Scent in Mail: The Effect of Scented Direct Mailings

The purpose of the present study is to test the effect of scent on response time and sales in direct mailings. It is proposed that congruent scent influences response time and sales in direct mailings. A total of 1571 direct mailings were sent out.The hypotheses were tested through a Mann-Whitney U test and a Wilcoxon signed-rank test. The results show that scent influences response time to direct mailings but that it does not influence sales.Future research should perform exploratory studies to test the effect of scent with regard to various forms of direct marketing and offerings as well as in general product and service contexts. Managers may want to consider scenting direct mailings when a fast response is critical. The result provides empirical support for the idea that scent has an effect on response time in direct mailings.


Introduction
Increasing competition inmany markets hasforced managers to find new ways to efficiently reach consumers (Morrison, Gan, Dubelaar, & Oppewal, 2011). One recent approach that companies have begun to use in order to tailor their offerings and their communication is sensory marketing (Spence, 2012). Sensory marketing is aimed at stimulatingthe consumer through the five senses: sight, hearing, touch, taste and smell (Lindstrom, 2005). Stimulation of consumers' senses can evoke feelings and memories, which in turninfluencesconsumer behavior (Hultén, Broweus, & Dijk, 2009). Certainly, many practitioners are aware that sensory marketing can have an effect on sales, product evaluations and customer satisfaction (Bellizzi, Crowley, & Hasty, 1983;Turley & Milliman, 2000) (cf. Davies, Kooijman, & Ward, 2003;Mattila & Wirtz, 2001;E. Spangenberg, Crowley, & Henderson, 1996). Researchers have manipulated sensory stimuli such as music, colors and various shop fittings (cf. Turley & Milliman, 2000). In comparison to such stimuli, scent can be closely associated to specific products (Parsons, 2009) (cf. Odeh & As'ad, 2014). The aroma offreshly baked breadin bakeries or butter popcorn at movie theaters is closely associated with products, and the scents are perceived by many consumers as natural elements of the environment (cf. Bone & Ellen, 1994;Verma, 2014). However, the effects of scent have often been over looked in research on sensory marketing (Bone & Jantrania, 1992;Gulas & Bloch, 1995;Hultén et al., 2009)(cf. Cirrincione, Estes, & Carù, 2014), and very much so in research on direct mailings.
Research on scent can at least be traced back to the early 1930's, when it was found (Laird, 1932) that a clear majority of subjects preferred socks scented with narcissus over of unscented socks. More than three decades later, a similar study (Cox, 1967), this time using the scent of oranges, found that almost 90 percent of the respondents preferred the orange-scented socks over the unscented duplicates, as the scented socks were perceived to be better quality (Cox, 1967). In more recent studies it has been found that, for example, ambient congruents cents can increase consumer brand recall (Morrin & Ratneshwar, 2000). However, one problem of the previous research is that it was almost entirely limited to products, brands or venues (Teller & Dennis, 2012). Research that extends beyond product-specific environments (bakeries, cinemas, flower shops, etc.) is relatively limited (Bone & Ellen, 1998;Orth & Bourrain, 2005). To get a broader picture, further research should be performed outside the mentioned settings (Teller & Dennis, 2012). Indeed, the need for more research on the effect of scent has been noted (Knasko, 1995;Morrin & Ratneshwar, 2000)(cf. Jacob, Stefan, & Gueguen, 2014Spangenberg et al., 1996).One area thatremainsunexploredis directmailings.Direct mailings is a massive industry estimated to have generated, for example, 12 billion U.S. dollars in the U.S. and 2 billion GBP in the U.K.in 2014 (Statista, 2014), and importantly, the design of direct mailings is considered to be critical to their success (Feld, Frenzen, Krafft, Peters, & Verhoef, 2013). Consequently, this study tests the effect ofscent onresponse time and sales in reply to direct mailings.

Theory and Hypotheses
Within basic marketing, direct mailing is described as a cost effective way for businesses to reach out to targeted markets. Direct mailingmay also be referred to as direct marketing, which includes several approaches ranging from email, web ads, and television commercials,newspaper ads and billboards (Kotler & Armstrong, 2013). Naturally, direct mail is often perceived as unrequested mail where the sender has the intent to sell or provide information about products and services (Chang & Morimoto, 2003)(cf. Kumar & Sharma, 2014); therefore, it also often is perceived as unwanted (Kotler & Armstrong, 2013). As a result, the effectiveness of direct mailings is not a given. Managers want to make direct marketing more effective(cf. Durango-Cohen, Torres, & Durango-Cohen, 2013;Feld et al., 2013;Greenfield, 2004;Lorenxi, Friedmaun, & Paolillo, 1988;Malthouse, 1999). However, while recommendations on how to make direct marketing more effective are abundant and sometimes contradictory, empirical evidence is limited. There is empirical evidence for the importance of the visual design of the envelope as a driver of opening frequency. But the opening of a direct mailing envelope is merely a must for a response to the mailing; it is not in itself causing the response rate (Feld et al., 2013). Indeed, to improve response rates managers are increasingly making use of integrated campaigns, combining direct marketing with other channels (cf. Cooper, 2010). Developments in the field of direct marketing and its various offshoots are moving towards more specialized, differentiated approaches (Cooper, 2010), and one fruitful avenue may involve sensory marketing and scent.
