Improving Consumption and Purchases of Healthier Foods in Retail Environments: A Systematic Review

This review examines current research on manipulations of U.S. food retail environments to promote healthier food purchasing and consumption. Studies reviewed use marketing strategies defined as the 4Ps (product, price, placement, promotion) to examine results based on single- and multi-component interventions by study design, outcome, and which of the “Ps” was targeted. Nine electronic databases were searched for publications from 2010 to 2019, followed by forward and backward searches. Studies were included if the intervention was initiated by a researcher or retailer, conducted in-store, and manipulated the retail environment. Of the unique 596 studies initially identified, 64 studies met inclusion criteria. Findings show that 56 studies had at least one positive effect related to healthier food consumption or purchasing. Thirty studies used single-component interventions, while 34 were multi-component. Promotion was the most commonly utilized marketing strategy, while manipulating promotion, placement, and product was the most common for multi-component interventions. Only 14 of the 64 studies were experimental and included objective outcome data. Future research should emphasize rigorous designs and objective outcomes. Research is also needed to understand individual and additive effects of multi-component interventions on sales outcomes, substitution effects of healthy food purchases, and sustainability of impacts.


Introduction
The promotion of healthy purchasing in shopping environments is a focal point of public health and research efforts aimed at reducing obesity and improving health outcomes. In the U.S., 71.2% percent of adults and 41.0% of children ages 2-19 have overweight or obesity, a condition that increases risk for cardiovascular disease, cancer, and diabetes [1,2]. Recent examination of American diets found most Americans eat more total calories, saturated fat, salt, and added sugar than they need, and do not consume enough fruits and vegetables, and whole grain products [3]. The majority of food purchasing occurs in supermarkets, which are uniquely positioned between the consumer and food purchasing decisions [4]. In addition to providing access to food, the in-store food retail environment is recognized for its influential role in dietary outcomes [5]. In-store, food retail interventions influencing the food purchasing decisions of consumers have grown in popularity over the past 10 years. This shift is in part due to the popularity of behavioral economics as a foundation by which customers may be "nudged", though indirect suggestions, toward healthier products [6,7]. Most commonly, research on in-store approaches is characterized by the 4Ps of marketing (product, price,

Inclusion Criteria
The studies included are original empirical research published between 2010 and 2019, in English, and from the United States. Studies were researcher-or retailor-initiated, conducted inside the retail environment, and manipulated the retail environment. Evaluations could be quantitative or mixed methods and all interventions had to include at least one of the following outcomes: (1) Purchasing-related (i.e., objective store sales data, objective food purchasing data, customer receipts, and survey self-reported purchases or expenditures, store sales, or intent to purchase), and/or (2) consumption-related (i.e., food frequency questionnaire (FFQ), 24-h dietary recall, food diary, Veggie Meter TM or other biometrics, or other survey self-reported diet/consumption or intent to eat).

Exclusion Criteria
Interventions were excluded if they were implemented by an entity other than a researcher or retailer (e.g., price intervention at the wholesale level or front-of-pack labels initiated by a food company), if they did not occur inside the retail environment (e.g., restaurants, schools, mobile food trucks, online, and laboratory), or if they did not manipulate the retail environment (e.g., grocery store tours).

Search Terms and Databases
Nine databases (i.e., Academic OneFile, Business Source Premier, CAB Abstracts, Communication and Mass Media Complete, Family and Society Studies Worldwide, PsycINFO, PubMed, Sociological Abstracts, and Web of Science) from a variety of sectors (i.e., agriculture, business, communication, health, and psychology) were searched. Key terms were constructed based on three concepts: (1) Healthier food, (2) study design, and (3) setting. A variety of search terms were used to ensure articles would be included with nuanced differences in terms (e.g., healthy food vs. better-for-you) across sectors. The following key terms were used in all databases: Healthier food "health* food*" OR "healthy eating" OR "fruit*" OR "vegetable*" OR "low* fat" OR "low* sodium" OR "low* sugar" OR "low-fat" OR "low-sodium" OR "low-sugar" OR "better for you" OR "nutritio*" Study design "intervention" OR "pilot" OR "experiment*" Setting "supermarket*" OR "grocery store*" OR "corner store*" OR "bodega*" OR "retail environment"

