Information is essential to consumers when making buying decisions. With the exposures to so many information messages each day, one sometimes wonders if they make a difference. The quality, intensity, and recipient of these messages are obviously importance components in how consumers respond. "Flowers. Alive with Possibilities " has been the newest entry into the commodity information arena with the objective to reach flower buyers and change their purchasing habits. Details about the messages and quality of the information can be found on the web page of the Flower Promotion Organization (www.flowerpossibilities.com), the group responsible for the design and implementation of this new flower promotion program. Along with the creative aspects of the flower promotion theme, this report addresses the question if the messages have had an impact on the demand for flowers. Statistical methods are used to measure if consumer flower buying habits have changed as a result of the new generic promotion program.

FPO's programs were started in the fall of 2000 with the introduction of "Flowers. Alive with Possibilities" in five test markets. Existing female flower buyers were the primary target using television and radio along with some forms of print media. All efforts were in the non-special occasion periods with the focus on females buying flowers for self use with the key being existing buyers and purchases for self use. While there is nothing conceptually to o prevent expansion of the programs beyond the current focus, budgets and goals of the test program lead to the selection of these program targets.

How can buying patterns change? While one cannot measure all buying behavior, generally new innovative promotions could attract new buyers to the existing markets and/or change the frequency of transactions among existing buyers. Likewise, responses could be different among the targeted and non-targeted markets as well as differences according to the purpose of buying (i.e., self versus gifts). In order to analyze consumer responses to these new FPO programs, one must have data detailing purchases of flowers and have controls where the promotions did not take place.

Figure I is useful for illustrating how the impacts of these new generic promotions were measured. First, five target cities were selected using the criteria of geographic diversity and feasibility for implementing the promotions within the defined periods and budget. Second, as seen in this figure, 12 control cities were selected that were geographically diverse but far enough removed from the target cities to assure the minimum level of spillover of messages from the target cities. That is, the control cities were not exposed to the media messages. Using the American Floral Endowment and Ipos-NPD consumer tracking data, detailed monthly flower purchasing activities were recorded for both the target and control cities. These data included both the number of buyers out of the household population in the market and the frequency of transactions among existing buyers. Buyers were identified by gender, purpose, age, income, and city. Statistical models were then used to determine if the promotions impacted the entry of new buyers and changed the frequency of transactions among existing flower buyers. FPO's activities were divided into three phases or periods of promotions: Phase I - September/October 2000; Phase 11 - March/April 200 1; and Phase III - September/October 2001. Overall expenditure intensity is illustrated in Figure 2. Approximately $5.4 million was spent during the three phases with 66 percent of the monies going to local television advertising. The actual media mixed differed within the target cities.

Performance Measures

In order to measure any respond to the FPO promotions there must be very specific criteria for measuring buying activity. Hence, two performance measures were defined:

Within an average month, typically 5 to 7 percent of the households buy some flowers with the percent depending on the demographics of the household. Similarly, atypical buying household may have around 1. 5 transactions per month again depending on the demographics of the household. From a performance perspective the main issue is to determine if the existence of the FPO promotions changed the level of market penetration and the frequency of transactions. Clearly, there should be no changes in the control cities since FPO promotions were not present. If changes were seen in the control cities associated with the precise periods that FPO activities were in place, then something other than the new promotions could have caused any measured shift in either performance value. This is why it is so essential to have the control cities in the analysis to make sure that any conclusions about the impact of the "Flowers. Alive with Possibilities" theme is not picking up the effects of something else.

Market Penetration - Statistical Model

While buyer penetration obviously changes over time and across demographics, the FPO commercials did not target new buyes. Statistically, there was no evidence that the promotion theme attracted new buyers within the target markets. Table I includes the statistical coefficients and t-values for measuring the impact of the FPO promotions in both the target and control cities. The t-values are important in that Table 1. Market penetration.

the values are considerably less than 2.0, the general criteria level for being statistically significant. In every case. there is no statistical evidence of changes in market penetration that could be attributed to the generic promotions. In the target cities the effects are positive as seen with the signs of the coefficients (middle column in Table I) but statistically are not different from zero. These promotions did not target new buyers and there is no statistical evidence of attracting new buyers within the targeted cities.

