New Product Research - Test Marketing
New Product Research - Test Marketing
In terms of providing a realistic evaluation of a marketing program, there is no substitute for conducting an experiment in which the marketing program (or perhaps several versions of it) are implemented in a limited but carefully selected part of the market. The impact of the total marketing program, with all its interdependencies, is determined then in the market context as opposed to the artificial context associated with the concept and product tests that have been discussed.
There are two primary functions of test marketing. The first is to gain information and experience with the marketing program before making a total commitment to it. The second is to predict the program's outcome when it is applied to the total market.
The test market fulfills the first function by being a pilot operation for the marketing program, just as a pilot production facility might serve to debug a manufacturing process. There are all sorts of possible and unanticipated problems associated with a marketing program. The physical problems of transportation, handling, stocking, and shelf life, for example, can generate difficulties. One prominent manufacturer tested a small compact box of facial tissues and found that, in the South, the tissues absorbed dampness and caused some boxes to explode after being on the shelf for a month. Clearly, without such a test the manufacturer might have gone national immediately, with disastrous consequences. It was only by actually
having the product on the shelf for a month that the problem was uncovered. When the goal of the test market is to try out the marketing program and when prediction is not required, it is not so necessary to develop elaborate experimental designs, although there should be some concern that the test is general enough to expose problems. Thus, if the tissue manufacturer had not bothered to test the program in the South, the problem with the dampness would not have been uncovered.
The second function of a test market is to predict the outcome of the market program when it is applied to the total market. Although the test market does provide a realistic test of the impact of a marketing program, it also has a variety of methodological problems associated with it that make prediction difficult, as we shall see. As a result, the ability of test markets to predict is much less than one would expect.
There are really two types of test markets, the sell-in test market and the controlled-distribution scanner markets. The sell-in test markets are cities in which the product is sold in just as it would be in a national launch. In particular, the product has to gain distribution space. The controlled-distribution scanner markets are cities for which distribution is prearranged and the purchases of a panel of customers are monitored using scanner data. First we will discuss the design of sell-in market tests.
Designing the Sell-In Market Test
The market test design involves several elements. First, the test cities need to be selected. Second, the market programs need to be implemented and controlled in each city. Third, the length of the test market must be established. Finally, a set of measures must be devised to evaluate the program.
Selecting the Test Cities City characteristics that are usually relevant in the selection of cities to use in the test market include:
1. Representativeness. Ideally the city should be fairly representative of the country in terms of characteristics that will affect the test outcome, such as product usage, attitudes, and demographics. Of course, the results can be adjusted to compensate for differences that are well known, if their impact upon sales can be estimated. For example, southerners use more biscuits, teenagers drink more soda, older people have more use for pharmaceutical products, and some towns have harder water (which could affect a bath soap test).
2. Data availability. It often is helpful to use store audit information to evaluate the test, as it provides safes data adjusted for inventory changes and gives other useful information such as shelf facings and in-store promotions. If so, it would be important to use cities containing retailers who will cooperate with store audits.
3. Media isolation and costs. It is desirable to avoid media spill over. Media that "spill out" into nearby cities is wasteful and increases costs. Conversely, "spill in" media from nearby cities can contaminate a test. Media cost is another consideration. Some media begin to charge exorbitant rates when they know they are in popular test markets.
4. Product flow. It may be desirable to use cities that don't have much "product spillage" outside the area.
Another issue is the number of test cities to use. Although most test markets use two or three test cities (or pairs of cities if two marketing programs are being tested), too many tests are conducted with a single test city. A leading marketing researcher, Valentine Appel, has observed that a single test city can lead to unreliable results because of the variation across cities of both brand sales and consumer response to marketing programs.27
Appel illustrates his point with an example of an advertiser that tested whether doubling the $26 million ad budget would arrest the brand's market share decline. Ten matched pairs of markets were generated in each of ten sales districts and one market in each pair was assigned to the heavy advertising treatment. The average share improvement was .81, under that required to justify the added budget. However, the share change ranged from - .22 to 2.43 and for three city pairs the change exceeded 1.5 percent. If one of these three cities had been the basis of the test, a very different conclusion would have resulted.
How many test cities should be used? One rule of thumb is to use around three. A more precise recommendation using the methods set forth in Chapter 13 will require an estimate of the variation in the test results across cities. One indicator of that variation is the month-to-month changes in sales for similar brands. Also required, of course, is a judgment about how accurate the test market prediction is.