Smell has been considered to be one of the most enigmatic senses in humans (Hultén et al., 2009). Unlike sight, touch and hearing, the sense of smell (along with taste) is a chemical sense(e.g. Morrin & Ratneshwar, 2000). That is, the sensory receptor cells in the nose are activated by chemical molecules(e.g. Cain, 1988). Although human survival no longer depends on the sense of smell,it is still very important. Children learn early to use scent cues in their environment, including the pleasant scent of fresh food orthe unpleasant odors such as those of foul food (Goldstein, 2009). Studies on this fundamental level has found that in general, a pleasant fragrance produces approach behavior while an unpleasant odor leads to avoidance behavior(e.g. Levin & McBumey, 1986;Takagi, 1989).
(1998)revealed thatambientscents made respondentssolve mathematical problemsmore quickly. And BaronandKalsher(1998) found that a pleasant scentcan improvesome aspectsof performance with regard to driving a vehicle. Likewise,physical performancehas been shown toimprove withthe help ofscent; Raudenbushetal. (2001)showed that the scent ofpeppermintcouldincrease physicalperformance inathletes. In terms of consumer behavior, Spangenberget al., (1996) testedwhetheran ambientscentin a store would affectconsumer perceptionof time and their number of purchases. The resultsshowed that with a scent present, consumers had a shorter perception of time, but there was nosignificant effect on sales. Orth & Bourrain (2005) found thatthe scent of lavenderhada positive effect onconsumerperception ofa brand,but thatrisk-taking,and again, saleswere not affectedsignificantly (cf. Herrmann, Zidansek, Sprott, & Spangenberg, 2013;Madzharov, Block, & Morrin, 2015;Suha, Moonb, Han, & Ham, 2014). Additional research has shown that mood can be influenced by means of scent (Diego et al., 1998). Studies by Roberts & Williams (1992) and Warm et al. (1991) revealed that respondents' mood was positively affected by subjecting them to different scents such as vanilla, chamomile and peppermint. Lorig& Schwartz (1988) found that subjecting respondents to the scent of eucalyptus, lavender and apple increased their brain activity, decreased stress levels and ultimately had a relaxing effect. Such findings contribute to a relatively uniform perception that fragrances can have an effect on people's mood and well-being (Ehrlichman & Bastone, 1992;Hultén et al., 2009;Lorig & Schwartz, 1988).
The SOR(stimulus, organism, response) model emerged in the 1960s against the background of findings in the social sciences, especially in psychology (Jacoby, 2002)(cf. H. Ehrlichman & Halpern, 1988. The model implies that environmental stimulus (S) causes emotional reactions in the organism (O) and thus influences the organisms' behavioral response (R)(e.g. Mehrabian & Russell, 1974). This logic is certainly in line with general behaviorism (cf. Skinner, 1974). However, the approach reflects ideas of hedonism and enjoyment as a major motive(H. Ehrlichman & Halpern, 1988)(cf. Goi et al., 2014). Specifically, the concept of stimulusorganism-response (SOR) has been deployed as a basis for explaining how pleasant fragrances and their effect can taint consumers' perceptions of products or services (Diego et al., 1998). According to the SORmodel pleasantfragrancesshould have a positive impactwhilstunpleasant odorsshould have anegative effect (Ehrlichman & Bastone, 1992). Over the years, the model has received backingfrom a variety ofresearch results (Chebat & Michon, 2003;Orth & Bourrain, 2005;Spangenberg et al., 2006;Turley & Milliman, 2000). For example, Chebat & Michon(2003) studied consumer behaviorin shopping centersand found thatpleasantscentsof peppermint, flowers, sandalwoodand citrushad apositive impact onhow consumersperceivedstore environmentand service. Likewise, the research of Orth & Bourrain(2005)showed that the scent of lavender had apositive effect onconsumers' perceptionof brands. Spangenberg et al. (2006)deployed the SORmodelas a basis for testing the effect of scent on variables such as perceived time spent in the environment, sales (number of items purchased as well as money spent).The results showedthat there was a positiveimpact onboth perceived time spent in the environment and sales (both on number of items purchased as well as money spent) (ESpangenberg et al., 2006). The results with regard to perceived time spent in the environment can be related to the research of Lipman(1990), who provides evidence for how pleasantscents may increase the timecustomersstayin a venue. Spragenberg(1996)relates the effect on perceived time to how consumers receiveanenhanced experiencewhen they visita storewith an ambientscent, which in turn contributes to their perception of time. Comparable effects on time, or on the perception of time,have been shown inother contexts, ranging fromin chessgames (Francis, 1987)to working hours (Lefevre, 1988).Likewise, a studydonebyHirschand Gay(1991) showed thatthetime spentin ascentedcasino was longer (compared to unscented), which in turnled toincreasedrevenue(cf. Knasko, 1995;Leenders, Smidts, & Langeveld, 1999;Nixdorf, Teerling, & Köster, 1992). In general, it may be that the effects of scent with regard time have to do with information processing. A study (Mitchell et al., 1995)showed that scent can affectconsumerinformation processing; added scent resulted in increasedtime spentprocessing when evaluatinga product, which in turn contributed toimproving the evaluation ofthe product.
