of nonadopters are positive for some crucial structural variables, while they are negative for a firm's own-grapes crushed and small-size firms. Among the former, the differences between the variables' average values are statistically significant for PDO area under cultivation, higher firm commercial size (i.e., large, and very large firms), number of winery employees and young commercial and sales staff, young owner, share of production under contract, own grapes crushed, and winery pressure tank sizes. Second, when detailing the analysis of the leading foreign markets, only the US shows a significant difference that is higher for Rive SA. Only the difference in large-scale retail market share is negative and almost significant. Third, the level of involvement in communication activities such as Vinitaly and Primavera del Prosecco (i.e., Prosecco Spring) show significant differences between the groups of firms.
Therefore, these first outcomes appear to confirm hypotheses 1 and 3 previously formulated (see paragraph 2) but seem to be inconclusive for hypothesis 2. The following estimates from multivariate logistic regression contribute to providing further insight into the relationship between the set of independent variables and the dependent binary variable.

Factors influencing Rive sub-appellation adoption
The results of logistic regression are presented in Table 3. According to Pregibon [60] and Mehmetoglu and Jakobsen [61], the assumptions that the model is good and correctly specified were The larger the Prosecco base wine production is, the more likely the Rive SA choice is; i.e., one hundred thousand hectolitres increase the probability of adopting the Rive SA by approximately 7 percentage points. The presence of a young owner was a less significant variable (p = 0.060). Wineries with the last feature have a propensity to adopt the Rive SA that is 2.7 times greater than for those who do not. Marginal effects show that, ceteris paribus, the probability of the Rive SA, when a young owner is present, increases by 11 percentage points.
Among the independent variables relating to market share, only the use of LSR appears to affect the propensity to adopt Rive SA by undermining it, although less relevantly than the other factors. More specifically, the analysis of the marginal effects for different values of LSR share shows that when this variable is above 10%, the probability of the Rive choice becomes almost zero. It may be argued that the role of distribution channels is overshadowed by the firm size, given a sort of channel specificity according to the size. However, we found no large variance inflation factors (all lower than 3), suggesting that collinearity was not a substantial problem between firm size and distribution variables 3 .
Primavera del Prosecco, which is a proxy for wine tourism events, is positively and significantly related to Rive SA adoption. For wineries rating Primavera del Prosecco from important (3 points) to extremely important (5 points), the propensity to use the Rive SA increased by a minimum of 6 to a maximum of 9 times in comparison to wineries poorly scoring this event 4 . The marginal effects show that the probability of adopting the Rive SA increases by 21-25% if the winery is involved in the Primavera del Prosecco and gives this event an importance score equal to or higher than 3 points. 3 The variance inflation factors for reported distribution channels (2.48 for Direct sales; 1.80 for Hotellerie-Restaurant-Café and Wine Bar; 1.90 for Large retail scale) when firms' size was included in the model (1.76 for medium-sized; 1.52 for large-sized; 2.92 for very large-sized) were quite small. 4 Using the Primavera del Prosecco score as multiple binary variables has performed better than dealing with it as a single categorical variable.  [62] and Pregibon [60]. The goodness-of-fit test was tested following Hosmer et al. [63]. Akaike's information criterion (AIC) and Bayesian information criterion (BIC). Sensitivity = 21 firms out of 38 (55.3%), specificity = 113 firms out of 120 (94.2%), positive predictive value = 21 firms out of 28 (75.0%), negative predictive value = 113 firms out of 130 (86.9%), predictive accuracy = 134 firms out of 158 (84.8%).

Drivers of the Rive SA market share
The goodness of the link test has provided evidence that the model was correctly specified concerning linear predictors (p = 0.000), regardless of whether the regression equation specification error test indicates that the model has no omitted variables. Other statistic indicators appear satisfactory as well; among them, results did not show a large value of the condition index (17.1), implying that multicollinearity is not a problem [64]. Table 4 presents the GLM regression results for Rive SA market share determinants.
