sku analytics loyalty nz sunz 2012
Post on 19-Oct-2014
1.025 views
DESCRIPTION
TRANSCRIPT
SKU Analytics and the Triumph Over Stone Age Segmentation Methods
Vince Morder, Loyalty NZ
Milo Davies, SAS
2012 SUNZ Conference, Te Papa, Wellington
SAS and Loyalty - A Great Partnership
Rich data +
Loyalty’s Techniques +
SAS tools
=
Loyalty Data (Historical)
Customer
Demographics
Age
Income
Gender
Address Email
Mobile
Other Sources
Lifestyle Survey
Census
KCUBE
LINZ/QV
Motor Vehicle
Final Prepare
Example: RFM Segmentation
• RFM segmentation is a behavioural based segmentation built on:
Frequency: How many visits have they taken?
Monetary Value: How much does a customer spend each visit?
Recency: When was last transaction customer did with you?
• A segmentation is built across all customers for a particular retail partner over some determined observation period.
RFM Segmentation
1 visit only Customers
27% of
all spend (Not part of this RFM analysis)
Low Value Infrequents
30.6% customers 0.6% spend
Medium Value Infrequent
26.2% customers, 3.1% spend High Value Infrequent
7.6% customers 27.6% spend
Medium Value 11.6% of base 7.4% of spend
Low Value Frequents 17.5% customers
8.7% spend
High Value Frequents 6.5% customers, 52.6% spend
Low Monetary Value High Monetary Value
Low Frequency
High Frequency
• To add further depth and insight, we can profile the demographics of each segment
• We can also track movement over time.
Example: Tracking RFM over time
HVF’s are mostly females. All others are greater proportion males.
There is a strong skew in highest income areas, lowest deprivation deciles towards higher rfm segments
HVF’s are most predominant in the 40-60 age range, HVI are older (retired age), spends lots, but less frequently
Example: Profiling the RFM segments
Along Came SKU….
(S)tock
(K)eeping
(U)nit
Literally, billions of records at the basket level
Loyalty Data (Current)
Customer
Behaviour (SKU)
Outlet
Basket Value
Items Points
collected
Time & Location
Demographics
Age
Income
Gender
Address Email
Mobile
Other Sources
Lifestyle Survey
Census
KCUBE
LINZ/QV
Motor Vehicle
Final Prepare
New methodologies using SKU data
• SKU data enables us to get an even better view of shoppers in the retail market.
• If used correctly, it can help us to understand the motivations behind buying decisions.
• If we can improve our understanding of our customers’ motivations we can become a lot more sophisticated in our decision making and our ability to keep customers engaged and loyal to retailers.
View of the customer using traditional data
Profiles using SKU data
Let’s take a look at some examples….
Milo’s Supermarket Receipts
ORGANIC
GLUTEN FREE
FRESH
HIPPIE !!
Milo’s Supermarket Receipts
GLUTEN FREE
READY MADE
FANCY BEER
NAPPIES
Milo’s Electronic/Whiteware Purchase History
• Focused on healthy/diet eating
• Happy to buy premium products
• High end, yet stylish, hardware
Example: Milo
• Except for that beer!
• Vacations involve going overseas
• Buys big pack items
• Buying for a family/kids
• Prefers convenient, easy cook meals
Example: Vince
• Low end electronics
• Vacation locally
• We have just looked at two different customers with two very different sets of products purchased with our partners.
is it healthy product?
Is it for a family?
is it expensive?
Is it functional vs.
showy, or both?
Etc...
But how to make sense of all these products and all these customers?
• Before we can understand our customers we must first understand the types of products they buy (rather than the product themselves) and be able to answer questions like:
What conclusions could we draw
• What is likely to be relevant and engaging to Milo is unlikely to be relevant or engaging for Vince
• The SKU data has the potential to help us identify these different customers so we can be relevant and engaging to both these customers.
• LNZ is in the process of classifying our partners retail products against our ideal dimensions.
Kids
Quick Gourmet Healthy
High Price
Budget
Alcohol Fresh Organic
Scratch
Enter the ideal dimensions
Showy
• E.g. Tuatara would have a high association with alcohol as well, but also, load quite highly on the ‘Showy’ dimension as well. Low loadings for Tuatara on the scratch dimension
• The double oven could load high on gourmet, scratch, showy, and high price.
• Points in the direction of a perfect representation of something we imagine.
The Secret Sauce
• There are tens of thousands of products across our partners and it would be impossible to manually try and classify all of them.
• To make it more difficult what I think is ‘healthy’ - you might disagree!
• E.g., This pulse monitoring watch could be for a health nut or someone who just suffered from a heart-attack.
• Instead we rely on an algorithm that sorts through characteristics of products to statistically determine how much they load on to our designated ideal dimensions.
• We then trawl and loop through the entire retailers’ transactional database to ‘score’ all the products customers are purchasing.
Milo’s Shopping Profile • Once we have scored all products we bring it all together and create a shopping
profile for Milo
• Looks like we don’t need to worry too much about giving specials to Milo!
Vince’s Shopping Profile • Once we have scored all products we bring it all together and create a shopping
profile for Vince
• Vince might need to get his cholesterol checked!
How this helps our partners
• We can apply cluster analysis to group together customers who share similar motivations.
• By understanding our customers’ primary motivations we can apply it across our business by:
Increasing the relevance of marketing activity through the clustered segments or leveraging one of the attributes.
Improving the targeting, personalisation and relevance of our communications.
Get greater insight into the profile of shoppers visiting different stores. Can assist in areas from ranging to more relevant ATL offers.
Example: Applying to campaigns
• A DM was sent to 10,000 existing retailer customers to promote a high end product X. The campaign targeted two audiences:
Customers who purchased product X.
Customers who purchased other similar speciality products.
• The campaign generated an average response rate of 6.9%
What happens when we overlay our “gourmet” attribute?
• We allocated each customer a HML segment based on their gourmet attribute score.
• Heavy gourmet customers responded at nearly double the rate of the next closest segment.
• Based on new dimensional profiling techniques, product X has a high gourmet attribute score.
Example: Enhanced Communication
• Question: Because you spend a lot at the retailer, does that mean you will have an interest in a their magazine?
• Not necessarily – you may spend a lot at retailer but heavily focused on value/everyday items because you’re shopping for a large family.
• To increase the relevance of the magazine, we can overlay customers’ behaviour dimensions in combination with the RFM to give a much more optimised target group.
Value (RFM)
+ Behaviour = Relevant & Optimised
• One of our retail partner’s magazine is an upmarket communication originally planned to target the most valuable customers based on their RFM segment.
What’s next?
LNZ is currently working with it’s partners to implement and begin leveraging these behavioural dimensions.
Plans are in place for our analytics to extend to
Social network data
Mobile applications
The vision for the LNZ Customer Intelligence Team is to be the undisputed Customer Loyalty experts