Metricsday 2019 - Cómo Media Markt utilizó Dynamic Yield
1 like373 views
The document summarizes how leading retailers like MediaMarkt, Sephora, and IKEA are personalizing customer experiences using Dynamic Yield's personalization platform. It provides examples of how a large fashion retailer fully personalized their homepage by continuously optimizing and targeting content to audience segments. It also discusses how recommendation strategies combined with predictive targeting increased revenue per user by 54.1%. The document highlights how Dynamic Yield uses machine learning to automate testing, recommendations, and predictive targeting to provide more relevant personalized experiences that boost customer engagement and sales.
![Machine Learning at the Core of Dynamic Yield
Experiment Automation
Multi-armed bandit algorithms: Estimates each variation’s probability to be best and allocates
to each variation a fraction of traffic proportional to its probability score.
Predictive Targeting
Proactively identifies opportunities in campaigns to provide users with more relevant experiences
that boost revenue and engagement.
Recommendations
Collaborative Filtering Strategy: Identifies variations most likely to convert a visitor based on the
exhibited preferences of similar visitors.
User Affinity Strategy: Recommends the highest rated variations based on a visitor’s individual
affinity profile and preferences.
Automated Strategy Selection: Our recommendations engine automatically learns which
recommendation strategies performs best for each customer for optimal engagement and results.
Predictive Audiences [In Development]
Automatically generates new, high-value segments based on a customer’s existing audiences.](https://image.slidesharecdn.com/metricsday-anderorcasitas-dynamicyield1-190314224733/85/Metricsday-2019-Como-Media-Markt-utilizo-Dynamic-Yield-30-320.jpg)
Metricsday 2019 - Cómo Media Markt utilizó Dynamic Yield
- 1. CONFIDENTIAL Personalization Pioneers How mavericks like MediaMarkt, Sephora and IKEA personalize each touchpoint Ander Orcasitas Dynamic Yield ander@dynamicyield.com
- 2. “Tomorrow, through personalization, online commerce will accelerate the very process of discovery. Amazon.com uses the internet to create real value for its customers and, by doing so, hopes to create an enduring franchise, even in established and large markets.” – Jeff Bezos 1997 Letter to Amazon Shareholders
- 3. Acquisition For every $92 spent acquiring customers Onsite Personalization Just $1 is spent converting them Today
- 4. The #1 inquiry into Gartner is about Personalization Source: Gartner 2017
- 5. Personalization Impact: • 5%-15% Revenue uplift • 10% to 30% increase in marketing spend efficiency Source: McKinsey, “Marketing’s Holy Grail: Digital Personalization at scale”
- 6. “Dynamic Yield is a Leader… Marketing Leaders seeking a scalable personalization solution suitable for a variety of use cases should consider Dynamic Yield” Gartner Named Us a Leader in Personalization Engines
- 7. Predictive Targeting A/B Testing with CRM & Browse Segmentation Machine-Learning Based Decisioning
- 13. A large fashion retailer has converted every piece of content and banner on their Homepage into Dynamic Content, continuously optimized and targeted at their varied audience segments. The order of the banners and their selected variations are synchronized to guarantee an optimal and consistent experience. Full Homepage Personalization Homepage For internal use only
- 14. Easy Hero Banner Optimization & Personalization Our Homepage Hero Banner is connected to a variations feed, from which Dynamic Yield automatically serves the most appropriate selection of banners to each of our most valuable segments We Continuously Personalize the Hero Banner, Automatically
- 17. “Just For You” Shopping Zone 15% Uplift in revenue per session vs. control group with no “for you” zone
- 18. Social Proof & Urgency Notifications 10% Uplift in revenue per user + 6% uplift in total purchase
- 19. Loyalty based Personalization Highlighting loyalty program with personalized messaging
- 20. Predictive Targeting A/B Testing with CRM & Browse Segmentation Machine-Learning Based Decisioning
- 22. Live Results Similar All slots show DY similar Similar + Bought Together 2 slots similar + 3 slots bought together Control No recs
- 23. Testing Recommendation Strategies Similar All slots show DY similar Similar + Bought Together 2 slots similar + 3 slots bought together Control No recs Revenue per User: (Baseline) Revenue per User: (+27.1%) Revenue per User: (+34.4%)
- 24. Recommendation Strategies with Predictive Targeting Similar + Bought Together To all users Revenue per User: +34.4% Revenue per User: +54.1% Similar Audience “PDP frequent visitors” (39% of visitors) Similar + Bought Together Rest of the users (61% of visitors) Predictive Targeting
- 25. Dynamic Yield Personalization Platform (web, mobile web, apps emails, ads) Personalization Recommendations Optimization Messaging Triggering Web Mobile Web Mobile Apps Email Advertising Segmentation First Party (3rd party data marketplaces, policy controlled data sharing) 2nd & 3rd Party (CRM, ESP, tag managers, POS, loyalty, etc) Enterprise Systems
- 26. Dynamic Yield’s Machine Learning Engine: Everything Is Tested In order to make sure that each user receives the optimal experience, everything is tested in real-time. Dynamic Yield optimizes 1:1 experiences for hundreds of thousands of users at every given moment.
- 27. Predictive Models Probability to Purchase Category & Brand Affinity Price Sensitivity Expected Lifetime Value Increase Retention Improve Customer AcquisitionPredictive Modeling Engine Optimize Conversion Probability to Churn
- 28. The Evolution of Testing with Machine Learning Traffic Allocation ● Manual Results ● Winner takes all ● Winning variation is statistically significant Traditional A/B/n & Multivariate Testing Classic Multi-armed Bandit model Dynamic Yield’s Predictive Targeting + Automation Traffic Allocation ● Dynamic Results ● Winner takes all ● Lower statistical validity + Personalization Traffic Allocation ● One click (Predictive Targeting) Results ● Predictive Targeting identifies winning variation for each audience ● Predictive Targeting’s suggestions are statistically significant
- 29. Automated Segment-Driven Strategy Selection Gathered information per user Revenue Revenue Per User Dynamic Yield’s recommendation engine automatically assesses the relevant data for each individual user and deploys the most appropriate recommendation strategy for optimal results. New Users Occasional Users Frequent Users Savy Users
- 30. Machine Learning at the Core of Dynamic Yield Experiment Automation Multi-armed bandit algorithms: Estimates each variation’s probability to be best and allocates to each variation a fraction of traffic proportional to its probability score. Predictive Targeting Proactively identifies opportunities in campaigns to provide users with more relevant experiences that boost revenue and engagement. Recommendations Collaborative Filtering Strategy: Identifies variations most likely to convert a visitor based on the exhibited preferences of similar visitors. User Affinity Strategy: Recommends the highest rated variations based on a visitor’s individual affinity profile and preferences. Automated Strategy Selection: Our recommendations engine automatically learns which recommendation strategies performs best for each customer for optimal engagement and results. Predictive Audiences [In Development] Automatically generates new, high-value segments based on a customer’s existing audiences.