Implementing and Controlling the Test A second consideration is to control the test by ensuring that the marketing program is implemented in the test area so as to reflect the national program. This step is not as easy as it may sound. A national program may not be defined precisely enough or it may not decompose easily to the local level. A national advertising budget, for example, may not be easy to allocate to a local level or it may not be well defined in terms of exactly how the television coverage is to be scheduled.
The test itself may tend to encourage those involved to enhance the effectiveness of the marketing program. Salespeople maybe more aggressive in obtaining distribution. Retailers may be more cooperative than usual because they are told it is an "interesting and important test market." Those implementing a test public policy program may be more motivated and involved than could be expected when the program goes "national." Considerable discipline must be applied to ensure that the marketing program is applied as intended.
There is also the reaction of competitors. At one extreme they can destroy the test by deliberately flooding the test areas with free samples or in-store promotions. More likely, however, they will experiment with retaliatory actions and monitor the results themselves. The question is: what impact will their actions have on the results? Even if they do nothing, there is a concern with how the results will change when the program goes national and they do react. Each test market should be monitored carefully so that competitive reaction and other relevant market parameters are tracked over time and their impact on the market response detected.
Timing A third consideration is timing. If possible, a test market normally should be in existence for one year. Even after a year has passed, if the program goes national, the test market should continue to be monitored to detect the impact of changes in the environment. An extended time period is needed for several reasons. First, there are often important seasonal factors that can be observed only if the test is continued for the whole year. Second, initial interest is often a poor predictor of a program's staying power. There is usually a fatigue factor that sometimes can take a long time to materialize. Cereal with dried fruit is one product that did very well during a short test but fizzled later because customers tired of it. Third, it is useful to allow the competition and other market factors to adjust to see what impact they will have on the results. An analysis of 141 test markets by the A. C. Nielsen Company indicated that, after the first six months of testing, the chance of adequately predicting market share at the end of 12 to 18 months is only one in two, but this rises to about two in three after eight months.28
Measurement A crucial element of the test market is the measure used to evaluate it. A basic measure is sales based on shipments or warehouse withdrawals. One problem with such information is that inventory fluctuations can distort this measurement, and it is not a very sensitive measure of consumer response.
Store audit data provide actual sales figures and are not sensitive to inventory fluctuations. They also provide variables such as distribution, shelf facings, and in-store promotional activity. Knowledge of such variables can be important in evaluating the marketing program and in interpreting the sales data.
Measures such as brand awareness, attitude, trial purchase, and repeat purchase are obtained directly from the consumer, either from surveys or consumer panels. Such variables as brand awareness and attitude also serve as criteria for evaluating the marketing program and can help interpret sales data. The most useful information obtained from consumers, however, is whether they bought the product at least once, whether they were satisfied with it, and have either repurchased it or plan to.
There is also the potential to project the levels of awareness, trial rates, and repeat purchase rates that will be achieved by a new product in a test market using survey data from the first three or so months of the test. A series of surveys would be conducted starting just prior to the test market launch. BBDO has developed a set of models termed the NEWS/Market system to do just that.29
In the NEWS/Market system, the surveys monitor new product awareness (aided and unaided recall), receipt and use of samples or coupons, trial purchases, and repeat purchases. Advertising and promotion expenditures are also monitored. Some relationships such as the drop in awareness among those not reexposed to the advertising, are estimated from prior research on other products. Of 34 NEWS/Market projections for which an ultimate test market share or sales were available, the average predictive error was 17.5 percent of the actual market share achieved. In 70 percent of the cases the prediction was within one share point, and in 90 percent of the time it was within 1.5 share points.
Predicting share from panel data will be discussed under controlled distribution scanner test markets because panel data is readily available in that context. The early prediction of test market results is important because it provides the potential to reduce the cost and time associated with test markets.
Costs of a Test Market
In making cost benefit judgments about test markets, all costs need to be considered. Many costs are relatively easy to quantify such as the development and implementation of the marketing program, preparation of test products, administration of the test, and collection of data associated with the test.
The costs and risks associated with delaying the launch of a new product are more difficult to quantify. If a new product launch is delayed by six months or a year, an opportunity to gain a substantial market position might be lost. Even worse, another competitor may preempt the market while the test is being pursued. For example, while Hills Brothers was testing High Yield Coffee, Folgers preempted with their Folger's Flake and when Proctor and Gamble was testing Bounce, Calgon introduced Cling Free.
