The intensive discussion about customer behavior (emotions, motivations, buying behavior – through shopper research and Shopper Marketing) and assortment control or tuning (data analysis, data warehouse, measures – through Category Management) is often seen with different priorities but a total common perspective.
GS1 Germany describes the demand: "In the long term successful Category Management is based on quantitative and qualitative research, the shopper research, as well as the systematic assessment of each category by sales data and panel data ... Successful Category Management is supported ... by a strategic Shopper Marketing and solid Customer Relationship Management (CRM)”. (Source: Category Management cooperation for success - http://www.gs1-germany.de/gs1-solutions/category-management/)
So far shopper research is pushed with special testing or analysis methods. Eye tracking, consumer interviews or in-depth analysis of panel (household) data are just a few examples of shopper research.
Based on the given analytical usage of cross selling analysis (SPSS nodes or algorithms – A-priori node, CARMA node or sequence node), the integration of shopper research (Shopper Marketing) and assortment management through category management is increasing. Important steps to deepen cross selling analysis with high shopper research potentials include the collection of detailed sales data from the identified customer (customer loyalty programs, etc.) over an extended period of time, as well as the systematic accumulation of sales data with more relevant additional information.
Some analytical cross selling scenarios can indicate the potential benefits for shopper research very well:
Evaluation of influencing factors on customer behavior by "virtual article or item”
- Base: A Priori node, CARMA node (SPSS Modeler algorithms)
Drivers for the customer behavior can be very multifaceted. Relevant elements and the derived “virtual item” at the time of purchase can be for example:
- store: attributes of site, store marketing and competitors (specific store values like park space size, assortment size, advertising, competitive actions...)
- placement of the product: shelf, special placement,
- weather: each distinct value (temperature, precipitation, etc.)
- events: local or global
- time specific influence: time of day, seasonal aspects
- important, visible characteristics of the purchase situation: for example, purchasing with / without child or children, individual yes/no, perceived age of customer
Selected drivers or elements can be used and included as a "virtual" article or item - limited in time - in the receipt (the sales data). This item will become one element of a possible cross selling analysis. An example in association analysis:
“Discounter ABC below 5 km distance” (virtual item)
>> Sales of >> Sugar 500 gr. pack (Confidence 15 %, Lift 3,1)
“Discounter ABC - 10 km + distance” (virtual item)
>> Sales of >> Sugar 500 gr. pack (Confidence 10 %, Lift 2,3)
These analysis results show that the distance (close neighborhood) of a specific discounter will possibly have an impact on selected items (in example: sugar). If you know this, you can initiate different actions with regard to retail marketing (pricing, promotion, placement etc.).
Customer with child“ (virtual item) Store 34
>> Sales of >> assortment group sweets (Confidence 35 %, Lift 1,6)
Customer with child“ (virtual item) Store 12
>> Sales of >> assortment group sweets (Confidence 25 %, Lift 1,8)
These analyses show that possibly the store layout in the different stores have a powerful influence on the customer behavior of defined customer groups (with children). In the analysis of other influential factors further evidence can be analyzed with regard to the customer behavior.
To be careful: The use of the construct "virtual item" will include "artificial" elements in the receipt or sales data and distort the key figures of the cross selling analysis. It is important to decide whether targeted selections of receipts, samples or store comparisons can be used more meaningful.
The concept of "virtual item” requires a well-designed analysis model and the usage and "enrichment" of data (receipts, transactional data) has to be harmonized with the analysis objective. To have this in mind will ensure that the analysis results will extend the findings of detailed shopper research activities.
Discovery of brand loyalty or churn
- Base: Sequence node (SPSS Modeler algorithms)
Based on identified customers (loyalty program) and their receipts over time (for example: 2 years) it is possible to generate "models with sequential rules" (Sequence node) which will show changes in the underlying buying trends or behavior. The interesting information in the sequential association rules are the specific buying trends over time (all purchase events) and thus extend the known cross selling method of analysis (reference there: all transactions without time sequence).
For example the subsequent analysis is possible:
Customers buying “Persil - detergent”, buy very rarely “Ariel – detergent” (or vice versa) – in a 2 years’ time frame
(proved by the specific metrics from sequential pattern analysis)
If the “sequential pattern analysis” is limited to a point in time with or without promotion or advertising for each item (Products: Persil / Ariel) you will find out how "stable" brand loyalty is under the effect of different marketing activities or price variations.
Customers buying "Eau de toilette > Brand Boss - Name ABC “, buy very often over time the "Eau de toilette > Brand Calvin Klein - Name XYZ " – in a 2 years’ time frame
(proved by the specific metrics from sequential pattern analysis)
You possibly will find out that there is an ongoing brand migration or erosion affected by advertising of the manufacturer or changing customer preferences. Additional aspects are variations in shelf presentation or retail / CP marketing or other actions. Often you can see creeping (underlying) changes in behavior, some of which have a “general meaning” others are only applicable for the specific retailer.
Often these analysis are very important with respect to the selection or optimization of assortment and are often indicators of changes, the retail can respond early on.
To develop or improve shopper research this way (above examples) will have a specific relevance and will discover the changing customer behavior in correlation with existing shopper research activities. The idea of the "virtual item" and the use of sequential shopping patterns can be complementary and will bring new knowledge to this discipline.
Creativity and targeted development of the potential benefits of the cross selling analysis provide a wide variety of interesting and extended analysis capabilities, combined with the existing shopper research mechanisms.