When it comes to the relationship between scent and sales,it has been argued that hedonisticrelevance isrelatively weak (Turley & Milliman, 2000). The majority ofthe research in the field deployseitherthe SORmodelorcongruenceas a theoreticalstarting point (Morrison et al., 2011). Thus, as an alternative tothe SORmodel (and hedonistic relevance),several scholars deploy congruenceas a starting point for their reasoning(cf. Bone & Ellen, 1998;Bone & Jantrania, 1992;Mattila & Wirtz, 2001;Mitchell et al., 1995). Acongruentscentis consistent with theconsumer's perception ofwhat, for example, a productshouldsmell like (Parsons, 2009). A fragrancethat is not consistent is referred to as incongruent (Bone & Jantrania, 1992).Research that deployscongruence as a starting pointmeasures the effect of the scentin relation toits various contexts (cf. Teller & Dennis, 2012). Such approaches are often based on the congruence or incongruence of scent in relation toa particular product orservice.Ascent thatdoes not matchits context (e.g. a particular product)is referred toasincongruent.It is generally implied that a congruentscent hasa positive impact oncustomers' purchasing behavior and product evaluation (Parsons, 2009;Spangenberg et al., 2006).  (coconut and  lemon respectively)were deemed more favorableby consumersin comparison to incongruentcombinations (Bone & Jantrania, 1992). Several studies(e.g. Hultén, 2012;Morrison et al., 2011;Spangenberg et al., 2006) have deployed the scent of vanilla as an independent variable. It appears that vanilla can have a similar effect onboth men and women; it is congruent in several contexts andgenerally perceived aspleasant (Hultén, 2012;Spangenberg et al., 2006). Indeed, vanilla has been found to haveaslightly positive impact onconsumers'purchasing behavior (Morrison et al., 2011). Comparable results have been achieved with lavender scent (Guéguen & Petr, 2006). Researchhas shownthat acongruentscent can cause consumers toremain longerina store (Schifferstein & Blok, 2002). As mentioned, it is implied that consumersevaluate contexts (and products)more positivelywhena matching scent is present. In an opposite manner,anincongruentfragrancethat does not fitthe contextcauses confusion (Parsons, 2009;Spangenberg et al., 2006). However, it has been argued that in comparison to no fragrances at all, in congruent scents still can have a positive impact on consumer behavior (Parsons, 2009).That is, it is important to consider that evenincongruentscentshelpus to recognizeand reinforceexperiences in retail environments,which in turn can be utilizedtoinfluence consumer behavior (Davies et al., 2003).
From the preceding discussionit is clear that scent can influence the amount of time consumersperceive they spend or the amount they actually spend in a setting (cf. Francis, 1987;Lefevre, 1988;Lipman, 1990;Spangenberg et al., 2006)and the effect may be understood in terms of an enhanced experience (Spragenberg, 1996). In addition, increased time spent has been associated with increasedrevenue (Hirsch & Gay, 1991)(cf. Knasko, 1995;Leenders et al., 1999;Nixdorf et al., 1992). These effects of scent with regard to time may have to do with information processing. That is, added scent results in increasedtime spentprocessing when evaluatinga product, which in turn improves consumers' evaluation ofthe product (Mitchell et al., 1995). Therefore, in regard to the relationship between scent and reply time to direct mailings we propose:

Methodology
Selection: To compare the response time and sales in reply to unscented direct mailings to the response time and salesin reply to scented direct mailings, a total of1571 direct mailings were sent out to randomly selectedstart-up companies (they were allregistered at theCompanies Registration Officein April2014). Randomizationresulted in two samplesof n=793forthe experimental groupand n =778for the control group (the uneven numbers resulted from external falloff as a result of invalid orincorrect addresses revealed after the randomization process).