Primarily because of the risks of being preempted, many brands, such as Quaker's Chewy Granola Bars, Sara Lee's Frozen Croissants, and Pills-bury 's Milk Break Bars have elected to bypass test markets.
Controlled Distribution Scanner Markets
The basic characteristics of controlled distribution scanner markets (CDSM) such as IRI's BehaviorScan were described in Chapter 5. Basically they tap into the retail distribution system, collecting data on all grocery store purchases and in a growing amount of drug and other retail store types where CDSM exists. They are termed controlled distribution because there are generally agreements with retailers to allow new products under test to have access to shelf space.
IRI's BehaviorScan, for example, has a panel of 3000 households established in ten communities such as Pittsfield, Massachusetts; Rome, Georgia; and Salem, Oregon. Each panel member carries an ID card that he or she presents to supermarkets (and some drugstores) when buying something allowing IRI to monitor all purchases. Knowledge of exactly when a panelist bought and repurchased a new product is thus available. In addition, the in-store conditions such as price, promotions, and special displays are also controlled and monitored. Further, the panelist has a device connected to his or her TV set that not only allows the channel selection to be monitored, but also allows the advertiser to substitute one advertisement for another in what are called "cut-ins."
CDSMs have four major advantages over test markets. First, they are less expensive. Although it is difficult to generalize, they are probably from one-sixth to one-third of the cost of a full test market. Second, there is the potential to do more experimenting with marketing variables in a CDSM.
The advertising seen by panel members is controllable. Further, the in-store activities such as promotions and pricing is under more control than it would be in a sell-in test market. Third, the scanner based data are probably more accurate, timely, and complete than that generally available in a sell-in test market. Fourth, there is the potential to provide accurate early estimates of the test market results using the consumer panel information.
The most obvious disadvantages of a CDSM is that it provides no test of the product's ability to gain shelf space, special displays, in-store promotions, and so on. Since gaining distribution can be a crucial issue for some products, leaving it unaddressed can be troublesome. Of course, many distribution problems may not surface in a few sell-in test markets.
Another major CDSM disadvantage is the limited choice of test cities. Several questions arise. Do the available test cities have adequate project-ability for a particular product? Typically one or two cities become so exposed to new products that their reaction begins to differ from those in cities not associated with a CDSM system?
Projecting Trial, Repeat and Usage Rate Using Panel Data The most accurate way to project trial and repeat purchase in a test market is by using panel data where the purchase and repurchase decisions of individual consumers can be monitored.
To estimate the ultimate trial level the percentage of product class buyers who will try the new brand at least once is monitored over time. Figure 21-2 provides a graph of the trial rate for a new toilet soap (Brand T) by four-week periods starting from the launch date. The initial time period is then projected to estimate the ultimate trial level.
Each person who tries the new product is then monitored and the time between the first (trial) purchase and second purchase is noted. The percentage of new product triers who rebuy the product (the repeat rate), is plotted against the time between the first and second purchase and projected. Figure 21-3 illustrates.
The market share estimate is thus the product of the two projections. In the example illustrated by the figures it would be .34 times .25 (which equals .085) or a market share projection of 8.5 percent.
There are a number of refinements that can be made to this basic approach:
1. Sales will depend on usage levels as well as market shares. If the initial buyers on average are heavy product users the estimate could be adjusted.
2. For some product classes (snack foods or cereals, for example) new brands may gain one or even two repeat purchases only to have customers lose interest in them. In those cases, it is important to not only estimate first repeat, but to also estimate second repeat (the percentage of those buying the new brand twice that buy it a third time) and even third and fourth repeat.
3. Accuracy can often be enhanced, if repeat is estimated separately for those who try the brand early in the test from those that try it later. Early triers are generally more enthusiastic about a new brand and tend to have higher repeat rates.
4. If there is no well-defined product class, trial rates are defined as the total number of triers instead of the percent of product class users who try the new brand. An analysis similar to that illustrated in Figure 21-2 is still used.
Market share and sales projections made using this modeling logic can be very accurate as the experience of Parfitt and Collins illustrate.30 They predicted the market shares of 24 successful, new, frequently purchased consumer products. In all 24 cases, their predicted market share, made 24 weeks after launch, was within the highest and lowest actual monthly market share encountered between months 12 and 18 after launch. Projections of trial and repeat have obvious diagnostic value. Weakness in achieving trial is usually due to a problem with the introduction effort, whereas low repeat usually signals a basic product problem.
