Data Science in action: Helping a major National
Charity to increase revenues
Dr Marcus Brownlow Dr Inna Kolyshkina
Dr Marcus Brownlow is an advanced analytics and data visualisation professional with 20+ years of academic, government and industry experience in the management, analysis and communication of complex data. He has been a SAS software user since 1988 and has delivered data-driven insight in scientific research, telecommunications, insurance and other industries. He likes the stories that numbers can tell, and likes them more when they have a business or other relevance. Depending on the context, the stories can be about risk or opportunity, profit or loss, money or people, business or society. The important thing is to be able to convey this information as succinctly as possible.
He is a Fellow of the Governor's Leadership Foundation and holds a PhD from The University of Adelaide. Dr Marcus Brownlow
Dr Inna Kolyshkina is a data science and business analytics professional with 20+ years of government and industry experience in the management, analysis and communication of complex data. She has been a SAS software user since 1997 and has delivered data-driven consulting projects in the industries such as insurance, banking, telecommunications, Government (State and Federal), transport, FMCG, airlines etc. Inna is a Founding Chair of Institute of Analytics Professionals of Australia, and Adjunct Senior Lecturer in University of South Australia and holds a PhD from The University of Kharkov, Russia.
Dr Inna Kolyshkina
Bequest modelling for a large national charity
Background ● A large national charity maintained a register of all its donors. Some of the donors would leave money in their Will (bequest).
● The charity tried to identify such donors and ensure that they get reached by the right amount and channel of marketing activity.
Client Issues Client issues The organisation faced the following challenges:
● ●
● ●
Difficulties in identifying donors who might become bequestors and who are the best target for marketing to Risk of donor attrition caused by the donors being over marketed to or receiving wrong marketing messages Limited call centre resources and hence the need to allocate the resources better in order to achieve “more with less” Donor data was of limited use as it was kept in a difficult to work with sources
The client needed to improve their marketing ROI and minimise donor attrition by optimising their targeting, marketing activities, marketing spend and resource allocation
Project outcomes
Benefits to the Client Benefits to the client 1. We segmented their donor base on likelihood of bequest. This allowed the organisation to ● optimise their marketing activities and spend by targeting the right donors instead of approaching everyone on the list ● optimise resource allocation so only the donors with high propensity to respond were approached by phone calls and visits ● optimise their donor acquisition strategy 2. We recommended important data improvements such as data storage, data enrichment that allowed easier insight driving in future
Bequest propensity model More detail on segments quality
How diditit How we we did
Step 1. Data pre-processing typically in Analytics projects right data pre-processing accounts for 75% of the project success
Data Background ● Good data management, but operationally-focussed
● Designed for call centre staff o o
execute marketing campaigns update donor records
● Operational CRM(ish) design, but also o o o o
donor demographic data campaign response donation history contact history
Analytical base table
Matrix containing attributes for each client: ● demographic (age, marital status, suburb …) ● communication (preferences, contact frequency, contact type …) ● engagement (response to various marketing campaigns… ) ● donation (frequency, type and timing of donation(s) …)
Data reality Contacts
Campaigns
Demographics
Donations
Data Pre-processing All done using data step and Base SAS procedures Repetitive tasks automated using %macro processing
Structural transformation
Stacking: proc transpose Concatenate and merge: data step
Data profiling
proc freq, %macro
Data cleansing and standardisation
data step
Data transformation and recoding
data step
Step 2. Analytics modelling
Segmentation via Analytics Modeling When the data was pre-processed, a combination of predictive analytics modelling approaches were used to derive insights from the data by segmenting the donor base into the segments with similar behaviours and propensity to bequest
Contact Us Dr Marcus Brownlow
[email protected]
Dr Inna Kolyshkina
[email protected]