First, the coefficient related to small-sized firms exhibits a positive sign and significant relationship with market share (p = 0.006), which supports the statement that smaller sizes create greater value in wine portfolios through the growth of Rive SA.
Second, the results show positive and significant effects of direct sales (p = 0.006), while for the role of Italian market, a negative impact is shown, which indirectly confirms that product growth is currently not linked to other domestic channels, which do not have closer relationships between producers and customers as much as direct sales [65]; Finally, the positive coefficients of the organized wine tourism events suggest that the Prosecco Road and the Primavera del Prosecco are positive and significant tools (p = 0.061 and p = 0.036, respectively) to promote growth and achieve the strategic objectives of Rive SA firms.  [59]. The goodness-of-link test was calculated according to Tukey (1949) and Pregibon [60]. Multicollinearity diagnostics followed procedures found in Belsley, Kuh, and Welsch [64]. Akaike's information criterion (AIC) and Bayesian information criterion (BIC).

Discussion
The three hypotheses stated at the beginning of the work find partial confirmation from the t-test analysis. The logistic regression model reinforces the validity of H1 and H3 by deepening the effect of the most explicative variables.
Given that Rive SA is a type of marketing innovation, younger entrepreneurs are more inclined to adopt it in comparison with those who are older. Moreover, it seems the only human factor that matters in the Rive SA choice. Among structural factors, both basic wine production and marketed bottles play the most important role. We observe that as winery size increases, it is more likely that the winery will include the Rive SA within its wine portfolio. When the number of end markets, channels or consumer segments grows, the need to rely on a wide diversified assortment of Prosecco by not simply using the traditional residual sugar content (brut, extra dry, dry) drives a winery to explore other products, such as those that can be produced in the Rive SA. Both the marketed bottles and basic wine enlighten a specific contribution to the model. In fact, to understand how they can coexist in the logistic model, we have to consider that the latter can only partly be transformed into bottles of sparkling CVPP to be sold by the winery: part can be sold as bulk wine to plain bottlers [5], part can be reclassified and sold under other appellations. As expected, the purchase of grapes from other estates does not favour a propensity for Rive SA adoption. The analysis of marginal probabilities for different percentages of purchased grapes shows that, for medium-high percentages, the likelihood the winery uses the Rive SA is reduced to nearly zero. Therefore, it seems quite clear that Rive SA is viewed as being strongly rooted in the terroir of the firm's vineyards and is closely linked to the concept of "estate-bottled" sparkling Prosecco.
If the t-test results show that wineries choosing the Rive SA are significantly more involved than others in the first Italian wine event (Vinitaly), where both wine differentiation and portfolio diversification play a crucial role in achieving successful public relations, the logistic model underlines that they are even more involved in the wine tourism business. In fact, the propensity to adopt Rive SA increases when wineries give a rating of 3 or more for the main wine tourism event, i.e., the "Primavera del Prosecco".
With reference to market features, hypothesis 2 seems to be rejected because neither the effect of the domestic market share, nor the penetration rate in the wine shop nor in the Horeca channels can be assumed to be significant factors driving the Rive SA choice. Only the share of bottles marked through LSR is somewhat important for this choice. Hence, because the share of bottles sold by other channels does not seem to significantly affect the choice of SA use, we can conclude that the distribution policy can be an obstacle to it only when the winery has a nonnegligible interest in the mass market. A similar consideration may be applied for sales in specific foreign markets. Their shares do not significantly influence the choice, a fact that could partially depend on the relatively new definition of the Rive brand and knowledge of it based mainly on local consumers.
Regarding the main factors affecting Rive SA market share, we observe that the two hypotheses previously formulated are well confirmed. Once a winery has chosen the Rive SA option, its weight in the total Prosecco wine supply increases with its cellar door share, which depends both on local customers and wine tourists, while it decreases with the firm's size. Therefore, considering previous results, we can conclude that while winery size positively influences the adoption of SA, once it has happened, it plays a negative role in the weight of the Rive SA in the winery sparkling wine assortment, where the Rive SA has to compete with an increased number of CVPP labels as the firm size grows. Moreover, being too focused on other domestic channels other than direct sales, where the interest in a SA subdivision is likely to be lower, may divert attention from improving the SA proportion in the CVPP portfolio.
