Data & AI Master Class - Jan Klawer, Artefact
6 likes869 views
Jan Klawer, Managing Director of Artefact Middle East and Africa, presented at the DigiMarCon Middle East 2018, focusing on how data and AI are essential for marketing. The document highlights the transformative impact of AI and data on consumer experience and marketing strategies, emphasizing the need for brands to adapt to new industry standards driven by leading tech giants like GAFAX. It discusses various applications of AI in marketing, outlines the marketing value chain, and suggests methods for automating and optimizing marketing processes to enhance efficiency.
Data & AI Master Class - Jan Klawer, Artefact
- 1. Jan Klawer MANAGING DIRECTOR MIDDLE EAST AND AFRICA ARTEFACT DUBAI, UAE ~ OCTOBER 23 - 24, 2018 DIGIMARCONMIDDLEEAST.COM | #DigiMarConMiddleEast Data & AI Master Class MASTERCLASS
- 2. How to Successfully apply Data & AI in the Marketing Value Chain Masterclass 23 October 2018
- 3. Presenting today. Jan Klawer Managing Director Artefact Middle East and Africa @klawer
- 4. New consumer opinions - new industry standards World’s most innovative companies Ranking 2017 Innovation is driven by GAFAx who are already well-prepared for the future Top 5 Market Capitalization, PWC 2017 Media business is mainly owned by GAFAx Top 5 of Most Loved Global Brands 2017 Consumer loves GAFAx
- 5. Technology has transformed consumers’ lives A new breed of brands such as Google, Apple, Facebook, Amazon, Netflix, Airbnb etc have captured consumers hearts & minds by disrupting the status quo across all sectors. These brands have defined higher standards of customer experience by continuously improving their products and services through digital, data & AI. All brands must now live up to these new standards.
- 6. Artefact is a digital company celebrating the long overdue marriage of marketers and engineers. Artefact : Marketing Engineers
- 7. A squad operating on all marketing needs Technologies Our highly accredited experts have 20 years experience in leading award- winning global digital marketing campaigns. Business oriented, data raised, digital native, marketing minded and IT trained unicorns, they live to put together the perfect team of our in-house talents. Our Full-stack Engineers will set-up end to end DataLake, with the right trade-off between speed and reversibility AI specialists with PhDs in Machine Learning developing unique proprietary solutions to fit our clients’ needs and optimize in-house processes. Media & Activation Creation DataLake Ecosystem Artificial IntelligenceStrategic Consulting Data Science Award-winning creatives, able to design concepts from branding to performance based on data. Marketing Consulting First class data experts from the largest international pool of giant tech certified tracking experts, Analytics Specialists, Data Consultants and Data Scientists.
- 8. To achieve your Companies’ Vision all Activities in the Data and AI Pyramid are Enablers: how? Choose the right technologies and implement them successfully within expert and non expert marketing teams. VISION USE CASES Define the future state of the business, the role of data in the transformation to come and help the organization change in order to drive it. Improve data accessibility, quality and privacy by defining the right infrastructure, processes and governance, and by implementing them. Create and launch new use cases to generate short term or long term impacts on sales and costs, and industrialize them. Generate deep insights and KPIs on customers or marketing operations to better understand who the customers are and how the business is impacted by them. ANALYTICS DATA SCIENCE TECHNOLOGIE S DATA INFRASTRUCTURE AND DATA GOVERNANCE AND TOOLS
- 9. Do you know? ...the percentage of data that was created in the last 2 years?
- 10. of the world’s data was created in the last 2 years 90%
- 11. Do you know? What % of marketing organizations are implementing or expanding AI & machine learning in 2018? *According to Forrester research in 2017. Survey among 150 Martech professionals in the US.
- 12. Source: Forrester Research 2017 “Marketers have reached a point where their ability to capture data has exceeded their ability to take data-driven action” “They are confident that AI enhances human decisions and insights.” 84%
- 13. Do you know? ...the amount of customer care interactions that will be managed by AI in 2020?* *as predicted by Gartner
- 14. of customer care interactions will be managed by AI in 2020* 88% Source: Gartner
- 15. Data & AI The future is here 15
- 16. Vast amounts of data are being produced and processed every minute. 16 1.3K2.4M347K 142M Half a trillion digital moments across devices processed by Google Analytics each day 50K Tweets Searches Uber rides $203K Amazon sales App Downloads Emails 1 MINUTE
- 17. Accelerated growth of data production. First even known written document information/volume timetoday201020063000 BC 1450 1900 1992 1999 Invention of Printing London library (500 000 books) Internet advent Each person produces around 250 Gb 12 times the distance between planet earth and the sun if we stock all the data in book format We are here! 90% of the data available in the world has been generated between 2016 and 2018 Skyrocketing quantity of data: birth of data science
- 18. Accelerated growth of science & technology.
- 19. AI Computing power + machine learning + data skyrocket innovation in marketing. Data is the Fuel Machine Learning algorithms are the engines
- 20. Automated orchestration Automated media mix optimization Manual reporting Automated single channel reporting Customer journey reconciliation Attribution analysis Manual orchestration The big tech giants are leading the way in using data to drive full marketing automation.
