Consumer Response to Alternative Selling Strategies: a Field Experiment

Arch G. Woodside, University of South Carolina (student), University of South Carolina
Robert E. Pitts,
ABSTRACT - The analyses of the separate and joint effects on sales of salesman perceived expertise, price, and the absence or presence of customer purchase pals are reported. The three main effects were statistically significant: sales increased with increases in salesman perceived expertise, the presence of purchase pals, and price increases ($1.98 to $3.98). The expertise-by-price interaction was statistically significant using multiple regression and multivariate probit analyses and four price treatments. The purchase pal-by-price interaction was statistically significant using three price treatments. The need for hypothesizing and empirically testing interaction effects on demand of marketing variables is emphasized.
[ to cite ]:
Arch G. Woodside and Robert E. Pitts (1976) ,"Consumer Response to Alternative Selling Strategies: a Field Experiment", in NA - Advances in Consumer Research Volume 03, eds. Beverlee B. Anderson, Cincinnati, OH : Association for Consumer Research, Pages: 398-404.

Advances in Consumer Research Volume 3, 1976      Pages 398-404

CONSUMER RESPONSE TO ALTERNATIVE SELLING STRATEGIES: A FIELD EXPERIMENT

Arch G. Woodside, University of South Carolina

Robert E. Pitts (student), University of South Carolina

ABSTRACT -

The analyses of the separate and joint effects on sales of salesman perceived expertise, price, and the absence or presence of customer purchase pals are reported. The three main effects were statistically significant: sales increased with increases in salesman perceived expertise, the presence of purchase pals, and price increases ($1.98 to $3.98). The expertise-by-price interaction was statistically significant using multiple regression and multivariate probit analyses and four price treatments. The purchase pal-by-price interaction was statistically significant using three price treatments. The need for hypothesizing and empirically testing interaction effects on demand of marketing variables is emphasized.

INTRODUCTION

Marketing decisions must he made in the context of insufficient information about processes that are dynamic, nonlinear, lagged, stochastic, interactive, and downright difficult (Kotler, 1972, p. xi).

Despite Kotler's famous quote of 1967 and 1972, few empirical studies exist on interaction and other effects on sales of pricing, product, promotion, and distribution levels which are downright difficult to explain or measure. Recently, Curhan (1974) and McCann (1974) used fractional factorial analysis of variance and regression analysis respectively to measure the main and interaction effects of multiple marketing decision variables; unfortunately, the discussions of their findings focused nearly exclusively on the main effects of the decision variables.

Kotler (1971) maintains that the market's response to variations in the level of any one marketing input is conditional on the level of the other decision variables. Furthermore, the variation of two or more marketing activities at the same time can have synergistic effects that are greater or less than the sum of the separate effects.

Some progress has been made on developing models of interaction effects on a conceptual level and measuring them on an empirical level (e.g., Kotler, 1964; Green, 1973).

This article reports on the analyses of the separate and joint effects on sales of two marketing decision variables and one consumer related variable. The analyses are based on a field experiment using two levels of a promotional variable, salesman expertise, and four price levels of a consumer semi-durable product. The presence or absence of other persons, i.e,, purchase pals, during the sales presentations was the consumer-related variable in the study.

HYPOTHESES

The general form of the hypothesized model of the effects on sales of salesman expertise, price, and purchase pal includes main and interaction effect terms:

Q = a + b1E + b2Pr + b3P + b4(EPr) + b5(EP) + b6(PrP) + b7(EPrP)    (1)

where:

Q = product sales

E = expertise

Pr = price

P = pal

bi = regression coefficients (B = standardized partial regerssion coefficients)

Intuitively, a positive main effect of perceived expertise attached to the salesman and sales while a negative main effect of price and sales would he expected.

H1: An increase in the level of perceived expertise attached to the salesman produces a greater likelihood of purchase by the customer.

H2: An increase in the level of price produces a decrease in likelihood of purchase by the customer.

As a rationale for H1, Woodside and Davenport (1974) have used Kelman's (1961, 1965) conceptualization of source credibility and internalization of the message by the receiver: the greater the communicator's perceived credibility by the recipient, the greater the likelihood that the recipient will accept the influencing message because it is congruent with the recipient's value system, i.e., internalization has occurred. Brock (1965) has also found that credibility of the source affected decisions made by customers (change from a customer's selected brand of paint to one advocated by the salesman).