Measures
Scent: Two focus groups consisting of seven and eight participants respectively (five women and two men aged from 28 to 53 and three women and five men aged 20 to55) were deployed to identify and selecta scent thatwas consideredthe mostcongruenttothe setting, i.e. to the directmailings. The selection ofthe fragrances evaluated in the focus groups were all well-anchoredinpreviousresearch: vanilla,citrus, lavender, peppermint and eucalyptus (cf. e.g. Bone & Jantrania, 1992;Cain, 1988;Morrin & Ratneshwar, 2000;Morrison et al., 2011;Raudenbush et al., 2001). The scents were sponsored by a company specializing in exposure andscentingof different environments.The groups were introduced to the purpose of the sessions as well as to the format of the direct mailings. The participants received paper sticks upon which the scents had been applied. The discussion then concerned the contextswith which the scents were congruent as well as which scent, if any, fit with the direct mailings. Both vanilla and peppermint was deemed suitable. Therespondents associated vanilla scentwithpeace andcredibility whilepeppermintrepresented energyand spontaneity.In the end, peppermint was chosen as it was consideredthe mostcongruentwith the direct mailings. Next a procedure for scenting the direct mailings was developed. Specifically, the directmailingswere authentic ditto consisting of twopagesand apostage-paidreply form(all in a single envelope), withinformation about the companyand their offer (the offer concerned website development and hosting).A pretest was performed by sending scented mailings through the mail to make sure that they were scented upon arrival. The results were deemed appropriate.
Response time and sales: Response time and sales were recorded for the experimentalandcontrol groups respectively.Sales could be made in two ways. One option was to submitthe postage-paidreply formthat camewith the mailing.The other option was to go through thecompany websiteonline wherethe sameresponse optionswere available. The date of each purchase was noted to establish the response time. A purchasewas recordedwhen ananswer sheetwas received or when an approvedregistration was madethrough the website.

Procedure:
The directmailings weresent out and replies were tracked for 14 days (no replies were received after 14 days). The mailings were visually identicalexcept for thepersonalizednames and addresses. Directmailingsforthe experimental group were scented: Each mailing was sprayedon the frontand on the backleaving a distinctpeppermintscent. Each envelope in the experiment group was sprayedon the inside. Next, the mailings were put in the envelopes and the envelopes were sealed.

H1-Response time:
The1571 direct mailings generated a total of 43 sales during the 14 days sales were recorded; 20 sales in the experiment group and 23 sales in the control group, hence each day represented a measure. The normality of the distribution was assessed through probability plots, kurtosis and skewness. The z coefficients for both the kurtosis and the skewness were within +/-2,58 (Table 1). However, the sig. values of the Kolmogorov-Smirnov and the Shapiro-Wilk tests were below 0,05 (Table 2), signifying that the distribution was non-normal (e.g. Hair, Black, Babin, Anderson, & Tatham, 2006), and therefore A Mann-Whitney U test was used to test the hypothesis(cf. e.g. Pagano, 1994).  The Mann-Whitney U testshowed that scent did elicit a statistically significant effect onresponse time in reply to direct mailings(Z = -1,972;p = 0,049) (Two tailed). The results reject H0. (See Table 3 and 4)  H2-Sales: Because the dependent variable was measured at the interval level and the independent variable consisted of two matching pairs, and the distribution of the differences between the two groups was deemed adequately symmetrical, a Wilcoxon signed-rank test was deployed to test the hypothesis(cf. e.g. Pagano, 1994). See Tables 5 and 6. The Wilcoxon signed-rank test showed that scent did not elicit a statistically significant effect on salesin reply to direct mailings (Z = -,406; p = 0,685) (Two tailed). The results reject H1

Conclusion
The results show that adding scentto the direct mailings had an effect on the response time in reply to the direct mailings.Specifically, the response time was shorter when a scent deemed congruent with the direct mailing by focus groups was added. The results show that adding scentto the direct mailings had no effect on sales in reply to the direct mailings. Specifically, both in absolute numbers (20 sales in the experiment group and 23 sales in the control group) and sales ranked with each day as a measure, more sales were found in the control group. However, the difference was not statistically significant.
