Click on an image to enlarge the picture.

Starting in the September of 2000, the Flower Promotion Organization (FPO) implemented the first phase of a new program to promote fresh-cut flowers. This program targeted five U.S. cities with the goal of increasing the frequency of buying fresh-cut flowers among existing female flower buyers in non-traditional holiday and event periods. The promotion periods were selected to correspond with these periods. These promotions have now been used in six separate phases with Phase VI ending in June 2003. As shown with Figure 1, the initial investment accounted for nearly 48 percent of the total $6.84 million spent on various types of promotions. The promotion dollars spent in subsequent periods generally tended to decline. Also, starting with Phase IV, Minneapolis/St. Paul was added as a sixth target city, accounting for around 16 percent of the dollars in those periods. As part of the overall generic promotion effort, a plan for evaluating buyer responses to the programs was put in place. Potential changes in market penetrations (i.e., buyers/households) and buying frequency (i.e., transactions/per buyer) were adopted as measurement criteria for judging consumer responses to the promotions. Furthermore, to provide a control group to compare with the target cities, 13 separate U.S. cities sufficiently removed from the target cities were used to provide a check against any responses seen in the target cities. Figure 2 outlines both the target and control cities.

To statistically measure the impact of FPO's programs, market penetration and buyer frequency models were estimated using household purchasing data from the target and control cities. With these models, one can determine statistically if the promotions have stimulated demand for fresh-cut flowers and then determine the value of any gains attributed to the promotions. Since the sixth city was added later in the program, the models are estimated for the original five and then the six cities.

Market Penetration
The FPO's primarily objective was to change the frequency of transactions among existing flower buyers. Yet, with the broad media coverage within the target cities there is always the possibility of attracting additional buyers with the promotions. This attraction is measured with buyer penetration where the number of buyers in each month was expressed relative to the total number of households. Penetration models were estimated for the target and control cities and the variable of importance in the model was the potential impact of the promotions. If the promotions were influencing market penetration, the impact should be positive and statistically significant with the significance being measured against a value near 2.0 (i.e., t distribution = 2.0 at the 95 percent confidence level).

Back To Top

In Figure 3 the estimated impact of the FPO promotions on market penetration is shown for the target and control cities across the six phase periods. For the models across all demographics, there is evidence of some gains in market penetration in the target cities but statistically one still cannot have confidence in the numbers. Whereas, when restricting the models to females buying for self use the results are quite different as shown in Figure 3.

Recall that the initial promotion efforts were largest during the first phase. Even so, there is no statistical evidence of any gains in market penetration attributed to the promotions. Starting with Phase III as shown in the left columns in Figure 3, there has been an increase in the impact of the promotions on market penetration among female buyers, buying for self use. By Phase IV the coefficient increased and, equally important, became much closer to being statistically significant as seen with the t-value 1.85 (last row in the figure). In the subsequent phases, the coefficient values remained nearly the same as did the t-values. Simply stated, there is increasing evidence that the FPO promotions have attracted additional buyers within the target cities.

As a check, the same market penetration models were estimated for the control cities (see Figure 2). Without exception, there is no numerical or statistical indication that changes in the corresponding phases were taking place in these non-targeted cities. Hence, the positive shifts attributed to the promotion in the target cities are supported by both the near significance in the target cities and the lack of any parallel changes in the controls.

Figure 3 is important because even with specific target goals of frequency increases, the promotions have had some impact in attracting additional buyers. With programs also targeting potential buyers, the analysis suggests that possibly even greater gains could be achieved in bringing more buyers into the fresh-cut flower market. These results are important in highlighting the additional gains above and beyond the primary promotion goals. Later we will use these penetration gains when calculating the overall rates-of-return to the promotions.

Buyer Frequency As stated above, the primary focus of FPO's promotions was to entice existing female buyers to purchase more flowers for self use. That is, through the promotions can the transactions per buyer be increased? Market frequency models were estimated across demographics for both the target and control cities. Without exception, the promotions consistently had a positive and statistically significant impact in the target cities. Likewise, there was a consistent lack of impact in the control cities, thus again pointing to the true measured impact of the promotions in the target cities.

Back To Top

In Figure 4 these strong statistical impacts are shown in the left columns and the totally insignificant responses for the control cities are to the right. For the target cities, the t-values remain above 3.6 and the coefficients are consistently positive and significant. Numerically comparing the left target cities columns to the control cities clearly reinforce the measured promotion impact. These target/control city differences are seen throughout all of the phases. Note, however, the slight decline in the later phases. Part of that must be attributed to the reduction in total effort (dollars).