The law of demand as expressed in H2 is open to challenge depending upon other factors operating in the market. Four possible reasons for expecting a positive price-demand relationship have been reviewed by Kotler (1971):

1. With the greater purchasing power following a substantial price reduction of a product purchased in substantial quantities, consumers may decide to switch to more expensive products, i.e., the income effect.

2. Higher price is likely to increase demand for certain goods because of the phenomenon of "conspicuous consumption," i.e., the Veblen effect.

3. An expectation effect: when a price reduction is seen as the beginning of a possible wave of further price reductions, many buyers may withhold their purchases.

4. The quality effect: in situations in which buyers are not well informed about the respective merits of competing products, and some risk is involved, buyers may take the price as an indication of quality.

Positive price and perceived-quality relationships have been reported in the marketing literature in situations involving low consumer knowledge and some risk. In a retail experiment, Woodside and Sims (1974) also report a positive price and quantity demanded relationship for a new product, an electric lunch box.

Intuitively, the presence of other persons accompanying a customer would produce a decrease in the effectiveness of a sales presentation, i.e., a decrease in the likelihood of purchase by the customer. A purchase pal may be viewed as the actor with the least at stake in the transaction, offering the customer a more unbiased view of the situation than the salesman; and therefore, the customer may be less affected by the sales presentation when a purchase pal is present since the pal may provide reasons for not purchasing the product.

H3: The presence of a purchase pal decreases the likelihood of purchase by customers following a sales presentation,

Some support of this hypothesis is provided by Bell's (1967) report that purchase pals (friends of relatives accompanying the customer) made automobile selling more difficult and less pleasant for the salesman.

H4: Decreases will be greater in the likelihood of purchase produced by increases in prices, the lower the level of perceived expertise of the salesman.

H5: Increases will be greater in the likelihood of purchase produced by increases in expertise if no purchase pal is accompanying the customer compared with increases when a purchase pal is present.

H6: Decreases will be greater in the likelihood of purchase produced by increases in prices if a purchase pal is accompanying the customer compared with decreases when no purchase pal is present.

H7: The presence of a purchase pal increases the changes in likelihood of purchase produced by price under low versus high perceived salesman expertise conditions.

Hypotheses 4 through 7 are interaction predictions of the three independent variable effects on sales. For specific price increases, the salesman of high versus low perceived expertise could decrease the customer's perceived risk in evaluating the product and consequently price changes should have a smaller impact on sales. This is the rationale for H4.

Increasing reliance in the salesman under high versus low expertise conditions is likely to occur when no purchase pal is present. If the customer is not able to seek help from others for making the purchase decision, the resulting internalization effect achieved by increasing the perceived expertise of the salesman has a greater likelihood of occurring. This is the rationale for H5.

Higher product prices may produce more comments by purchase pals of the economic risk in purchasing the product, thus the presence of purchase pals should increase the effect of price changes on sales, i.e., produce a greater price elasticity (M6).

The presence of a purchase pal provides the customer with an additional source of information which could increase the customer's perception of the product's price; consequently, changes in the likelihood of purchase produced by different prices for high versus low levels of salesman expertise are likely to be greater when purchase pals are present.

The seven hypotheses are shown graphically in Figures 1 through 4. The interaction hypotheses can be stated for the demand curve in the figures:

H4: [(Da-Db)/(Pr1-Pr2)] < [(Dc-Dd)/(Pr1-Pr2)]  from Figure 1.

FIGURE 1

HYPOTHESIZED EFFECTS OF PRICE AND SALESMAN EXPERTISE

FIGURE 2

HYPOTHESIZED EFFECTS OF EXPERTISE AND PURCHASE PALS

FIGURE 3

HYPOTHESIZED EFFECTS OF PRICE AND PURCHASE PAL

FIGURE 4

HYPOTHESIZED EFFECTS OF PRICE, EXPERTISE, AND PURCHASE PAL

H5: [(Da-Db)/(E2-E1)] > [(Dc-Dd)/(E2-E1)]   from Figure 2.

H6: [(Da-Db)/(Pr1-Pr2)] < [(Da-Db)/(Pr1-Pr2)]  from Figure 3.

