Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Stephen Kenwright - HeroConf - Machine learning tech you could and should use tomorrow

441 views

Published on

Edit's Marketing Director, Stephen Kenwright, takes you through a list of machine learning tech you could and should use during his #HeroConf London talk on 22/10/18.

Published in: Marketing
  • Be the first to comment

Stephen Kenwright - HeroConf - Machine learning tech you could and should use tomorrow

  1. 1. Machine Learning Tech you could and should use tomorrow Stephen Kenwright Marketing Director
  2. 2. Sheffield Hallam University Confidential
  3. 3. Confidential I studied this guy
  4. 4. This guy was one of my lecturers Ray was using machine learning to identify authorship @stekenwright
  5. 5. Devonshire Manuscript. Confidential
  6. 6. Confidential One potential author.
  7. 7. ConfidentialConfidentialConfidential Text mining.
  8. 8. Luminance. Confidential
  9. 9. How we use text mining First-party data like internal search queries, help desk questions, chat logs, phone logs Social media data, reviews and online coverage/links Search queries Confidential@stekenwright
  10. 10. What can we do with this data? PredictVisualise trends Connect data to customer records Compile text data in one place Confidential@stekenwright
  11. 11. Possible use cases Which keywords generate the most profitable customers? Does survey data match up to sales? Which customer complaints lead to churn? Confidential@stekenwright
  12. 12. ConfidentialConfidentialConfidential Identify pain points through call log mining.
  13. 13. Data Robot.
  14. 14. ConfidentialConfidentialConfidential We’re getting ahead of ourselves.
  15. 15. A quick history Confidential Early 1900s 1970s 1990s Now Intuition Statistical programming languages Automated machine learning Manual analysis Visual statistical software Using experience and judgement to predict outcomes Writing code to construct statistical models The software knows how to analyze your data and does it for you Manual calculations to predict outcomes Drag and drop workflows with menu driven commands to set up and statistical analysis @stekenwright
  16. 16. Today Confidential@stekenwright
  17. 17. ConfidentialConfidentialConfidential By 2020, 85% of interactions will be handled without a human. Gartner
  18. 18. Confidential Layered approach to implementation 4. One-to-one communications 3. Life style centred 2. Customer life cycle centred 1. Basic Increasingsophistication (Data,Audienceinsight, Technology) Example types of data:Example communications: Customer recognition Personalisation to drive relevancy and CTA Segmentation centred on lifestyle approach Fully personalised Basic information at purchase Customer life cycle position, purchase history Demographic: Age, affluence, geo- location, motivation Full benefits of Single Customer View @stekenwright
  19. 19. Netflix In the wild: Netflix.
  20. 20. Confidential Or Amazon. @stekenwright
  21. 21. “Machine learning is a core, transformative way by which we’re re-thinking how we’re doing everything.”
  22. 22. Machine learning in the Google stack In-Market Audiences Smart Display Campaigns Smart Bidding Confidential@stekenwright
  23. 23. Limitations You can create more customised audiences manually Brand safety has made marketers want more control over creative Seasonal campaigns are still manual Confidential@stekenwright
  24. 24. Choosing tech, according to Forrester Confidential@stekenwright
  25. 25. Everyone’s machine learning is, err…learning Pre 2010 2015 2017+ Confidential@stekenwright
  26. 26. Where we are now.
  27. 27. Confidential
  28. 28. Analytics maturity Market sizing Market sensing NBA modelling Proposition development Price modellingCampaign evaluation Data visualisation A/B testing PREDICTIVE Profiling / segmentation Customer lifetime value PRESCRIPTIVE Web analytics Propensity modelling Machine learning PRE-EMPTIVE KPI reporting Research analysis Upsell modelling DIAGNOSTIC DESCRIPTIVE COMMERCIAL VALUE COMPLEXITY Attribution / marketing mix Confidential@stekenwright
  29. 29. Developing personas Demographics • The youngest group • x% are families • Long distance travel Behaviour • Summer time travel • Book 4-6 months in advance • Less likely to visit in 12 months Lifestyle • Into music • Follow current affairs • Shop online • Lower income % of Base – a% % Value – b% Spend per visit - £c HH 3 year value - £d @stekenwright
  30. 30. Marketing mix modelling saves £2.4m PPC spend. Confidential
  31. 31. ConfidentialConfidentialConfidential Use email open rates to inform ad copy.
  32. 32. Phrasee personalises email using machine learning. Confidential
  33. 33. Self-serve data cleanse – puradata.co.uk 3. Dedupe
  34. 34. Olduvai gorge, Tanzania Confidential
  35. 35. Reducing response times for humanitarian aid. Confidential
  36. 36. Mara Tanzania project.
  37. 37. Basic image… with highlights.
  38. 38. Edit Classifier. • EditBuildings_1406043538 • "custom_classes": 5, “ • "class": "HomeComplex", • "class": "HomeSmall", • "class": "TinComplex“, • "class": "TinLarge", • "class": "TinSmall", Uses reinforcement learning to train the Watson image recognition and classification algorithm Confidential@stekenwright
  39. 39. Confidential Watson.
  40. 40. ConfidentialConfidentialConfidential 74% of organisations implementing AI who believe it’s making them more creative. Capgemini 2017
  41. 41. ConfidentialConfidentialConfidential A cautionary tale.
  42. 42. Confidential The Great Hall was filled with incredible moaning chandeliers and a large librarian had decorated the sinks with books about masonry. Mountains of mice exploded. Several long pumpkins fell out of McGonagall. Dumbledore’s hair scooted next to Hermione as Dumbledore arrived at School. The pig of Hufflepuff pulsed like a large bullfrog. Dumbledore smiled at it, and placed his hand on its head: “You are Hagrid now.” http://botnik.org/content/harry-potter.html
  43. 43. Let’s get started.

×