13.8 Discussion and next steps

The presented approach is a typical example of the normative usage of a GIS (Longley 2015).We combined survey data with expert-based knowledge and assumptions (definition of metropolitan areas, defining class intervals, definition of a final score threshold).It should be clear that this approach is not suitable for scientific knowledge advancement but is a very applied way of information extraction.This is to say, we can only suspect based on common sense that we have identified areas suitable for bike shops.However, we have no proof that this is in fact the case.

A few other things remained unconsidered but might improve the analysis:

• We used equal weights when calculating the final scores.But is, for example, the household size as important as the portion of women or the mean age?
• We used all points of interest.Maybe it would be wiser to use only those which might be interesting for bike shops such as do-it-yourself, hardware, bicycle, fishing, hunting, motorcycles, outdoor and sports shops (see the range of shop values available on the OSM Wiki).
• Data at a better resolution may change and improve the output. For example, there is also population data at a finer resolution (100 m; see exercises).
• We have used only a limited set of variables.For example, the INSPIRE geoportal might contain much more data of possible interest to our analysis (see also Section 7.2).The bike paths density might be another interesting variable as well as the purchasing power or even better the retail purchasing power for bikes.
• Interactions remained unconsidered, such as a possible interaction between the portion of men and single households.However, to find out about such an interaction we would need customer data.
  In short, the presented analysis is far from perfect.Nevertheless, it should have given you a first impression and understanding of how to obtain and deal with spatial data in R within a geomarketing context.

Finally, we have to point out that the presented analysis would be merely the first step of finding suitable locations.So far we have identified areas, 1 by 1 km in size, potentially suitable for a bike shop in accordance with our survey.We could continue the analysis as follows:

• Find an optimal location based on number of inhabitants within a specific catchment area.For example, the shop should be reachable for as many people as possible within 15 minutes of traveling bike distance (catchment area routing).Thereby, we should account for the fact that the further away the people are from the shop, the more unlikely it becomes that they actually visit it (distance decay function).
• Also it would be a good idea to take into account competitors.That is, if there already is a bike shop in the vicinity of the chosen location, one has to distribute possible customers (or sales potential) between the competitors (Huff 1963; Wieland 2017).
• We need to find suitable and affordable real estate, e.g., in terms of accessibility, availability of parking spots, desired frequency of passers-by, having big windows, etc.