To provide further insight, the frequency models were estimated across several demographics and purpose types within the target cities. If targeting via demographics/purpose works, there should be evidence of any differences when comparing the households. As particularly seen with the second left column in Figure 5, the largest impact is among female/self use buyers where the coefficient is near .40 compared to the other values. Note that all females and all self users are quite similar to the female/self since they are embedded in the female/self group. Also, for all female and self combinations, the coefficients are statistically significant. In contrast, when estimating just gifts and male buyers, the buyer frequency estimates are not significant. This is particularly true for male buyers. Hence, within the five cities the overwhelming conclusion is that targeting flower buyers works, at least when directed to females for self use.

Using the buyer frequency model, gains in frequency with and without FPO's promotions can be estimated. A household must have at least one transaction per month to be a "buyer" according to our definition. Typically, flower buyers average near 1.60 transactions in a month, recognizing that there are seasonal variations in the buying activities. These buyer frequencies are shown in Figure 6 for the average among the five original target cities. At this point Minneapolis/St. Paul is still excluded in order to provide continuity to the comparison over time. The bottom row in Figure 6 shows the month-to-month frequency without FPO and the next row gives the predicted frequency with the promotions over Phases I through VI. For example, in October 2000 the transaction frequency with FPO is 1.84 and then 1.44 without the promotions. This translates into a 28 percent increase in buyer frequency that is directed attributed to the FPO promotions. Each green bar addition (Figure 6) reflects the gains during a phase of FPO promotions and in every period there were positive values realized. The amount of frequency gain is obviously dependent on the level of promotion activities and the type of media used to delivery the message. Generally, the message theme - "Flowers. Alive with Possibilities" was the same throughout the program life. In Phases IV (June 2002), V (Sept.2002), and VI (June 2003) the percentage gains were 14%, 15% and 12% respectively in the five original target cities.

Back To Top

Total Buyer Transactions
While Figure 6 shows the gains per buyer, it is also instructive to calculate the absolute gains in total transactions over the FPO periods and within the five-target and then six target cities. These totals are going to differ across the cities because of population levels and differences in responses to FPO's efforts. During the six FPO phases within the original five cities, an estimated 2.02 million additional transactions were directly attributed to the FPO promotions. Approximately 32 percent of these gains were in the Chicago area and 27 percent in Philadelphia, then followed with San Diego (16.7%), Detroit (12.6%), and Houston (11.2%). Again, the percentages depend on both the city effectiveness and the absolute number of households in each city.

Figure 7 shows these total gains in transactions by city and across the six phases. What is particularly important is that even if the effectiveness in some cities is smaller, one must also consider the absolute size when judging the full impact of the programs. For example, stopping activities in Chicago even if the marginal responses were weaker could easily reduce the total gains simply because of the size of the market. Secondly, there are clear differences among the cities as seen with the variations in the relative gains over the six phases. Relative growth or at least some consistencies in the gains are seen in San Diego and Houston. Where as, both Chicago and Philadelphia showed substantial declines after Phase III. Initially, Houston showed no response during Phase I but, after some media adjustments, gains were registered. Finally, the first entry into the Minneapolis/St. Paul market failed to show any meaningful gains as shown in the bottom right graph in Figure 7. However, with the investment adjustments made in Phase VI, the Minneapolis/St. Paul response became positive, but still lack statistical significance. Given the increases in this city in Phase VI, one must conclude that at least part of the lack of early response was due to an under-investment in this sixth city. Recall that in Phases IV $39,800 was spent in Minneapolis/St. Paul; in Phase V the promotions were $44,655; while in Phase VI the investment was increased to $142,731. As shown, the impact moved in the correct direction after the additional efforts in this city. Still, one cannot attribute the same level of statistical confidence to the Phase VI numbers for this city relative to the original five target cities.

Unique City Responses - Detroit Case
In the original five cities most of the estimated frequency changes were statistically significant except for Detroit. In Detroit the frequency responses carried the correct signs but were generally statistically less significant than seen in the other four original target cities. Yet when comparing the levels of market penetration, the gains in penetration in the Detroit market were consistently positive and statistically significant. Figure 8 shows a comparison between the market penetration and frequency for the Detroit market which has always been somewhat puzzling during each evaluation period.

Back To Top

What can we conclude relative to the other target cities? In the Detroit market the estimates indicate that the promotions have attracted additional buyers and there is a high level of statistical confidence in the responses, while the gains in frequency, though positive, are not as statistically reliable. An interesting lesson from the earlier PromFlor analysis was that those promotions tended to attract additional buyers within the lower income groups but did not increase the frequency of buying. Similar responses may be occurring in this city.

