A well-known retail organisation was marketing their General Insurance products to a large database of potential customers and wanted to improve their ROI. StatCore was engaged in a project to gather insight that would help the client improve campaign performance by selecting a more effective, targeted subset of potential customers to contact.
Tony quickly realised the potential for the client to make better use of their historic campaign performance data to determine who received marketing messaging in the past and whether it affected their decision to purchase. Tony extracted and manipulated hundreds of millions of rows of data tables using complex and efficient SQL queries, then built new targeting models to guide the client to direct their messaging to the people who were most likely to respond positively.
Tony applied uplift modelling, which divides a sample group into four parts:
- Persuadables – people who would only buy when treated with the marketing message
- Sure Things – people who would buy anyway, even if they were not treated with a marketing message
- Lost Causes – people who would not buy, where marketing has no effect
- Do Not Disturbs – people who would buy if left undisturbed. If treated with a marketing message, they may reverse their decision to buy.
Once Tony’s new model was applied and the client targeted only the ‘Persuadables’, the ROI dramatically improved. A/B campaign tests confirmed that the Cost Per Acquisition (CPA) of Tony’s recommended ‘contact group’ was £100 compared with £312 for the ‘do not contact’ group.