Procedure of Article Search
RefWorks database was used to organize all articles. The searches were conducted by two authors and yielded 1231 studies (see Figure 1). After excluding 635 duplicate articles, two authors reviewed each full-text article to determine eligibility and excluded 548 studies. This review yielded 42 articles that met all inclusion criteria. Then, citation and bibliography searches were conducted with all 42 articles identifying an additional 22 articles for a final total of 64 articles (see Table 1).
After removing duplicates, two reviewers independently screened the title, abstract, and full text of the remaining 596 articles. Reviewers discussed any differences and consulted a third reviewer, when necessary, and a consensus was reached. One reviewer conducted forward and backward searches of the included articles. Titles and then full texts were reviewed to assess eligibility. Articles were abstracted and coded independently with two coders; discrepancies were discussed until a consensus was reached. Article abstractions included participants, study design, intervention description, 4 Ps, intervention setting, duration of intervention, data collection methods, outcome variables, and key findings. Our research reviewed studies and categorized them according to the 4 Ps: Product, price, promotion, and/or placement. Examples of interventions that were classified as product included determining how many and how much variety of a product to stock. Interventions that examined price included strategies such as price reductions and coupons. Furthermore, examples of interventions classified as promotion included shelf labels, recipe cards, and taste tests, and examples of placement strategies included altering the in-store location of products, such as moving to an endcap or to eye level. Our review included an examination for biases, with a focus on research design (eliminating confounders) and measures (i.e., self-report vs. objective data). Bias was assessed using the principles laid out in the Cochrane risk of bias tool [15].

Features of Included Articles
The primary intervention sites varied in terms of store size and included supermarkets (43.8%), corner stores (31.3%), grocery stores (26.6%), and/or convenience stores (9.4%) (see Table 1). Experimental designs accounted for about one-third (35.4%) of available studies, while the remaining were pre-experimental (33.8%), quasi-experimental (27.7%), or time series (3.1%). The most frequently used objective outcome data were store sales data (46.9%), while self-reported purchasing or expenditures was the most frequently used self-report measure (40.6%). Intervention length varied from 22 min to 3.5 years. Most studies (89%) incorporated promotion as a key component of the intervention, although efforts to address product (34%) and placement (31%) were also prominent. Relatively few interventions focused on price (16%). A total of 56 of 64 studies (87.5%) had at least one positive effect. When considering only objective measures of sales and more rigorous methods of determining dietary intake (i.e., 24 h recalls or biometric data), 100% (14 out of 14) had at least one positive effect.

Single-and Multi-Component Interventions
Thirty interventions were classified as single-component interventions because they only manipulated one of the four Ps, while 34 interventions were classified as multi-component. Over the past 10 years, the number of single-and multi-component interventions have both slightly increased (see Figure 2).

Features of Included Articles
The primary intervention sites varied in terms of store size and included supermarkets (43.8%), corner stores (31.3%), grocery stores (26.6%), and/or convenience stores (9.4%) (see Table 1). Experimental designs accounted for about one-third (35.4%) of available studies, while the remaining were pre-experimental (33.8%), quasi-experimental (27.7%), or time series (3.1%). The most frequently used objective outcome data were store sales data (46.9%), while self-reported purchasing or expenditures was the most frequently used self-report measure (40.6%). Intervention length varied from 22 min to 3.5 years. Most studies (89%) incorporated promotion as a key component of the intervention, although efforts to address product (34%) and placement (31%) were also prominent. Relatively few interventions focused on price (16%). A total of 56 of 64 studies (87.5%) had at least one positive effect. When considering only objective measures of sales and more rigorous methods of determining dietary intake (i.e., 24 h recalls or biometric data), 100% (14 out of 14) had at least one positive effect.

Single-and Multi-Component Interventions
Thirty interventions were classified as single-component interventions because they only manipulated one of the four Ps, while 34 interventions were classified as multi-component. Over the past 10 years, the number of single-and multi-component interventions have both slightly increased (see Figure 2).

Product
Of the 30 single-component interventions, only one intervention manipulated product [40]. The study utilized a pre-experimental design and found positive effects on produce sales after increasing stocking and availability of fresh produce [40].