In terms of providing a realistic evaluation of a marketing program, there is no substitute for conducting an experiment in which the marketing program (or perhaps several versions of it) are implemented in a limited but carefully selected part of the market. The impact of the total marketing program, with all its interdependencies, is determined then in the market context as opposed to the artificial context associated with the concept and product tests that have been discussed.
There are two primary functions of test marketing. The first is to gain information and experience with the marketing program before making a total commitment to it. The second is to predict the program's outcome when it is applied to the total market.
The test market fulfills the first function by being a pilot operation for the marketing program, just as a pilot production facility might serve to debug a manufacturing process. There are all sorts of possible and unanticipated problems associated with a marketing program. The physical problems of transportation, handling, stocking, and shelf life, for example, can generate difficulties. One prominent manufacturer tested a small compact box of facial tissues and found that, in the South, the tissues absorbed dampness and caused some boxes to explode after being on the shelf for a month. Clearly, without such a test the manufacturer might have gone national immediately, with disastrous consequences. It was only by actually
having the product on the shelf for a month that the problem was uncovered. When the goal of the test market is to try out the marketing program and when prediction is not required, it is not so necessary to develop elaborate experimental designs, although there should be some concern that the test is general enough to expose problems. Thus, if the tissue manufacturer had not bothered to test the program in the South, the problem with the dampness would not have been uncovered.
The second function of a test market is to predict the outcome of the market program when it is applied to the total market. Although the test market does provide a realistic test of the impact of a marketing program, it also has a variety of methodological problems associated with it that make prediction difficult, as we shall see. As a result, the ability of test markets to predict is much less than one would expect.
There are really two types of test markets, the sell-in test market and the controlled-distribution scanner markets. The sell-in test markets are cities in which the product is sold in just as it would be in a national launch. In particular, the product has to gain distribution space. The controlled-distribution scanner markets are cities for which distribution is prearranged and the purchases of a panel of customers are monitored using scanner data. First we will discuss the design of sell-in market tests.
Designing the Sell-In Market Test
The market test design involves several elements. First, the test cities need to be selected. Second, the market programs need to be implemented and controlled in each city. Third, the length of the test market must be established. Finally, a set of measures must be devised to evaluate the program.
Selecting the Test Cities City characteristics that are usually relevant in the selection of cities to use in the test market include:
1. Representativeness. Ideally the city should be fairly representative of the country in terms of characteristics that will affect the test outcome, such as product usage, attitudes, and demographics. Of course, the results can be adjusted to compensate for differences that are well known, if their impact upon sales can be estimated. For example, southerners use more biscuits, teenagers drink more soda, older people have more use for pharmaceutical products, and some towns have harder water (which could affect a bath soap test).
2. Data availability. It often is helpful to use store audit information to evaluate the test, as it provides safes data adjusted for inventory changes and gives other useful information such as shelf facings and in-store promotions. If so, it would be important to use cities containing retailers who will cooperate with store audits.
3. Media isolation and costs. It is desirable to avoid media spill over. Media that "spill out" into nearby cities is wasteful and increases costs. Conversely, "spill in" media from nearby cities can contaminate a test. Media cost is another consideration. Some media begin to charge exorbitant rates when they know they are in popular test markets.
4. Product flow. It may be desirable to use cities that don't have much "product spillage" outside the area.
Another issue is the number of test cities to use. Although most test markets use two or three test cities (or pairs of cities if two marketing programs are being tested), too many tests are conducted with a single test city. A leading marketing researcher, Valentine Appel, has observed that a single test city can lead to unreliable results because of the variation across cities of both brand sales and consumer response to marketing programs.27
Appel illustrates his point with an example of an advertiser that tested whether doubling the $26 million ad budget would arrest the brand's market share decline. Ten matched pairs of markets were generated in each of ten sales districts and one market in each pair was assigned to the heavy advertising treatment. The average share improvement was .81, under that required to justify the added budget. However, the share change ranged from - .22 to 2.43 and for three city pairs the change exceeded 1.5 percent. If one of these three cities had been the basis of the test, a very different conclusion would have resulted.
How many test cities should be used? One rule of thumb is to use around three. A more precise recommendation using the methods set forth in Chapter 13 will require an estimate of the variation in the test results across cities. One indicator of that variation is the month-to-month changes in sales for similar brands. Also required, of course, is a judgment about how accurate the test market prediction is.