- 21. What is AI? Weak AIStrong AI “The creation of computer programs (machines) that can (out)perform tasks that are currently completed more satisfactorily by humans”
- 22. A machine that can do everything a human can, including being self-aware What is strong AI? Timeline: 30 to 1000 years
- 23. 26/11/2016 “Intelligent” programs developed for specific applications that either do or don't reproduce human intelligence That's the focus today What is weak AI?
- 24. Todays examples of AI applications. Siri 2011 Now 2012 Cortana 2014 Alexa 2015 M 2016 Assistant 2016 Bixby 2017 Personal assistants Self-driving cars Augmenting humans Off-the-shelf AI for a wide range of industries: With Tesla taking the lead, they include all AI bricks ● Sensing: The camera turns video into objects and places ● Planning: I have to go from A to B via the best route ● Learning: I learn to manage any situation via millions of hours of driving ● Knowing & Reasoning: Moral driving rules Natural language processing and use of web resources Sentient Ascend: Tech to optimize the UX of e- commerce sites Khresterion: Real-time voice analysis tech Watson: Can help with medical diagnoses
- 25. Map of AI. Learning from a set of examples, growing over time ● Supervised ● Unsupervised ● Reinforcement Learns to recognize spam in an inbox AI Learning Sensing Processing signals from the outside world ● Computer Vision ● Speech recognition ● Natural Language Processing ● Feature construction ● … Planning Moving in an environment to attain a goal ● Monte-Carlo Tree Search ● Markov decision process* ● Genetic algorithms ● .... Knowing & Reasoning Deducing facts from other facts, rules, connections ● Rule-based engine ● Expert systems ● Ontologies ● …. Camera of a driverless car The robot must go charge itself then leave home I can prove a mathematical theorem from axioms * Stochastic model used to study optimization problems using dynamic programming or reinforcement learning algorithms
- 26. What is currently possible, and what isn’t (yet)? In the years ahead, specialists will keep striving to let AI perform tasks that the human brain currently processes in a second. Input Challenge Output What AI can do... What AI can't do... Are there any human faces (0/1)? Who’s on there? Will they pay back the loan (0/1)? What is that sound? Who’s talking? What’s the position of other vehicles and objects? Will it crash?
- 27. We’ve built AI solutions in order to inspire and free up time for customers and marketers, removing errors and repetitive tasks. Our AI roadmap Marketing as an API Automated reporting Siloed channels & devices customer vision Omni-channel & cross-device customer vision AI-based content production Media AI decision making process Automated orchestration
- 28. Where to apply Data & AI in digital marketing? 28
- 29. Your challenge: What tasks are appropriate to automate and optimize with AI? 522hours per year lost on repetitive tasks * *DJS Research 2017 DUMB WORK GUESS WORK Manual and time-consuming tasks can disappear thanks to AI AI can confirm or deny human intuition in the prediction of the optimal customer dialogue / journey 522 hours per year lost on repetitive tasks*
- 30. Thanks to AI, marketing departments can concentrate on higher value-added work: strategy and creativity
- 31. Tomorrow CMOs will formulate strategic and creative hypotheses and machines will do the rest of the job: it is our marketing as an API vision
- 32. Where to apply data & AI for marketing as an API? WHO - Audience segmentation Segment audiences based on: - Behavioral / relational likeness - Buying propensity - Reasons to buy - Churn propensity - Interest or channel affinity - Lead scores (engagement / conversion)
- 33. Activate message, content and offer based on: - Dynamic creative optimization - Dynamic content creation - Recommendation engine WHAT - Message / content / offer optimization Where to apply data & AI for marketing as an API?
- 34. WHEN/WHERE - Channel optimization Reallocate and activate budget based on: - Conversion attribution - Budget allocation and distribution - Reception recognition (offline impact) - Brand perception - Voice search / Personal assistants Where to apply data & AI for marketing as an API?
- 35. Time CustomerInteractionsandValue REACH ACT CONVERT ENGAGE Smart Content Curation Programmatic Media Bidding AI Generated Content Voice Search Propensity Modeling Ad Targeting Predictive Analytics Lead Scoring Re-targeting Dynamic Pricing Web & App Personalisation Chat Bots Predictive customer Service Marketing Automation 1:1 Dynamic content emails Who? Where? What? Demand generation and purchase intent 1st Purchase Indecisive Customer Loyal Customer Lapsed Customer Repeat Customer Applications of AI & machine learning in the customer journey.
- 36. Score AI Readiness explained Expert 31 > 40 Demonstrates excellent digital capabilities. Experts have already invested in AI and identified business cases for AI. Opportunist 21 > 30 Demonstrates good digital capabilities. Opportunist are already investing in AI, but are not taking full advantage of all possibilities. Novice 11 > 20 Demonstrates good basic digital capabilities. Perhaps already shown interest in AI and taken basic steps towards implementation. Laggard 0 > 10 Demonstrates relative poor digital capabilities. Laggards are probably in the early stages of gaining information and evaluating AI. Compare your score against our threshold and the benchmark based on competitor survey results, and see what you can do to adopt AI into your activities. 24 12.7 Threshold Current Survey Do the AI-readiness test! Your personal score AI: How ready are you? ?