Discussion: The fact that scent had asignificant effect onthe response timeunderlines the general conception that scent is related to time. One obvious explanation for the relationship draws on the SOR model and that scent enhance the experience(cf. E. Spangenberg et al., 1996), which in turn influences the time spent with the direct mailing(cf. Francis, 1987;Lefevre, 1988;Lipman, 1990;E. R. Spangenberg et al., 2006). That is, the respondents may spend more time in front of the direct mailing at the moment when it is received and thus decide faster. As a consequence, and in practical terms, managers may want to consider scenting direct mailings when a fast response from the target market is critical (as a response or intervention tothe actions of competitors).However, this explanation struggles when the model is extended to suggest that increasedtime spentprocessing when evaluatinga product in turn improvesthe consumer evaluation ofthe offering(cf. Mitchell et al., 1995), at least in the sense that improved consumer evaluation of the offering implies, in the end, increased sales. Added scent had no effect on sales. According tothe SORmodel, a pleasant scentshould have a positive impact, and indeed, Chebat&Michon(2003)found that peppermint is considered to be a pleasant scent. Thus, as the result show that scent had nosignificant effect onsales, it limits the applicability of SORmodel with regard to sales. This result is in line with several earlier studies which found no effect of scent on sales (e.g. Mitchell et al., 1995;Orth & Bourrain, 2005;E. Spangenberg et al., 1996).
Another interpretation of the observed relationship between scent and response time draws on congruence. One the one hand, congruence should contribute to an enhanced experience, in turn leading to the increased  Spangenberg et al., 2006). Thus, both in terms of the SOR model and in terms of congruence, the results problematizeany association between increased time spent and increasedrevenue(cf. Hirsch & Gay, 1991;Knasko, 1995;Leenders et al., 1999;Nixdorf et al., 1992). Previous researchbyRaudenbushet al.(2001)and Baron&Kalsher(1998) showed thatpepperminthad anactivating effectand thatthe scentincreasedrisk taking. Hence an alternative explanation for the faster response time is that certain scents can elicit specific responses in consumers and that the chain of causality has little to do with the SOR model or congruence.
As mentioned, Spangenberg et al.(2006)found a positive impact of scent on both perceived time spent in the environment and sales (both on number of items purchased as well as money spent).However, it appears that a critical difference between Spengenberg and earlier studies is that the former deployed gender congruent scents rather than merely ambient scent(cf. Spangenberg et al., 2006).Thus, to make the results conform to the notion of congruence, it may be argued that the added scent in the present study wasnot congruent enough. That is, it may be that the scent was congruent enough to enhance the experience sufficiently to cause the potential buyers who would buy the product anyway decide or act faster. But the scent was not congruent enough to elicit additional sales. While deemed congruent by the focus group, peppermint may not be congruent enough with the direct mailings to influence sales. Consequently, an apparent limiting factor includes the potential for establishing scents strongly congruent with an offering, which in turn is decided by the characteristics of the offering, unless congruence is sought with the recipient. Congruence with the recipient could, for example, entail a male or female scent for a male respective female target market.
Theresults obtainedand the discussion above suggests severalfruitful avenues for further research.The possibility that certain scents can elicit specific responses in consumers and that the chain of causality may have little to do with the SOR model or congruence warrants more explorative studies on the effect of scents (including degrees of unpleasant scents) on direct mailings, regardless of the scents are deemed congruent or pleasant. In terms of congruence, more research is needed into what types of offerings or direct mailings allow the establishment of clearly congruent scents. Issues such as what scents are congruent with, for example, a book, a magazine, a music CD, a computer game CD, or with various consumer electronics, may depend on the media, the medium, and the target audience(this study did not take demographics into account). Various degrees of congruence with classes of offerings or with explicit contexts may be established. A related factor to consider is the level of involvement associated with the product or service being offered. An offering which is expensive and visible to others (e.g. cars, jewelry etc.) is generally a high involvement purchase. Cheap and non-visible offerings, such as for instance toothpaste and sugar, are generally low involvement purchases. It is implied that the higher the involvement, the more rational information processing is done by consumers before making a purchase(e.g. Bolfing, 1988;Celsi & Olson, 1988;Kapferer & Laurent, 1985;Sarathy & Patro, 2013), which could have had an impact on the effect of scent. In the present study, website development and hosting can be considered to be a relatively high involvement purchase decision, promoting more rational information processing in turn possibly limiting the effect of scent. Thus fruitful avenues for further research include taking into account the offering or purchase situation in terms of involvement, as well as congruence, and pleasantness or unpleasantness of the scent. Moreover, assessing differences with regard to factors such as gender, age orother nominalaspects in relation to response time is likely both to help elucidate the relationships and have practical value for the direct marketing industry.