H7: {[(Da-Db)/(Pr1-Pr2)] - [(Dc-Dd)/(Pr1-Pr2)]}/P1 < {[(Da-Db)/(Pr1-Pr2)] - [(Dc-Dd)/(Pr1-Pr2)]}/P2   from Figure 4.

METHOD

The data of this study are taken from a field experiment set in a retail store from which some results have previously been reported (Woodside and Davenport, 1975). The research design used included two levels of salesman-perceived expertise and four prices which enabled sales comparisons to be made between treatment levels and across treatments. In addition, the salesperson recorded the presence or absence of a purchase pal after the customer left the store. A control group of subjects received no salesman presentation with the product priced at $1.98.

A salesperson attempted to induce purchase of a new product innovation among consumers shopping for 8-track stereo music tapes in a small music store in Augusta, Georgia. The salesperson was a woman in her late thirties who was the communicator for all eight treatment conditions. The salesperson was a high school graduate and had not been trained in marketing theory or sales psychology. She was not informed of the predicted results of the study and no additional compensation was offered to her.

Product

The product was the "HC-2001, Head and Capstan Cleaner Kit," manufactured by Becht Electronics, Burbank, California. The kit included two felt pads, head cleaning solutions, and cartridge to be used to clean 8-track players.

The product had been introduced to the market during the month of the experiment and none of the firm's six competitors carried the product. The six competitors were visited during the course of the study by a shopper wanting to buy "some type of cleaning stuff or kit" for his tape player. Two competitors had knowledge of such devices but stated that demand was not great enough to carry them.

The product was somewhat technically complex and its safe use was assumed to he important to a user since the product had to be connected to the tape player. The kit was mounted on cardboard, enclosed in a plastic container, and had a suggested retail price of $1.98. Operating instructions were printed on one side of the cardboard.

Procedure

The salesperson attempted to induce customers, who had just purchased one or more tapes, to make an additional purchase of the cleaning kit. Selected customers were randomly assigned to one of the eight treatment conditions and to the control group. Treatment conditions were typed and copies placed below the cash register in the store after the copies were randomly mixed. Space was available on the copies to record purchase information. Blank copies represented the control group assignment.

Customers examining 8-track tapes were selected unobtrusively as subjects in the experiment. While a selected subject was examining the tapes, the salesperson looked at the top treatment copy under the register and administered that particular treatment. A display box containing 20 of the cleaning kits was placed near the cash register throughout the experiment. A 6" by 6" card appeared on the display box. The words "8-Track Tape Cleaner Kit" and the price were placed on the card. Different cards were used for the different price treatments. Prices were not listed on the cleaning kits. Purchase response was recorded immediately after the customer left the store. The copy of the treatment administered remained in its top position until the next subject was selected. The salesperson removed the previous customer's treatment copy at this time, noticed her role for the new subject, changed the display card if necessary, and administered the treatment when the subject approached the cash register.

The treatments consisted of the eight combinations of the four price levels and the two expert - non-expert levels. A total of 30 customers were assigned to each combination and to the control group. Purchase behavior was the only dependent variable measured in the study.

The following price levels were included in the study: $1.98, $2.98, $3.98, and $5.98. This wide range in prices was used to attempt to insure for some significant price effects in the experiment. Hunch lead to the belief that demand would not be affected for a wide range of price. The product's price for the control group was $1.98.

The salesperson expressed prior purchase by herself of the musical tapes bought by the customer in all treatment conditions, except for the control group. Levels of perceived expertise were defined as the salesperson's oral instructions on how to operate the tape cleaner versus her expressed inability to operate the cleaner.

The salesperson's presentations to the customers were independently observed by one of the experimenters. The prior decision was made to discard the data for the first ten customers given the presentations. A number of small operational problems were corrected in these initial presentations and the salesperson became proficient in delivering the sales messages.

Customers in the control group did not receive any sales presentation but could purchase the product from the display box on the counter near the cash register. Any questions asked about the product by these customers were answered by the salesperson. No customer was rejected for requesting information. The salesperson responded to questions in the treatment role as required. The salesperson recorded whether or not the customer requested additional information. Few customers requested further information.

Store policy was to accept cash, check, or charge card and no discrimination was made for method of payment.

Further details of the field experiment procedure and specific wordings of the expertise conditions are given by Woodside and Davenport (1975). Different operational definitions of expertise could have been used in the experiment which implies theoretical and managerial constraints on examining the findings.