Implementing and Controlling the Test A second consideration is to control the test by ensuring that the marketing program is implemented in the test area so as to reflect the national program. This step is not as easy as it may sound. A national program may not be defined precisely enough or it may not decompose easily to the local level. A national advertising budget, for example, may not be easy to allocate to a local level or it may not be well defined in terms of exactly how the television coverage is to be scheduled.
The test itself may tend to encourage those involved to enhance the effectiveness of the marketing program. Salespeople maybe more aggressive in obtaining distribution. Retailers may be more cooperative than usual because they are told it is an "interesting and important test market." Those implementing a test public policy program may be more motivated and involved than could be expected when the program goes "national." Considerable discipline must be applied to ensure that the marketing program is applied as intended.
There is also the reaction of competitors. At one extreme they can destroy the test by deliberately flooding the test areas with free samples or in-store promotions. More likely, however, they will experiment with retaliatory actions and monitor the results themselves. The question is: what impact will their actions have on the results? Even if they do nothing, there is a concern with how the results will change when the program goes national and they do react. Each test market should be monitored carefully so that competitive reaction and other relevant market parameters are tracked over time and their impact on the market response detected.
Timing A third consideration is timing. If possible, a test market normally should be in existence for one year. Even after a year has passed, if the program goes national, the test market should continue to be monitored to detect the impact of changes in the environment. An extended time period is needed for several reasons. First, there are often important seasonal factors that can be observed only if the test is continued for the whole year. Second, initial interest is often a poor predictor of a program's staying power. There is usually a fatigue factor that sometimes can take a long time to materialize. Cereal with dried fruit is one product that did very well during a short test but fizzled later because customers tired of it. Third, it is useful to allow the competition and other market factors to adjust to see what impact they will have on the results. An analysis of 141 test markets by the A. C. Nielsen Company indicated that, after the first six months of testing, the chance of adequately predicting market share at the end of 12 to 18 months is only one in two, but this rises to about two in three after eight months.28
Measurement A crucial element of the test market is the measure used to evaluate it. A basic measure is sales based on shipments or warehouse withdrawals. One problem with such information is that inventory fluctuations can distort this measurement, and it is not a very sensitive measure of consumer response.
Store audit data provide actual sales figures and are not sensitive to inventory fluctuations. They also provide variables such as distribution, shelf facings, and in-store promotional activity. Knowledge of such variables can be important in evaluating the marketing program and in interpreting the sales data.
Measures such as brand awareness, attitude, trial purchase, and repeat purchase are obtained directly from the consumer, either from surveys or consumer panels. Such variables as brand awareness and attitude also serve as criteria for evaluating the marketing program and can help interpret sales data. The most useful information obtained from consumers, however, is whether they bought the product at least once, whether they were satisfied with it, and have either repurchased it or plan to.
There is also the potential to project the levels of awareness, trial rates, and repeat purchase rates that will be achieved by a new product in a test market using survey data from the first three or so months of the test. A series of surveys would be conducted starting just prior to the test market launch. BBDO has developed a set of models termed the NEWS/Market system to do just that.29
In the NEWS/Market system, the surveys monitor new product awareness (aided and unaided recall), receipt and use of samples or coupons, trial purchases, and repeat purchases. Advertising and promotion expenditures are also monitored. Some relationships such as the drop in awareness among those not reexposed to the advertising, are estimated from prior research on other products. Of 34 NEWS/Market projections for which an ultimate test market share or sales were available, the average predictive error was 17.5 percent of the actual market share achieved. In 70 percent of the cases the prediction was within one share point, and in 90 percent of the time it was within 1.5 share points.
Predicting share from panel data will be discussed under controlled distribution scanner test markets because panel data is readily available in that context. The early prediction of test market results is important because it provides the potential to reduce the cost and time associated with test markets.
Costs of a Test Market
In making cost benefit judgments about test markets, all costs need to be considered. Many costs are relatively easy to quantify such as the development and implementation of the marketing program, preparation of test products, administration of the test, and collection of data associated with the test.
The costs and risks associated with delaying the launch of a new product are more difficult to quantify. If a new product launch is delayed by six months or a year, an opportunity to gain a substantial market position might be lost. Even worse, another competitor may preempt the market while the test is being pursued. For example, while Hills Brothers was testing High Yield Coffee, Folgers preempted with their Folger's Flake and when Proctor and Gamble was testing Bounce, Calgon introduced Cling Free.
