Arthur W. Allaway, David Berkowitz and Giles D'Souza

Retail innovations, whether they involve new promotional efforts, new product lines, or even new locations, typically have only a short window within which to achieve profitable market penetration before they are duplicated or diffused by competitor reaction. Retailers have to concentrate on gaining the fastest spread and the deepest penetration of customer acceptance of new innovations across their market areas. The research field that deals with these phenomena is spatial diffusion. In theory, the spatial diffusion process evolves in three distinct stages, the drivers of which are information flows. These information flows have a strong geographical orientation and thus yield a predictable spatial pattern of consumer adoptions of the innovation.

This paper examines the processes and driving forces of spatial diffusion in a retail setting by modeling the spread of household-level adoption behavior across a market area following the launch of a new loyalty card program by a very large retailer in a major U.S. city. The loyalty card program was designed to build store traffic, increase basket size, and raise shopping frequency, and was launched with heavy promotion. A very detailed database which included both behavioral and geographical measures on nearly 18,000 adopters over the first year of the program was created automatically as part of the launch. The approximately 300,000 nonadopters in the market area were added to the database as well.

Preliminary analysis of this data supports the presence of an identifiable diffusion process in the speed and spatial pattern of loyalty program adoptions, apparently influenced by the store's marketing efforts and possibly influenced by early adopters of the program. To model the phenomenon, we represent the spatial diffusion process as an event history model, with the time in days between the launch date of the loyalty card program and an individual's adoption of the loyalty card as the dependent variable and a STORE DISTANCE EFFECT, a BILLBOARD EFFECT, and a NEIGHBORHOOD EFFECT hypothesized at major influencing variables.

Modeling results indicate that during Stage One of the diffusion process, probabilities of adoption were affected most by the STORE DISTANCE EFFECT and the BILLBOARD-EFFECT, although the NEIGHBORHOOD EFFECT- the influence of previous adopters in stimulating people around them - was already apparent. After Stage Two, the STORE DISTANCE EFFECT was similar to that of Stage One, the BILLBOARD EFFECT dropped in importance, and the NEIGHBORHOOD EFFECT emerged as a major influencing factor explaining adoption behavior. By the end of Stage Three, the importance of STORE DISTANCE EFFECT declined somewhat, the BILLBOARD EFFECT rebounded somewhat, and the NEIGHBORHOOD EFFECT became the major mechanism driving diffusion. Also in Stage Three, RADIO EVENTS, the special "radio remote broadcasts" held more than 100 days into the program to stimulate new memberships, were highly influential in attracting new consumers who had not yet adopted the program even though they had been exposed to information about it for nearly four months.

Overall, our findings indicate that:
(1) there are three very distinct spatial diffusion stages involved in the response of a market area to a new innovation;
(2) the spatial pattern is established very early in the process;
(3) although likelihood of adoption at any time is influenced by distance, the impact of the store itself, the billboards, and later on, the radio remotes, shows the potential power of the firm in accelerating the adoption process with a coordinated launch campaign;
(4) the earliest adopting group, the Innovators, appear to be the key NEIGHBORHOOD EFFECT "drivers" in influencing new adopters to follow them.

These findings generate significant implications for retailers. First, there appears to be value in a "wide-net" initial launch effort for many retail innovations. Innovators and early adopters, especially those who may have "been waiting for" a particular type of innovation, react very quickly to the launch and appear willing to travel significant distances if the innovation appears to offer immediate and significant value. If these early triers become "advocates" of the innovation, they can then become important (and inexpensive) drivers of the spatial diffusion process, creating secondary "pockets" of adoption around them with positive word-of-mouth, even in neighborhoods several miles from the store.

Second, with geographically selective marketing efforts such as billboards and neighborhood newspapers, retail managers might be able to generate specific spatial diffusion patterns for their particular marketing initiatives in the same way they currently select reach and frequency patterns for advertising strategies. Managerially, this is important. If there are neighborhoods which evolve more quickly than others, expenditures can be shifted from saturated neighborhoods within the market area to those neighborhoods still diffusing or those which have not taken off at all.

Finally, the widespread availability of scanner data and the development of large numbers of loyalty programs and other customer relationship management programs tying scanner data to individual customer characteristics should make this type of research easier in the future. The addition of Geographical Information System (GIS) software and the powerful data storage products and statistical modeling software available to managers and researchers can yield significant empirical research to both test existing theories and to create new ones.