Involvement in wine tourism is confirmed to be important in also determining the SA share. In fact, in addition to the variables included in the logistic model, the high scores attributed to both the Prosecco Wine Route and "Primavera del Prosecco" show a significant impact on the dependent variable. Hence, we can argue that, as the French wine classification system shows [66], a process of intense subdivision within a geographic region is much more effective when it is accompanied by sound wine tourism perspectives, which rely mainly on a considerable number of people willing to obtain a remarkable experience in the wine world.

Conclusions
This survey highlights the relevance of structural, market-related and wine tourism-related features in explaining SA adoption choice by measuring how these features can contribute to supporting and valuing a sparkling wine produced with considerable efforts by heroical vine-growers (Rive SA) within the CVPP.
Among them, we want to stress the increased likelihood of using Rive SA as business size (however measured) grows. If the linkage with wine tourism involvement was rather expected, this was not straightforward given that it could also be supposed that small wineries focused on niche markets would have better appreciated the new SA as a tool to better differentiate their small sparkling Prosecco products from that of other producers [67]. We can hypothesize that a core business based on a consolidated simple portfolio and a reduced level of flexibility in comparison with larger producers may contribute to explaining this empirical evidence.
Our study suggests the CVPP wineries key-conditions for benefitting of the Rive SA: -to be strongly rooted in their land (i.e., to be as much as possible self-sufficient for grape production); -to undertake a coherent strategy, based on the values of a heroic SA coupled with the prestige of the Unesco World Heritage, whereas the values preservation of the local culture is a core subject to work for the common good of that wine community; -to promote the knowledge on Rive SA, by boosting sales especially at the cellar door, which represents the best channel to communicate consumers the values embedded in the Rive brand; -to get actively involved in crucial wine tourism events to build and operate higher value-generating positioning (i.e., Primavera del Prosecco, Prosecco Wine Route, etc.).
In a scenario where Prosecco is becoming increasingly popular worldwide, the knowledge of factors that make Rive SA adoption likely may help the Tutelary Consortium define an appropriate promotion strategy to widen the use of Rive SA among CVPP producers as a tool for further differentiating their Prosecco along with their own brand. It should be based on four aspects to be considered: -to looks to a model for economically and technically sustaining heroic viticulture and vine-grower's backbreaking work, to preserve socio-economics and cultural values, landscape and biodiversity values [68][69][70]; -to implement distinguishing promotional activities with in-person information, exclusively in the cellar door and in Horeca channels, which are the most relevant in generating higher value for consumers [71]; -to improve the ties between Rive SA and landscape values, which can be grasped passionately by the territorial firms, thanks to key wine tourism events promoted in the Prosecco Hills of Conegliano and Valdobbiadene (i.e., Conegliano Valdobbiadene's Wine Festival, UNESCO World Heritage Site's Association, etc) [45]; -to maintain careful and close control of the quality from the grapes selection phase to the sales phase (i.e., through ad hoc members' courses and training, etc).
Our study has some limitations. First, this research is focused on CVPP firms that produce sparkling wines in a SA in Italy. Second, while the most significant drivers in Rive SA adoption were identified, they do not represent the whole of sparkling wine production in Italy or, for example, sparkling wines in other areas with steep-slope viticulture. Third, further research on this topic is necessary over time to capture changes in firm strategies depending on circumstances (i.e., due to the market or the Protected Designation of Origin's product specification rules) that can enable or constrain SA adoption.
Finally, the results should be viewed as a first step in the attempt to build a theory of SA economics concerning drivers that support the firm's choice of a SA. We believe that our study provides useful intuitions for those who would broaden this research strand in different countries with other specificities.