Placement
One study implemented a placement-only intervention [23]. This experimental study had mixed effects (positive + null + negative). Positive effects were found such that featuring healthy products in aisle endcaps increased sales of these healthy products. However, when healthy products and indulgent products were featured together in aisle endcaps, sales of indulgent products increased while healthy products did not increase [23].

Price
Three studies implemented price-only interventions [18,29,33]. One study had mixed effects (positive + null + negative) [29], while another had mixed effects (positive + null) [18] and one study had only positive effects [33]. Both experimental studies provided a 50% discount for fruits and vegetables [18,33] and found that customers who received the discount purchased significantly more fruits and vegetables than customers who did not receive the discount [18,33]. However, one study also found no sustained effect on participants' spending on fruits and vegetables from baseline to follow-up period [18]. In addition, one study used a pre-experimental design [29].
Four studies used experimental designs [17,18,22,34]. The interventions focused on nutrition shelf labeling [18], food samples [34], nutrition education [17], and a smartphone app [24]. One study found positive effects on fruit and vegetables purchases [34], while another study found mixed effects (positive + null) on food purchasing (e.g., positive effects on servings of fruit and no effect on servings of vegetables) [17]. Two studies found mixed effects (positive + null + negative) for the change in consumption and purchases of products authors classified as healthier (e.g., fruits, vegetables, and whole grains) as compared to products identified as less healthy (e.g., higher calorie products and sweets) [24] and for change in the sale of popcorn using different nutrition shelf labels [22]. One example of a study with largely positive results used a combination of shelf labels (e.g., "healthier option," "low sodium") in combination with education about the labels [17]. Positive effects were found such that customers purchased more servings of fruits and dark-green/bright-yellow vegetables. However, there were no significant differences between the groups on saturated fat, trans fat, and servings of vegetables [17].
Eight studies focused only on promotion, utilizing quasi-experimental designs [25,26,30,31,[36][37][38][39]. Of these, four studies tested shelf labels [30,36,37,39], one examined the effectiveness of nutrition information labeling [26], one utilized a mass media campaign [25], one tested food demonstrations [38], and one examined the ability of increased stocking and promotions to sell healthy items [31]. Three found positive effects on purchases of healthier products [36,37,39] and one found positive effects regarding self-reported fruit and vegetable consumption [38]. One study had null effects on healthy food purchases and consumption, using both self-report measures and a skin carotenoid test [31]. Another study found negative effects on the demand for healthy popcorn [30]. Two studies found mixed effects (positive + null + negative) on sales of milk and the influence of caloric information on purchases [25,26].

Other
In addition to the 4Ps, two studies did not fit into the standard 4P framework and therefore were classified as "other" [16,35]. Both utilized experimental study designs. One examined the effects of ambient music [16] and the other study analyzed effects of ambient scents [35]. Findings showed mixed effects (positive + negative) as lower volume music increased healthier purchasing patterns and higher volume music increased unhealthier purchases [16]. Additionally, findings showed positive effects when using an in-store indulgent scent (i.e., chocolate chip cookies), which led to increased purchasing of healthier foods, and decreased purchasing of unhealthy foods [35].
An experimental design was used in only one study [46]. The intervention included a food marketing campaign (inclusive of food demonstrations, recipe cards, and an audio novella) featuring fruit and vegetable characters in tiendas [46]. Positive effects were found on daily fruit and vegetable intake but not variety [46].
Three studies employed quasi-experimental designs [53,54,58]. One intervention manipulated the in-store location of produce (i.e., moving pre-packaged produce near checkout lines), added shelf labels, and distributed recipe cards [58]. Another intervention focused on the effects of promoting meal bundles through in-store displays [54], while another examined the effects of pre-packaged produce packs moved to aisle endcaps packages [53]. One study found that shoppers who were exposed to the intervention were more likely to purchase produce [58], and another found that moving the pre-packaged produce near checkout lines increased healthy purchasing [53]. However, displaying meal bundles was ineffective in increasing healthy item sales [54]. One study used a pre-experimental design [57].
Two studies with time series designs addressed the effects of using behavioral nudges [56] and implementing a healthy food kiosk coupled with food sampling [55]. Results showed positive effects for healthy food sales when multiple behavioral nudges were implemented simultaneously [56] and when food sampling was combined with featured food kiosks [55]. Null and negative effects were found for healthy item sales when intervention tactics were isolated as well as among certain foods [55,56].