The Rate-of-Return to FPO
Concluding with Phase VI, the empirical evidence points to positive gains from the FPO experiment during the last three years. Furthermore, some of the gains now appear to be from attracting additional buyers in addition to the frequency gains. In order to fully estimate the rate-of-returns to the generic promotions, both the frequency gains and penetration need to be considered. Contrary to earlier summaries the dollar gains to FPO are now estimated in two ways:

1. Gains = Frequency Change x Buyer Base x Average $/per occasion
2. Gains = Frequency Change x (Buyer Base + Buyer change) x Average $/per occasion

Note that Method (1) is identical to prior evaluations while (2) includes any increases in the number of buyers that are attributed to the promotions. As seen in Figure 9, around 12.5 percent of the total gains (see Table 1) resulted from increases in buyer penetration while most of the gains are from the frequency changes. Given the message focus on existing buyers, this distribution is as expected. In fact, the penetration is somewhat of a bonus since the goal was to gain additional transactions among existing fresh-cut flower buyers.

Back To Top

Adopting Method (2) the ROIs have been estimated as reported in Table 1 (Bottom of page) following the format used in prior summaries. Since the Minneapolis/St. Paul numbers were still not statistically significant, the gains from FPO in Table 1 are still based on the five original target cities. Rows (1) through (7) show in promotion dollars for each Phase with Row (7) giving the total dollars ($6.61 million) for the five cities. Recall that the total six cities advertising investment was $6.84 million and, here, we are only considering the five-city gains. With both the frequency and penetration gains, Column 6 and Row 14 shows the total estimated gain without any correction for the coverage factor associated with the Ipsos sample. For a coverage adjustment equaling one, a total for $33.38 million additional dollars of retail household expenditures is shown. Without the market penetration, this value would have been $31.4 million or about a $2 million dollar gain attributed to the additional penetration. Since it has always been known that the Ipsos data does not give complete coverage, a range of coverage adjustments has been reported in all of the executive summaries. The coverage factors range up to 5.78 based on comparing the Ipsos numbers to those reported by USDA. While subjective, the 5.78 seem high and the actual coverage probably lies somewhere between these ranges from 1 to 5.78. In Row (14) the retail dollar gains are adjusted according to the coverage factor (e.g., for coverage 2 the gains of $33.37 are doubled).

Rows (15) through (21) show the corresponding rates-of-return at the retail by simply expressing the average number of additional dollars in household expenditures generated per dollar of FPO promotions in the five cities. For example, for the same coverage factor of 2 each FPO promotion dollar generated $9.5 additional dollars of expenditures on fresh cut flowers in the target area. Obviously, the gains would be higher with larger coverage adjustments. Next, the same gains are expressed at the wholesale level using a market adjustment of 3.67. That is for each wholesale dollar, the retail price is increased by a factor of 3.67. If this margin adjustment factor is too high, then the wholesale gains will be underestimated and, if too low, the wholesale will be over estimated. From prior analysis, the expectation is that 3.67 is on the high side. Similarly, the wholesale gains are adjusted using the wholesale/producer markup of 1.84. Again, if this is too high, then the producer gains are underestimated. The bottom line gains to the flower producers are reported in Rows (29) through (35) and the ROIs are in Rows (36) through (42). Note that during Phase VI the ROI dropped off after achieving gains in the ROI in prior periods. Yet, even by Phase VI at the producer level, the ROI was greater than 1 without any coverage adjustments. With the coverage adjustments the ROI increases accordingly.

The wholesale factor of 3.67 represents the markup from wholesale to retail for all flowers including those with considerable value added such as arrangements. As suggested this value is most likely on the high side. Given the structural shift to supermarkets and to more bunches, the value added should be considerably less than seen for arrangements. Using arguments adopted with earlier PromoFlor analyses, a wholesale/retail markup of 3.00 is probably closer to a realistic number when the flowers include less value added such as the bunches and single stems. Recognizing that the exact factor is not known precisely, the ROIs reported in Table 1 have been recalculated in Table 2 in order to show the gains with the 3.00 factor. First, Tables 1 and 2 will be identical up to Row (21) . Then starting with Row (22) the total dollars back to the wholesale level increase with the lower adjustment factor. The producer ROIs will be adjusted accordingly given the higher wholesale revenues.

Back To Top

To facilitate interpreting Tables 1 and 2, a selected set of ROIs for the coverage factor of 4 have been plotted in Figure 10. This coverage factor was used only to illustrate the gains. Clearly, the ROIs will differ depending on the coverage scale as discussed earlier. As a general rule there is around a 22 percent increase in the producer ROI when using the lower wholesale/retail adjustment.

Conclusions
After six Phases of FPO promotions in selected U.S. cities, the overall conclusion is that the promotions have impacted the demand for flowers through increasing buyer frequency and, to some degree, attracting additional buyers. Around 87 percent of the gains are from the increased transactions per buyer. The ROIs are positive but still the magnitude of gain is dependent on both coverage and markup adjustment factors.

Across the demographics the greatest response is among female buyers who purchase for themselves. This is consistent with the focus of the campaign target group. Similarly, gains are seen in each of the original five target cities and no comparable frequency gains are registered in the control cities. This control group provides strong evidence that the measured response is truly due to the promotions and not something common to all of the cities during the promotion periods.