Pre-Test of Appeals

Perceived differences of the communicator's message for the expert and non-expert treatments were examined in a pre-test. Students in one class at the University of South Carolina rated the salesperson using the messages on the following qualities for 7-point Likert scales: cold-warm, non-expert - expert, makes me not want to buy-makes me want to buy, and salesperson not familiar with my needs-salesperson is familiar with my needs.

The written appeal given the students began with the following:

Assume you have just purchased some 8-track stereo tapes, and as you are paying for them, the salesperson says:

The students were randomly assigned to receive one of the treatments by randomly mixing the copies of the appeals. The students were told that answering the questions was not connected with the course, not to place their names on the paper, and their questions would be answered afterwards. The $1.98 price was used for both treatments in the pre-test.

All results were either statistically significant (p < .05, t-tests) or supported preconceived hypotheses in directions of mean differences from scores on the Likert scales. The salesperson in the expert treatment was statistically significantly rated more expert, makes me want to buy, and familiar with my needs compared with the salesperson in the non-expert treatment. Further details of the pre-test are discussed in Woodside and Davenport (1974).

RESULTS AND DISCUSSION

The complete set of data of the experiment are presented in Table 1. The three main effects of expertise, price, and purchase pal were statistically significant. Results are shown in Table 2 of chi-square analyses of these main effects.

TABLE 1

CUSTOMER PURCHASE BEHAVIOR FOR FOUR PRICE, TWO SALESMAN EXPERTISE, AND TWO PURCHASE PAL CONDITIONS

TABLE 2

MAIN EFFECTS OF PRICE, SALESMAN, EXPERTISE AND PURCHASE PAL ON CUSTOMER PURCHASE BEHAVIOR, IN PERCENT

The first two hypotheses were supported by the analysis. The increase in perceived salesman expertise produced a substantial increase in sales (30%). Only 8.3% of the customers in the $5.98 price treatment purchased the product, while 55% of the customers in the $1.98 treatment purchased the product. However, the differences in percent of product purchases between the $1.98, $2.98, and $3.98 treatments were slight. More than 50% of customers in these treatments purchased the product.

Counter to the third hypothesis, a significant positive relationship between purchase pals and customer purchase behavior was found. Over 61% of the customers shopping with purchase pals purchased the product versus 32.5% purchases when the purchase pal was not present. Two post hoc explanations of this finding can be suggested. A purchase pal may provide an immediate means for social validation (Howard, 1965) that the purchase is a correct choice and thereby reduces the perceived risk in the act of purchase and increases the likelihood of purchase. Secondly, the presence of a purchase pal provides the customer with a more conspicuous role to play in deciding whether or not to purchase the product; thus, the customer may experience a greater need to provide reasons for not purchasing the product. Consequently, the purchase of the product may be the easier action to take for the customer.

Multiple regression analysis and multivariate profit analysis were used to test the significance of the interaction terms as well as the terms for the main effects in equation 1. A quadratic term for price (Pr2) was included in the equation because a curvilinear relationship was believed to likely result from the wide range of prices used in the experiment ($5.98 -$1.98).

The dichotomous dependent variable was nonpurchase versus purchase. Effect coding (-1, +1) was used on levels of expertise and no purchase pal versus purchase pal. Effect coding is similar to dummy coding except that the former generates regression coefficients which reflects the linear model. The standardized partial regression coefficients of the independent variables produced using effect coding are equal to the correlation coefficients (phi-coefficients) of the independent and dependent variables, if the independent variables are orthogonal (Woodside, Pitts, and Gewirtz, 1975).

Multivariate probit analysis was used to overcome some of the limitations of using multiple regression analysis on a dichotomous or ordinal level dependent variable. Probit analysis assumes a linear effect on each independent variable as well as a series of break points between categories for the dependent variable. Maximum likelihood estimators are found for these coefficients, along with their asymptotic sampling distributions, as well as an analogue of R2 is defined to measure goodness of fit (McKelvey and Zavoina, 1975; Kau and Hill, 1972).