Promotion and Price
Only one study examined the effects of promotion and price and it used an experimental design [52]. The intervention examined the effects of healthy food consumption education and coupons with mixed effects (positive + null) on healthier purchases. Combining education and coupons was the most effective intervention group for increasing healthier purchases while null effects were largely observed for education and coupon only groups [52].
Five studies were experimental [59][60][61][62][63]. The interventions included adding point-of-purchase promotions, changing the store structure and environment (e.g., adding a buffet bar or refrigerator, grouping products in a display), and altering the in-store location of products (e.g., multiple facings, prime placement, secondary placement, checkout aisle end-caps), and increased stocking of healthier products [59][60][61][62][63]. All five studies found mixed effects for improving the purchasing and consumption of healthy food. For example, Foster and colleagues (2014) implemented an intervention to increase the purchases of specific healthier foods through shelf tagging promotions and by altering the shelf placement of products [59]. In intervention stores, sales of 2% milk, whole milk, two targeted cereals, and one of three promoted frozen meals remained the same, while sales of skim milk, 1%, and two out of three frozen meals increased [59].
Four studies utilized quasi-experimental designs [65,[67][68][69]. Two studies added point-of-purchase promotions, changed store structure and environment, altered in-store location, and increased stock of fresh produce [67,69]. Another study introduced healthier products to checkout lanes and added point-of-purchase promotions [68], and another changed store structure, increased media coverage about healthier choices, and offered in-store education sessions. Two studies found null effects on consumption and purchasing of fruits and vegetables [67,69], one found mixed effects (positive + negative) on consumer purchasing of healthy foods in healthy vs. standard checkout lanes [68], and one found positive effects of store owners' perceptions of changes in sales of promoted healthy foods [65]. Of these four quasi-experimental studies, two interventions were Proyecto MercadoFRESCO [67,69]. Both studies found null effects, such that there were no significant differences in consumption of and dollars spent on fruit and vegetables [67,69].
Six studies in this category used a pre-experimental [71,73,74,[76][77][78] design. Similar to previous studies, strategies added point-of-purchase promotions, changed store structure and environment, altered in-store location, and increased stock of fresh produce [71,74,[76][77][78]; one study implemented these strategies and paired urban farms with corner stores such that corner stores sold products obtained from urban farms [73]. Three studies found positive effects on purchases, sales, consumption, and intent to purchase healthy food [71,73,76].
Of the three studies, two studies used a quasi-experimental design [64,66]. Both were multifaceted interventions that included increased stocking of healthy foods, point-of-purchase promotions, and price reductions/incentive cards [64,66]. One of the studies found when shelf labels were consistently used (high fidelity), positive effects on sales of the promoted, healthy items were found [66]. The second quasi-experimental study found mixed effects (positive + null): shelf labels on healthy items led to participants purchasing more promoted foods but did not change consumption. However, the study authors did not observe changes in healthy food consumption. Finally, one study used a pre-experimental design [72], with mixed results.

Promotion, Placement, and Price
Two studies examined the effects of promotion, placement, and price strategies [70,75]. One study found null effects [70] and the other found positive effects [75]. Both studies used similar interventions, Plate It Up Kentucky Proud [75] and Plate It Up [70], which added point-of-purchase promotions, altered product placement, and offered coupons and discounts [70,75].
One study used a quasi-experimental design [70]. The results showed null effects on fruit and vegetable consumption. The study authors found no difference in the percent of food purchasing dollars spent on fruits and vegetables between control and intervention groups [70]. In addition, Liu and colleagues (2017) used a pre-experimental design [75] and found that recipe cards had a positive effect on customers' purchases of recipe ingredients and increased consumption of fruits and vegetables [75].

Intervention Including 4 Ps
Finally, only one study utilized all four Ps [79]. The study used an experimental, participatory design and found null effects for fruit and vegetable consumption. However, there was a significant decrease in the consumption of some unhealthy foods (e.g., chips) [79]. The intervention increased stocking of healthy foods, altered the in-store environment, added point-of-purchase promotions, and included discounts [79].

