Business & Data Fail - The Little Pink Skirt
Don't you know those issues when you are shopping offline or online? Inexplicable shortages, "economic" / marked down products that are more expensive than the original ones, or Kafkaïan conversations with support after a failed digital sale?
I have the same, and as an experienced data practitioner in the retail field, every time it happens I try to understand what has gone wrong, on the business side as well as the technical/data side of things.
This is the first edition of the…
Recently, I had to find a little pink skirt for my daughter who is joining a classical dance course.
I turned to my usual go-to store for everything related to sport - an international retailer specialized in sport items.
In short, what happened?
I have been visiting the same store for weeks before the start of the school year and the right size was never in stock.
It was available online, so I ordered it along with other items to be delivered in my store...
Everything seemed ok but the pink skirts were not delivered: it turned out that they were removed from my order because the warehouse didn't have them...
I finally went for a 50 minutes ride to find them in one of the few stores that had an homogeneous stock coverage of the different sizes.
All of this provides us a nice setup to investigate business & data issues:
demand planning at the national level and its breakdown in sizes
stock balancing in this retailer's stores
how the architecture around the order management system may have led to this disappointing situation
the missing link between order management and CRM to handle order processing issues
Demand Planning
Let's put ourselves in the shoes of the supply manager working with her data team on the Dance accessories product category. What could be her approach to ensure that the overall stocks for each size covers the needs of the whole French stores network?
Little pink skirt sizes for children are based on ages: 3-4 years, 4-5 years, ... up until 13-14 years, after which girls start to pick their clothes in the adult sizes.
What is the pattern of the number of little dancers by age in France?
Luckily, the INJEP (institut national de la jeunesse et de l’éducation populaire) collects data from sports federations and publishes them as open data.
Let's have a look at the data published by th)e INJEP for 2021:
Focusing on children:
That is around 21,000 for all kinds of dance, not only classical. We can safely assume that contemporary and modern jazz dancers are not wearing pink tulle skirts.
Our retailer probably has a better idea of the total number of pink dresses it can sell in a year. Our concern is more focused on the breakdown in sizes, so let's make a strong an simple hypothesis that around 50% of dancers are classical dancers.
This gives us a French market of 10,000 pink dresses in sizes ranging from 4 - 5 y. to 14 - 15 y.
Can we use this data to plan at the item size level?
Yes and no. INJEP is not providing data for every age, and the french dance federation hasn't responded to my data request. I believe that, if not already done, there is a win-win to find between retailers and federations to leverage aggregated licensees figures.
We can make a few assumptions to build an initial guess that only the retailer will be able to confirm:
licensee numbers start quite strong at the age of 5;
peak dance frenzy is around 9 - 10
then there is a growing churn when getting closer to and entering high school
Let's quickly mix two statistical distributions and it could look like this:
We can make it quite fit the INJEP data if we bin ages the same way:
It could be a reasonable first guess for production orders and planning warehouse stocks.
How do you think it compares to this retailer's yearly planning?
Stock balancing
Let's continue the exercise of putting ourselves in the shoes of this retailer's business teams. How are stocks balanced in the various stores?
Our story starts too late to know what where the stocks when stores were first delivered the goods for the start of the school year. These days it starts even before kids are even out of the previous school year.
What we know is:
For 3-4 weeks, since the beginning of August, I have been going to the same place in the store to check the available skirts. For 3-4 weeks, no skirt under 10 y.o: nearly all skirts were for 13-14 y.o.
At the time of this writing, in all dance schools in the region, courses have just started, or are starting, and most girls have their dance skirt. It is rather useless but you know, the others have one.
I just checked a few days ago the stock levels in 60 stores, sorted by increasing distance from a point I located arbitrarily:
(The store I visit every week is number 18. It looks like I have no luck because it is one of the least stocked in terms of pink dance skirts.)
What do we see here and what can we say about it?
Some stores don't have any stock at all, but this might be an assortment choice based on location or the size/concept of the store.