Primarily because of the risks of being preempted, many brands, such as Quaker's Chewy Granola Bars, Sara Lee's Frozen Croissants, and Pills-bury 's Milk Break Bars have elected to bypass test markets.
Controlled Distribution Scanner Markets
The basic characteristics of controlled distribution scanner markets (CDSM) such as IRI's BehaviorScan were described in Chapter 5. Basically they tap into the retail distribution system, collecting data on all grocery store purchases and in a growing amount of drug and other retail store types where CDSM exists. They are termed controlled distribution because there are generally agreements with retailers to allow new products under test to have access to shelf space.
IRI's BehaviorScan, for example, has a panel of 3000 households established in ten communities such as Pittsfield, Massachusetts; Rome, Georgia; and Salem, Oregon. Each panel member carries an ID card that he or she presents to supermarkets (and some drugstores) when buying something allowing IRI to monitor all purchases. Knowledge of exactly when a panelist bought and repurchased a new product is thus available. In addition, the in-store conditions such as price, promotions, and special displays are also controlled and monitored. Further, the panelist has a device connected to his or her TV set that not only allows the channel selection to be monitored, but also allows the advertiser to substitute one advertisement for another in what are called "cut-ins."
CDSMs have four major advantages over test markets. First, they are less expensive. Although it is difficult to generalize, they are probably from one-sixth to one-third of the cost of a full test market. Second, there is the potential to do more experimenting with marketing variables in a CDSM.
The advertising seen by panel members is controllable. Further, the in-store activities such as promotions and pricing is under more control than it would be in a sell-in test market. Third, the scanner based data are probably more accurate, timely, and complete than that generally available in a sell-in test market. Fourth, there is the potential to provide accurate early estimates of the test market results using the consumer panel information.
The most obvious disadvantages of a CDSM is that it provides no test of the product's ability to gain shelf space, special displays, in-store promotions, and so on. Since gaining distribution can be a crucial issue for some products, leaving it unaddressed can be troublesome. Of course, many distribution problems may not surface in a few sell-in test markets.
Another major CDSM disadvantage is the limited choice of test cities. Several questions arise. Do the available test cities have adequate project-ability for a particular product? Typically one or two cities become so exposed to new products that their reaction begins to differ from those in cities not associated with a CDSM system?
Projecting Trial, Repeat and Usage Rate Using Panel Data The most accurate way to project trial and repeat purchase in a test market is by using panel data where the purchase and repurchase decisions of individual consumers can be monitored.
To estimate the ultimate trial level the percentage of product class buyers who will try the new brand at least once is monitored over time. Figure 21-2 provides a graph of the trial rate for a new toilet soap (Brand T) by four-week periods starting from the launch date. The initial time period is then projected to estimate the ultimate trial level.
Each person who tries the new product is then monitored and the time between the first (trial) purchase and second purchase is noted. The percentage of new product triers who rebuy the product (the repeat rate), is plotted against the time between the first and second purchase and projected. Figure 21-3 illustrates.
The market share estimate is thus the product of the two projections. In the example illustrated by the figures it would be .34 times .25 (which equals .085) or a market share projection of 8.5 percent.
There are a number of refinements that can be made to this basic approach:
1. Sales will depend on usage levels as well as market shares. If the initial buyers on average are heavy product users the estimate could be adjusted.
2. For some product classes (snack foods or cereals, for example) new brands may gain one or even two repeat purchases only to have customers lose interest in them. In those cases, it is important to not only estimate first repeat, but to also estimate second repeat (the percentage of those buying the new brand twice that buy it a third time) and even third and fourth repeat.
3. Accuracy can often be enhanced, if repeat is estimated separately for those who try the brand early in the test from those that try it later. Early triers are generally more enthusiastic about a new brand and tend to have higher repeat rates.
4. If there is no well-defined product class, trial rates are defined as the total number of triers instead of the percent of product class users who try the new brand. An analysis similar to that illustrated in Figure 21-2 is still used.
Market share and sales projections made using this modeling logic can be very accurate as the experience of Parfitt and Collins illustrate.30 They predicted the market shares of 24 successful, new, frequently purchased consumer products. In all 24 cases, their predicted market share, made 24 weeks after launch, was within the highest and lowest actual monthly market share encountered between months 12 and 18 after launch. Projections of trial and repeat have obvious diagnostic value. Weakness in achieving trial is usually due to a problem with the introduction effort, whereas low repeat usually signals a basic product problem.
Comments
Post a Comment