Marketing + Media Alliance (MMA) created the Consortium for AI Personalization (CAP) in 2022 with the desire to test whether AI Personalization technology can create omnichannel experiences that outperform the ads and communications customers and prospects get today by a large enough margin to justify any additional costs associated with the technology.
CAP is a major initiative that MMA believes could deliver 30-60% increase in marketing performance. We believe that if fully rolled out, this could move the stock price of a company through increases on marketing productivity. We have completed tests with 7 marketers and the results have been even stronger than expected, with average performance gains measured against the marketer’s chosen KPI exceeding 100%.
We continue to run tests with more marketers, to validate these findings at greater scale and to test the approach with new media types and by making different data available to the ML model. We will also test using AI to drive personalization deeper into the customer experience, from landing pages and the purchase process to optimizing across channels.
Research Objectives:
- Quantify the value of an intelligent (AI) experience delivery engine driven by contextual data.
- Derive benchmarks and economic impact projections of this technology and case studies
Research Questions:
- Can AI metacontextual personalization of creative assets improve performance, and by how much (est. 30-60%)?
- To the extent that AI personalization comes with an extra charge, is lift worth the upcharge (ROAS)?*
- How do #1 and #2 vary...
- by product category, ad format and/or channel and what about optimization at different points…
- CTR -> Page Visits -> Transactions -> Margins
- And is there possible impact on business valuation
Immediate Return on Investment: We see on average a +195% lift from AI-driven Personalization for creative ad placement.
With seven studies completed so far, results show that the upside is big.
 |
 |
 |
 |
 |
 |
 |
|
|---|---|---|---|---|---|---|---|
| KEY INSIGHT | Upside is BIG | Audience Matters | Creative Diversity | Power of Interactions | Constant optimization | Gen AI + AI Optimization | Diversity in Versions is Key |
| KPI Improvement | +259% | +136% | Test 1: +68% Test 2: No lift | +188% | 161% | +197% | +65% on average |
| Media Type | Display | Display | Video | Audio | Display | Audio | Display |
| KPI being optimized | Webpage visit | Web form submit | Webpage visit | Website visit & App install | App registrations (UK) | Quote Starts | Ticket Orders |
| # versions | 72 | 81 | 15 | 16 | 90 | 96 | 2,000+ |
| Ecosystem | Open web | Open web | Open web | Open web | Open web | Open web | Open web |
| Data available to ML model | DMA, Time, Day of Week, Device OS, Connection type (cable, mobile, corporate, etc.) | DMA, Time, Day of Week, Device OS, Connection type (cable, mobile, corporate, etc.) | Segment (core, youth), Device OS, US geo, Connection type (cable, mobile, corporate, etc.) | DMA, Time, Day of Week, Connection type (cable, mobile, corporate, etc.) | Time, Day of Week, Connection type (cable, mobile, corporate, etc.) | Time, Day of Week, Connection type (cable, mobile, corporate, etc.) | DMA, Segment, Time, Day of Week, Device OS, US Geo, Hardware type, ISP, ASN |
Two-Phased Approach
Learning Agenda
- Phase I (“Walk”) - Using metacontextual data (non-PII) to drive AI-driven Personalization
- Phase II (“Run”) - Using other data signals and identity, targeting lower-funnel KPIs and more
AI Techniques
AI Techniques
This application of AI/ML works by identifying cohorts and optimizing pre-approved elements and creative to each. MMA's approach uses two ML techniques combined with unsupervised learning:
- One-Hot Encoding
- K-Modes Clustering