Market Frequency - Statistical Model

FPO promotions targeted existing buyers will) the message to buy more flowers for self use. If the promotions were successful in changing the frequency of buying, any relationship between the promotion dollars (Figure 2) and the frequency of transactions should be positive and statistically significant. Similarly, there should be no response in the control cities. Table 2 reports these coefficients and the corresponding t-values, again with the critical t-value for significance being around 2.0.

The evidence is overwhelming in showing the positive gains in frequency attributed to the FPO promotions as illustrated in Table 2. The t-values for each phase in the targeted cities are considerably greater than 2.0 and the coefficient signs are positive as they should be if the promotions increased the frequency of buying. Likewise the coefficients for the control cities are statistically no different from zero. Clearly, the gains attributed to the FPO promotions are not reflecting some external event that took place in the total flower market during the promotion periods. Otherwise, one would have seen similar gains in the control cities. Comparing these t-values between the target and controls add strong statistical confidence to the overall conclusion about the positive gains attributed to the new generic promotions of flowers.

Since Tables I and 2 show the gains in frequency but not in attracting new buyers, the remaining discussions concentrate on the frequency without further consideration for new buyers.

Market Frequency - by Demographics

Using the statistical models one can next calculate the transactions per buyer with and without the present of the FPO promotions. The difference in frequency then would be the gains attributed to the promotions. Since specific demographic groups were targeted, it is useful to show these gains by selected demographic groups. Figures 3a shows the transaction frequency A with and without FPO for all demographic groups combined. Figure 3b shows the same frequencies among the targeted group of females for h purchases. Corresponding numerical values are printed at the bottom of each figure to give the exact frequency values.

Across all demographic prior to the FPO startup, the frequency among existing buyers was between 1.5 to 1.6 transactions per month, depending on the month considered. In Phase I (Sept/Oct. 2000) the frequencies ranged from 1.66 to 1.84. Without FPO the estimates are that the frequencies would have been 1.44 to 1.50 transactions per buyer. Specifically, in October the gain was .40 transactions representing a 28% increase in the frequency of transactions. The amount of gain will differ depending on the level of promotions for each month as presented in Figure 2. One can simply move across the bars or values in these figures and see the gains for Phases 11 and 111. A simple average over the three phases points to around a 19% increase in the frequency of transactions that can be attributed to the new flower promotion programs in the 5-city target area.

While the amount of gain is important, of equal interest is that fact that households have been targeted by demographic profiles and the promotions worked to simulate that group. It is clearly feasible to target existing flower users and they will respond as seen with Figures 3a and 3b.

Promotion Response by Income

Earlier evaluations of the now terminated PromoFlor showed that buying behavior differed by income groups as one would expect. For that campaign, there was basically no response to Mr. Buzz among the highest income groups (above $75,000). FPO's campaign targeted a different audience, yet there were expectations of differences across incomes. One approach used was to estimate the frequency responses among households with less than $50,000 per year compared with those above that level. Once the responses are known then it is possible to simulate the gains attributed to FPO over a range of potential expenditures.

Figure 4 shows the gain in the frequency of transactions for the two income categories. Without the promotions the frequency stands at 1.37 and 1.54 for the low and high income groups. respectively or a 12% difference in frequency of transactions. Next let FPO promotions increase to $2.5 million for the October 2000 period example. Frequency gains among the under $50,000 group increased from 1.37 to 2.08 for a 0.71 gain while the higher income group showed a 0.22 frequency gain. The difference in response is easily seen in Figure 4 over the range of FPO simulated expenditures with the larger gains among the under $50,000 group. In some ways this response is similar to what was seen with PromoFlor where the lower incomes tended to respond more than the higher income groups.