McKelvey and Zavoina (1975), and Bettman (1974) review some of the problems of using multiple regression analysis with a dichotomous dependent variable: heteroscedastic disturbances, expected value of the error term greater than zero, and non-normally distributed error term. However, multiple regression is quite robust and departures from its underlying assumptions are unlikely to produce substantially different conclusions in the comparison of the relative importance of the independent variables. The partial regression coefficients and R2 estimated by multiple regression are more conservative than the coefficients of the coefficients and R2 produced by probit analysis (McKelvey and Zavoina, 1975).

TABLE 3

ANALYSIS OF "BEST" MODEL WITH FOUR PRICES

TABLE 4

ANALYSIS OF "BEST" MODEL WITH THREE PRICES

The resulting equations from multiple regression and probit analyses are shown in Tables 3 and 4. The equations in Table 3 are based on data of all four prices while data for $1.98 to $3.98 were used to compute the equations in Table 4. A stepwise solution was used to enter the variables in the multiple regression equations in Tables 3 and 4.

The regression equation in Table 3 produced an R2 of .32 (F = 22.07, d.f. = 5,234, p < .001). The R2 = .47 following probit analysis was also significant with the test statistic distributed as X2 (l* = 89, d.f. = 5, p < .001).

Coefficients of the three main effects of price, expertise, and purchase pal were each statistically significant (p < .01, two-tailed tests) in both the regression and probit equations in Table 3. The coefficients of price-squared was also statistically significant (p < .02). The expertise-by-price interaction coefficient was the only interaction coefficient statistically significant (p < .02 from the regression coefficient).

No other variables were statistically significantly related to purchase (p < .10) based upon analysis of variance of the data.

R2 and R2 in Table 4 are also statistically significant (p < .001). The standardized partial regression coefficients with price instead of price-squared are similar to the coefficients found in Table 4:

Z'y = .582ZE + .261Z(PrP) - .268Z(EPr2)

Z'y = .788ZE + .260Z(PrP) - .460Z(EPr)

The main effects of price and pal were not statistically significant from an analysis of variance of the data using three prices. The price-by-pal (p < .001) and expertise-by-price-squared (EPr, p < .07) interactions were statistically significant.

The interactions effects on purchases of price and expertise are apparent in Figure 5. For four prices and high expertise the significant stepwise multiple regression is Z'y = -.515ZPr2 and for four prices and low expertise Z'y= -1.480ZPr2+ 1.227ZPr a positive price slope occurs from $1.98 to $3.98 for the low expertise condition versus a negative price-squared slope for the high expertise condition.

FIGURE 5

EFFECTS OF PRICE AND SALESMAN EXPERTISE

For the purchase pal conditions and four prices:

Z'y = -.361ZPr2    (No Pal)

Z'y = -.868ZPr2 + 1.525ZPr    (Pal)

The quadratic relationship between price and demand for the four price treatments when purchase pals were present is shown in Figure 6. Both stepwise multiple regression functions for the absence or presence of purchase pals were statistically significant (p < .01).

FIGURE 6

EFFECTS OF PRICE AND PURCHASE PAL

While some interaction terms were found to be significant, H4-7 were not supported by the analysis. For H4, decreases in the likelihood of purchase produced by increases in prices were not greater, the lower the level of perceived expertise of the salesman. In fact, purchases increased with price increases from $1.98 to $3.98 for the low expertise condition.

Decreases in the likelihood of purchase produced by increases in prices were not greater when a purchase pal accompanied the customer compared with decreases when no purchase pal is present. In fact, purchases increased with price increases from $1.98 to $3.98 when a purchase pal was present.

CONCLUSIONS

Customers had little opportunity to have prior knowledge concerning the product used in this study and they may be assumed to have engaged in Extensive Problem Solving decision making (Howard and Sheth, 1969), i.e., well-defined and structured choice criteria have not yet developed for the customers. Consequently, factors other than price, e.g., salesman expertise and advice of purchase pals, may affect purchase behavior to a greater extent, within a range of relevant prices.

Purchase pals may have a positive influence on customer purchasing behavior for some situations. Previous research (Life Insurance Agency Management Association, 1966) has found that when the wife is present during a sales presentation of life insurance, the likelihood of purchase by the husband is substantially increased (47% versus 32% in the study). Percent of purchases increased to 56% when the wife took an active part in the sales discussions. Thus, the presence of purchase pals in Extensive Problem Solving situations might be expected to produce greater likelihoods of positive sales outcomes. However, further research is necessary for different products, consumers, and decision making processes before any firm conclusions can be reached.