- 37. Creation of new uses (Chatbots) Algorithmic Omnichannel Orchestration Sales Predictio n Dynamic Persona Creation The way we use AI in our day to day activities at Artefact. Predictive life moments detection Client segmentation Dynamic Creative Optimisation (DCO) Content Indexation Automated Mix Media Optimization Automated Trafficking Bid Optimization Automated Reporting Cross-channel Synergy detection ROPO Measurement INSIGHTS CREATIO N ACTIVATION MEASUREMENT
- 38. Data & AI in the Marketing Value Chain 38
- 39. The marketing value chain is composed of repetitive tasks that are prone to error. Plan. Setup. Control. Optimize. • Identify the best audiences to target • Find the best budget allocation • Ad creation • Ad trafficking • Consistent campaign tagging • Tracking consistency • Data verification • Data (PII) leakage • Error tracking (payment, 404) • Attribute value across channels & reallocate budgets • Optimize for the best audiences, creative, content & offer • Automated reporting
- 40. Common challenges in the planning phase. Plan. Setup. Control. Optimize. • Identifying the best target audience • Finding the best budget allocation and distribution
- 41. We have automated audience clustering combined with creative optimization. RESULTS Conversions ROI Reach x2 x2,3 x1,4 ACTIVATE THE BEST SEGMENTS WITH THE BEST MESSAGES
- 42. We seek for patterns amongst ‘convertors'.
- 43. How did we do it: Audience Builder
- 44. Common challenges in the setup phase. Plan. Setup. Control. Optimize. • Ad creation • Ad trafficking • Consistent campaign tagging
- 45. Tasks performed by an Ad Trafficker in DoubleClick. Time-consuming Manual Create a campaign Upload creativeCreate placements Create ads Verify impressionsDownload tags
- 46. Solution: Trafficking 2.0 A stand-alone interface
- 47. Media plan is ingested and processed by our algorithm into a traffic sheet RELIABLE Easily prevent human mistakes FASTER Automatically launch campaigns within minutes, in a few clicks Trafficking 2.0: 3 steps to reduce complexity Automate 98% of your campaigns. USER- FRIENDLY 1 easy-to-use interface Errors are automatically detected and detailed within the interface Detect errors that impact performance Turn media plans into campaigns 21 Automatic campaign setup: the algorithm creates ads and their backups. Traffic campaigns at scale 3
- 48. Common challenges in the Control phase. Plan. Setup. Control. Optimize. • Tracking consistency • Data verification • Data (PII) leakage • Error tracking (payment, 404)
- 49. PowerScan: validate quality of tag implementation through actionable reports. Automate tags verification at scale Detect wrongly implemented tags and variable labels and values Automatically scans thousand of URLs DETECT TAG ERRORS SCAN ALL URLS ON YOUR WEBSITE 21 Detailed reporting with actionable insights: - Errors: no call, wrong variable label, wrong value - Cause: technical or analytics - Evaluate and refine GET A FULL REPORTING AND TAKE REQUIRED ACTIONS 3
- 50. Common challenges in the Optimize phase. Plan. Setup. Control. Optimize. • Attribute value across channels & reallocate budgets • Optimize for the best Audiences, Creative, Content & Offer • Automated reporting
- 51. 1 Demand is sent to the agency 2 The agency writes a new scenario 3 Preview link is sent to the client 4 Validation of the scenario by the client 5 The new scenario is activated 8 emails 2 calls 3 mobilized resources 3 days of process & work BEFOREBEFORE The creative optimization process.
- 52. AFTER 1 person 1 hour A creative editor facilitates manual testing of different scenarios.
- 53. Intuition is being compared to historical performance and machine learning algorithm.
- 54. At Artefact, we have built a suite of AI-based products OUR PROPRIETARY TECHNOLOGY IS AT THE HEART OF OUR DAILY DECISIONS Audience Builder Media Plan Recommendation SEAds Generator URL Builder Trafficking 2.0 PowerScan PII Tracking URL Checker Octopus SEA Optimization SEA/SEO Synergies SEA exclusion/targeting Automated Reportings Artefact Ingestion ToolDPP Display Optimization Plan Setup Control Optimize Activation MyDCO Infrastructure Log File Scan Page Speed Check
- 55. Plan. Setup. Control. Optimize. • Identify the best audiences to target • Find the best budget allocation • Ad creation • Ad trafficking • Consistent campaign tagging • Tracking consistency • Data verification • Data (PII) leakage • Error tracking (payment, 404) • Attribute value across channels & reallocate budgets • Optimize for the best audiences, creative, content & offer • Automated reporting By automating tasks in all these phases...
- 56. ...we gain time to re-focus on Data-Driven Strategy & Creativity 56
- 57. Key takeaways. Start breaking down you every day activities into clusters of repetitive tasks that could potentially be automated. Determine where you could introduce some 'intelligence' to support decision making. Start deploying algorithms for clustering/predicting what could be a profitable audience, message/creative or media mix. Don't be intimidated by the hype around AI/machine learning.
- 58. Thanks for your attention