Over 60 stores, only 4 seem to have enough stock in sizes up to 8 - 9 (n° 1, 3, 21, 53, maybe you could add n°27),
All other stores have very few stock left with an inhomogeneous distribution of sizes. It won’t be enough to keep the category for the rest of the school year.
Skirts starting at size 10 - 11 are much more difficult to find. They were either voluntarily or unvolontarily understocked.
We can’t exactly say if the items are pushed to stores, pulled by stores, or a mix of both depending on product categories.
What do we need to know to build our planning?
We would like to know the sell-through potential of each store, which means working on estimating the size of the following concentric potentials:
estimate the trade area covered by each store and the corresponding number of households. This could be drafted with a map of stores - including competing ones - and the categories they cover.
determine the ratio of little dancers in these households. Sociodemographic data from INSEE about household compositions and ages should help, along federations data if available.
refine this estimation by modelling customer premium choice. Analyzing internal sales and external data, can we say where parents will more likely go to either Repetto or a discount all-purpose store to get little skirts?
This 3-level exercise is not specific to pink little skirts, it could and should be processed and industrialized for all kinds of product categories.
See what gold open data provides us about practitioners of each sport... and this is only aggregated data over the whole country:
Getting advantage from all these data would help build a much more robust estimate for the physical retail network:
less missed sales
less slow movers in stores
... and more stock available for e-commerce!
Order Management
And this is the last part of our investigation: how could I order skirts from the website if there was no available skirt for e-commerce sales?
In theory (see drawing below):
inventory availability data should be fed in real-time to the sales channel (here, the e-commerce part) through the Order Management Software, so that I could not even order an unavailable item in the first place
sales orders should be sent in real-time to the Warehouse Management System so that ordered items are immediately reserved/sent to the shipment area... and for example are not used to fulfill pull orders from stores. (That's where quotas may be involved.)
What happened here?
The inventory availability data published by the WMS to the OMS and then by the OMS to the e-commerce front-end was possibly one or more days old. A daily batch job?
The order fulfillment status, also published by the WMS to the OMS and then by the OMS to the e-commerce front-end, took 4 days to get into my e-mails after my order, because it was only communicated to me when my whole order was delivered in-store.
How is inventory availability data currently fed to the digital sales channel?
Event management and CRM
Did you say silos ?
I received two e-mails about my order:
August 26th: "Your order is confirmed!", with the skirts seemingly available
August 30th: "Your order is available in store!", starting with the lines:
"It's waiting for you!"
"Order status: available"
After this, who would scroll until the end of the e-mail to see a section named ”Missing items”?
There seems to be a missing link between order management and marketing, and a missing CRM scenario.
Between these two e-mails, there wasn't an alert saying "Dear customer, we're really sorry but something wrong happened on our end".
The only communications I had were perfectly aligned with the status, correctly structured to fit the technical reality. But it didn't surface what was important for me, the customer, and didn't provide me what I would have liked to know:
there was an issue on the retailer's end
my card was only charged for delivered items, or my money was sent back
ideally, there would have been an hybrid, long tail solution to fulfll my order from stocks in other stores
Do logistics teams work with CRM teams to build a scenario for handling unfulfilled orders? (at least, while the project for getting robust availability data is underway)
That's all for today! What do you think?
I focused on one particular experience with one particular retailer but this kind of issue happens at LOTS of B2C companies.
Starting from one bad customer experience, we uncovered:
demand planning strategy and the data to be ingested and analyzed to help
balancing and again, the data that could be leveraged
possible architectural and implementation issues in order management
the importance of communicating significant customer events between two units
I hope this example showed how every issue is at the same time a business and data issue and how making operational excellence a reality is a goal for both of them, together.
If you are a business leader or a data leader and you could relate theses issues with some of your own, feel free to reach out!
If you are leading the efforts on one of these topics at this retailer and you want to discuss today’s article, drop me a word privately on LinkedIn or at gansanay AT gmail DOT com.