Price changes had no substantial effects on demand from $1.98 to $3.98, while demand was substantially decreased when the product was priced at $5.98. This finding indicates that a broad range of acceptable prices may exist for consumers for some products, i.e., products for which consumers do not have prior price information.

More formal attention to significant interaction effects of marketing decision variables and other variables is needed in the study of consumer behavior. The study of such interaction effects will help pinpoint the conditions when one "common-sense" or intuitive belief will occur versus an opposite but equally plausible belief.

LIMITATIONS

The findings presented in this study are limited to the customers participating in the experiment. Other forms and degrees of the expertise treatments could easily have been developed. Unfortunately, the discussions between the purchase pal and the customer, if any, were not recorded in the study. Also, Capon (1975) found a significant interaction effect between salesman and sales treatment which indicates that perception of and success of the salesman cannot be separated from the sales message (and vice versa). Thus, future studies of salesman-buyer interactions should include more than one salesperson and the resulting data should be analyzed for possible interaction effects between salespersons and sales messages.

REFERENCES

Gerald D. Bell, "Self-Confidence and Persuasion in Car Buying," Journal of Marketing Research, 4 (February, 1967), 46-52.

James R. Bettman, "Threshold Model of Attribute Satisfaction Decisions," Journal of Consumer Research, 1 (September, 1974), 30-35.

Timothy C. Brock, "Communicator-Recipient Similarity and Decision Change," Journal of Personality and Social Psychology, 1 (June, 1965), 650-654.

Noel Capon, "Persuasive Effects of Sales Messages Developed from Interaction Process Analysis," Journal of Applied Psychology, 60 (April, 1975), 238-244.

Ronald C. Curhan, "The Effects of Merchandising and Temporary Promotional Activities on the Sales of Fresh Fruits and Vegetables in Supermarkets," Journal of Marketing Research, 11 (August, 1974), 282-294.

Paul E. Green, "On the Analysis of Interactions in Marketing Research Data," Journal of Marketing Research, 10 (November, 1973), 410-420.

John A. Howard and Jagdish N. Sheth, The Theory of Buyer Behavior (New York: John Wiley, 1969).

Paul Kau and Lowell Hill, "A Threshold Model of Purchasing Decisions," Journal of Marketing Research, 9 (August, 1972), 264-270.

Herbert C. Kelman, "Processes of Opinion Change," Public Opinion Quarterly, 25 (Spring/1961), 57-78.

Herbert C. Kelman and Alice H. Eagly. "Attitude Toward Communicator, Perception of Communication Content and Attitude Change," Journal of Personality and Social Psychology, 1 (January, 1965), 63-78.

Philip Kotler, Marketing Management (Englewood Cliffs: Prentice Hall, 1972).

Philip Kotler, Marketing Decision Making (New York: Holt, Rinehart and Winston, 1971).

Philip Kotler, ".Marketing Mix Decisions for New Products," Journal of Marketing Research, 1 (February, 1964), 43-49.

Life Insurance Agency Management Association, The Family Financial Officer (Hartford: Life Insurance Agency Management Association, Report 5, 1966).

John M. McCann, "Market Segment Response to the Marketing Decision Variables," Journal of Marketing Research, 11 (November, 1974), 399-412.

Richard D. McKelvey and William Zavoina, "A Statistical Model for the Analysis of Ordinal Level Dependent Variables," Journal of Mathematical Sociology, 4 (in press, 1975).

Arch G. Woodside and J. William Davenport, Jr., "The Effect of Salesman Similarity and Expertise on Consumer Purchasing Behavior," Journal of Marketing Research, 11 (May, 1974), 198-202.

Arch G. Woodside and J. William Davenport, Jr., "The Effect of Price and Salesman Expertise on Customer Purchasing Behavior," Journal of Business (in press, 1975).

Arch G. Woodside and J. Taylor Sims, "A Retail Experiment in Pricing a New Product," Journal of Retailing, 50 (Fall, 1974), 56-65.

Arch G. Woodside, Robert E. Pitts, and Aaron Gewirtz, "Multiple Regression Analysis of Interaction Effects in Multidimensional Category Data," unpublished working paper, College of Business Administration, University of South Carolina, 1975.

---------------------------------------