Care must be taken when comparing these gains in frequency alone since the distribution of buyers within the two income groups are quite different. Approximately 57% of the total number of flower buyers are in the above $50,000 income group. Hence, while the frequency gain is less for the higher income group, the actual number of buyers is much larger. For the example illustrated in Figure 4 for October 2000. one needs to multiply the frequency times the number of buyers in each income group. Since market penetration was shown not to be influenced by the promotions, the gains in terms of total transactions is simply the number of buyers x the frequencies. For example in October 2000 there were 77,000 household buyers under $50.000 in income and 93,000 flower buyers in the target markets with incomes $50,000 or more. Without the FPO promotions the total transactions equal 105,472 and 143,298 between the low and high income groups. With the promotions set to $2.5 million the transactions increased to 160,154 and 163,726. While the gains in transactions were greater among the low income group, the absolute number of transactions achieved was greater targeting the higher income group. Interestingly, the total transactions between the two income groups differed by almost 38,000 without the promotions hut only 1.572 with the promotions. Ultimately one must also calculate the amount of monies spent per transactions between the two income groups.

Figure 4 also shows the declining incremental gains associated with addition promotion expenditures. This type of response is typical of most promotion programs where incremental gains drop off as additional monies are spent. This basically says that the initial large gains cannot be expected to continue by simply adding more and more money to the same market. The increment or marginal gains will continue to decline as the total promotions are increased. At some point one would be better off by spreading the promotion dollars across more cities that just putting greater amounts within the same cities.

Differences Across the Phases

The models were also estimated attempting to capture differences among the three promotion periods. Total frequency gains initially shown in Figure 2 can differ because of the levels of promotions over the three phases and if the responses to the promotions changed as new phases were added to the analysis.

Figure 5 shows the simulated gains using the response coefficients as new phases were added to the analysis. The lower response curve is for Phase I and the calculated frequency with the actual promotion level is predicted to be 1.84 transactions (point 1). Total expenditures were decreased in Phase 11 but the response to the promotions actually increased as seen with the upper curve. At point (11) the transactions were calculated to be 2.06 for the lower expenditure level. Finally, during Phase III the response dropped off but still remained slightly above the initial phase. The total dollars spent for Phase III in October were less while the response was slightly higher than the initial efforts. The frequency for Phase III then equaled 1.76 for October. A declining marginal gain to the generic promotion efforts is again seen as more monies are spent. This figure points to potential adjustments as programs mature and highlights the importance of how the monies are allocated across markets and time.

Promotion Gains by Outlets

An issue of major interest is whether the gains seen with FPO's efforts are or are not distributed equitably across outlets for flowers. The American Floral Endowment data show that 75% of all fresh cut flowers transactions are through florists and supermarkets. Supermarkets have 48% and florists account for 26.5% of the total transactions. In terms of expenditures, florist account for half of the total flower purchases.

Models similar to those presented earlier were estimated for each of the two major outlet types and the results are illustrated using Figure 6 for all FPO phases combined. Averaged over the three phases, the transaction frequency through florists was 1.37 transactions per month and 1.5 1 transactions through supermarkets assuming no FPO promotions. Using the estimated models but now specific to each outlet, the models show transactions of 1.62 and 1.80 with the promotions. Both outlets shared in the gains in transactions and on a relative basis the percentage gains in transactions were very similar as shown with the two bars in Figure 6. Florists experienced a 25% gain and supermarkets, a 29% gain overall. Figure 6 points to a reasonable level of equity in sharing the retail market transaction gains between the two major dominant outlets for flowers. Supermarkets gain slightly more in transactions but the average transaction value is less than for florists.

Conclusions

FPO has now finished three phases with its generic promotion of flowers. Selected consumers in targeted markets have been exposed to a new theme and the empirical evidence points to positive gains. With these programs, existing buyers were targeted for self purchase and that is where the gains were realized. New buyers were not attracted to the market in any significant way. Likewise, control markets were used to confirm that the measured FPO gains were real and not just some random event influencing all the flower markets.

The analysis shows the importance of having a continuous data set such as supported by the American Floral Endowment consumer tracking study.

Ronald W. Ward is a professor in the Food and Resource Economics Department, University of Florida. Gainesville, Florida. RWWard@mail.ifas.ufl.edu.