Discussion
This review, which examined the scope and impact of in-store marketing strategies related to healthy food sales, purchasing, and measures of diet, yields several important conclusions. One key finding of this recent review of literature is that both single-and multi-component interventions have become equally common focal points of research. Approaches provide evidence that increasing access to healthy food products in stores, particularly while utilizing promotion strategies, increases healthy food sales, purchasing, or improves dietary outcomes. While prior reviews found that positive outcomes were more common in studies utilizing multiple Ps [12,13], ours found more parity, even when considering the level of rigor applied to research designs and outcome measures. Overall, positive results were found in 27 of 30 single-component interventions as compared to 29 of 34 multi-component interventions, despite that multi-component interventions reported results related to a higher quantity of outcome measures.
Promotion efforts, including shelf labels, call out messages, and sampling products, continue to show promise as an important mechanism to improve purchasing. In-store promotion interventions are increasingly common, often with positive effects, either in combination with other approaches, or used alone. Previous reviews have found that older interventions, specifically those prior to 2008, were more likely to manipulate promotion, most often in single-component interventions [9,11]. In the more recent studies examined in this review, promotional interventions were frequently paired with placement and product strategies in multi-component interventions, for example including the coupling of a shelf labeling intervention with an end of aisle display, yielding positive effects.
Prior literature has identified multi-component interventions' added complexities in deciphering effects of its individual components [4,11]. There are two reasons for this complexity. One is the layered nature of multi-component interventions which by definition result in activities such as taste-testing, coupled with an end-cap placement and a shelf tagging, which make it difficult to decipher how components work together or separately to influence purchasing. It is possible for example that similar effects could be seen from just a single-component intervention, rather than multiple, though such impacts are difficult to decipher. Future multi-component interventions should consider alternative research designs where elements of the intervention are incorporated at different times and in different combinations, and then removed and then incorporated again in order to understand collective and individual effects, such a 2 × 2 factorial design or an ABA design [80].

Limitations and Future Directions
Of 64 studies reviewed, 24 in total (38%) were conducted without a control or comparison group. Only 14 of the 64 studies were experimental and included objective outcome measure data. The lack of a control group in more than one-third of studies displays the limitations of food environment research. Studies conducted with control groups, using store sales outcome data, and using rigorous dietary outcome measures are needed. Further research is also needed to better understand the individual and additive effects of multi-component interventions on outcomes like product sales.
The literature is limited in its ability to capture the extent to which increased healthy food sales results in overall less healthy food purchases. While several studies examine interventions in terms of specific product substitutions, for example by testing whether promoting a healthier item in a category results in changed sales in that product and a less healthy alternative (e.g., replacing higher fat popcorn with low-fat popcorn), few studies examine how targeted product sales relate to sales in other product categories (i.e., a spillover effect; e.g., increase in fruit sales associated with increase in low-fat dairy sales). Future research is needed to understand how increases in healthy food purchases do or do not serve to substitute for less healthy foods.
In addition to better understanding the marketing mechanisms that work best to shift purchasing, future research should examine the extent to which interventions yield sustained effects. Our review found that less than 20% of studies examined impacts beyond three months and only 4.5% considered impacts beyond one year.
It is unclear how the current COVID-19 context will continue to impact in-person food sales as compared to online sales and the extent to which product promotion and placement strategies can, or will, translate into online environments. Future work should seek to better understand how online food purchasing environments, including virtual supermarkets and real-world e-commerce platforms, can incorporate the four Ps to increase access to affordable foods.

Conclusions
Efforts to improve consumption and purchases of healthier foods in retail environments are diverse, even within the framework of the 4Ps. Considering these marketing strategies, this review found that promotion was the most commonly utilized strategy for single-component interventions, and manipulating promotion, placement, and product was the most common strategy used for multi-component intervention. In addition, interventions included in the review often employed pre-experimental or quasi-experimental research designs and relied more on self-report data rather than objective data. New research should implement interventions using rigorous designs and objective outcomes in order to advance the field. Further, given the large proportion of studies that implemented multi-component interventions, research is also needed to understand the individual and additive effects of approaches that use more than one of the 4Ps on objective sales outcomes, substitution effects of healthy food purchases, and the sustainability of impacts.
Author Contributions: A.K. contributed to the study concept and design. K.M. was responsible for screening. K.M. and H.W. extracted and coded the data, analyzed the data, and drafted the manuscript. A.K. critically reviewed all drafts. A.K. and K.G. approved the final version submitted for publication. All authors have read and agreed to the published version of the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.