Department stores and fashion retailers have specific data idiosyncrasies which can limit the value of insights and analytics initiatives. If you are looking to increase the capabilities of your department store customer analytics initiatives, below are a handful of simple tips which can effectively convert some of these challenges into opportunity and competitive advantage.
CUSTOMER VISIT VERSUS TRANSACTIONS
For the department store embarking upon customer-centric analytics versus more traditional product-centric analytics, a good place to pause for consideration is to be very deliberate about the distinction between the concept of a transaction and a customer visit. Most non-department store retailers have one bank of cash registers where customers pay, generally on the way out – meaning that the customer will travel around the store, fill their basket and pay for all purchased product at one time.
This is generally not the case in department stores, where it is more likely that the cash registers are spread around the respective departments. The multi-department shopper behavior in a department store then is characterized by a shopper picking up goods in one department, paying for them there, then moving to the next department and paying separately at the next department. Therefore it is quite conceivable that a single unique shopper could be behind (say) three distinct transactions within a store in one hour. If this is not considered, it will be very easy to over-estimate this shoppers visit frequency (by a factor of three), as well as to have a very limited insight of what their entire basket looked like on this single trip.
The solution to this is generally to design a (very simple) algorithm which allows that any unique tracked shopper (e.g. via loyalty program) who had multiple transactions within (say) 2 hours had those transactions consolidated into one transaction. That is to say some ETL is performed to create a ‘synthetic transaction/basket’ which contained all of the SKUs that shopper purchased in that 2 hour period. In simple terms you will be consolidating all of the shoppers individual transactions in that time block into one ‘super transaction’ which includes all the SKUs from the component transactions). The determinant of how many hours should be the cut-off is a matter for you to determine, based off what behavior you observe in your individual shoppers. I’d suggest that being broad in this is superior to being too tight – unless you have reason to believe your shoppers are routinely visiting you more than one time per day.
It would be fundamental to have color as an attribute of department store clothing articles. The utility of knowing if black jackets or red jackets are selling better and trending up or down are clear. However in many cases the retailers product attribute database can contain thousands upon thousands of colors (effectively every color that every designer has thought to name an article, making this attribute somewhat useless. What is important to remember is that these colors while useful for marketing are of little help in insights. The original colors should definitely remain intact, however ideally they will be augmented with a meta description of color (e.g. ‘Light Blue’, Dark Blue, etc). The goal is to streamline your hundreds or thousands of colors down to a manageable number which are pure signal, as opposed to distracting noise. (11Ants Retail Insights Cloud user tip: This field will be something like CL_Summary Color).
Similar to the above on colors, in many retailers the actual sizes becomes unwieldy (disparate sizing conventions, repetition with minor typos, etc), and valuable signal is lost at the expense of noise. (11Ants Retail Insights Cloud user tip: This field will be something like CL_Summary Size.)
COLLECTIONS: SEASONAL & OTHER
The nature of collections can create complexity that is often hard to manage. e.g. a SKU is found the Spring 2020 Collection, the Summer 2021 Collection and the also the Fall 2021 Collection. Traditional ways of managing this as regular attribute (e.g. 11Ants Customer Label (‘CL_’) could mean hundred of columns of attributes. What is preferable is to use the concept of a tag (think hashtag), and then every article can have a list of the tags that are relevant to it. (11Ants Retail Insights Cloud user tip: for the above example populate the column tags with the following value “Spring 2020 Collection | Summer 2021 Collection | Fall 2021 Collection’ The pipe symbol is simply applied as the separator, as many times as required. )
HIERARCHY STRUCTURE TO ACCOMODATE MAXIMUM FLEXIBILITY OF DRILL DOWN
Hierarchies are pretty straightforward and no different than for any other retailers at the top end (i.e. Department > Category > Sub Category), however by the time you get to the article in fashion retail, you have still got some way to go. What I mean by that is that ‘Coca-Cola Regular 2L’ is pretty much the leaf of the tree for a grocery retailer. However ‘Butterfly Denim Jacket’ for the fashion retailer is by no-means sufficient. Some of the ways you will inevitably want to look at this:
– How many Butterfly Denim Jackets did I sell?
– How many of each color did I sell?
– How many of each size did I sell?
Consider building multiple product hierarchies e.g. H1L1_Department > H1L2_Category > H1L3_Sub Category >H1L4_Article > H1L5_Color > H1L6_Size and then another which is identical, however follows this opposing pattern at the bottom of the tree: H2L4_Article > H2L5_Size > H2L6_Color. ((11Ants Retail Insights Cloud user tip: for the most part you will be able to analyze higher levels of the hierarchy just as if they were SKUs, so there is no problem if the level above the SKU is redundant, if it helps maintain consistent architecture of the hierarchies).
If you are interested in learning more about how 11Ants Retail Insights Cloud can help you reduce time to insights by tens of thousands of hours per year and transform your department store customer insights initiatives, please contact us to learn more.