![Approach: Multi-lingual, rule-based, deep natural language
processing
All our analysis capabilities are provided for multiple languages and multiple data
sources
– There‘s value beyond English tweets
SMA uses the BigInsights text analysis engine for base text processing, and runs
SMA‘s text analysis rule resources in this engine for „our“ capabilities
My Nikon D5000 doesn‘t work well in fuzzy light conditions. Text
[My] [Nikon] [D5000] [does] [n‘t] [work] [well] [in] [fuzzy] [light] [conditions] Tokens
Part-of-SpeechPP NP NP VAUX NEG VB ADV PREP ADJ NN NN
[My Nikon D5000] [doesn‘t work well] [in fuzzy light conditions]. Syntax Analysis
[My Nikon D5000] doesn‘t work well in fuzzy light conditions. Behavior
[My Nikon D5000 doesn‘t work well] in fuzzy light conditions. Sentiment
BigInsights
SMA](https://image.slidesharecdn.com/1fe98e55-2fcd-4746-ab0c-5efb3648d8ce-160306103722/85/Bigdata-and-Social-Media-Analytics-28-320.jpg)
Bigdata and Social Media Analytics
- 1. IBM Social Media Analytics 1.2 A Big Data Application Presenter Dillip Kumar Majhi –Bigdata And Social Analytics Solution Architect
- 2. Agenda The scope of social media analytics Demo IBM Social Media Analytics: Architecture and Framework Filtering the noise: configuring relevant social media analytics A deeper look at some of SMA‘s built-in analytics capabilities 2 A deeper look at some of SMA‘s built-in analytics capabilities No silos: Integrating Social into a 360 degree customer view Our successful journey to „Application as a Service“
- 3. “Social Business” – more than just a facebook page Uses for social business1 Respond to customer questions Capture customer data 78% 60% 79% 55% 3 33 3 Source: Institute for Business Value; Business of Social Business Study, 1 Based on responses from individuals having personal experience with customer-related social business activities (n = 599) Solicit customer reviews and opinions customer data Identify and manage key influencers Today Next two years 79% 71% 44% 69% 37% Cognos Consumer Insight 1.1 IBM Social Media Analytics 1.2 Cognos Consumer Insight 1.1 IBM Social Media Analytics 1.2
- 4. Organizations are moving beyond promotion towards driving sales and support Promote events/ marketing campaigns Generate sales leads and revenue 83% 71% 74% 51% 4 4 Provide product and services support Sell products directly to customers Today Next two years 69% 46% 61% 35% Source: Institute for Business Value; Business of Social Business Study, 1 Based on responses from individuals having personal experience with customer-related social business activities (n = 599) Cognos Consumer Insight 1.1 IBM Social Media Analytics 1.2
- 5. IBM Social Media Analytics (SMA) targets many Business Issues - driving actions, not just listening What are your How can I identify sales leads ? Marketing SalesAre my suppliers getting a bad reputation? Supply ChainSupply Chain 5 What are your questions? How do consumers feel about our new messages / ads? What are emerging customer questions to prime our call center agents with? Customer Service 5 What features are our customers asking for? Product Development
- 6. A Social Media Analytics application requires more than Analytics Deliver results to more than one employee Provide meaningful Social Media Metrics Take IT out of the picture for daily operation 6 Text Analytics - Detect Brands, Products - Sentiment - Author Geo, Demographics, Behavior Collection Analysis - Detect emerging topics in Social Media Statistics - Automatically discover relationships& affinities Influencer Identification
- 7. Agenda The scope of social media analytics Demo IBM Social Media Analytics: Architecture and Framework Filtering the noise: configuring relevant social media analytics A deeper look at some of SMA‘s built-in analytics capabilities 7 A deeper look at some of SMA‘s built-in analytics capabilities No silos: Integrating Social into a 360 degree customer view Our successful journey to „Application as a Service“
- 8. Demo Scenario: Addressing new customer segments for a grocery store Business problem: Imagine you‘re Wegmans... ...and you want to enhance your market positioning around „healthy foods“ Approach: Analyze competitors, as well as some key aspects of shopping experience in 8 Approach: Analyze competitors, as well as some key aspects of shopping experience in social media
- 9. Start: SMA Welcome Page All your analysis projects in one place 9 Save and share configurations across projectsLaunch the relevant UI
- 10. SMA Reporting Application Dynamically select analysis criteria 1 Guided Navigation according to Social Media Framework Individual KPIs
- 11. Share of Voice – Check the overall „Brand Health“ I (Wegmans) „lag“ behind my peers in this space 1 Less positive sentiment, compared to peers Certainly room for improvement
- 12. Reach – what are the Top 10 sites where people talk? Twitter? No surprise – What‘s tripadvisor doing here – and why am I missing? 1 Twitter? No surprise – let‘s focus on other sources....
- 13. Proximity to a „Healthy Food“ store is seen as a plus: The underlying data is always only one click away 1 Start co-marketing activities with certain hotels to pull „healthy food“ shoppers to nearby Wegmans locationsAction
- 14. Sanity check – are there certain aspects of the shopping experience we should improve? Narrow down to a certain aspect Positive/Negative 1 Positive/Negative sentiment ratio significantly worse Product recall Store Layout
- 15. Shaping my message: who‘s talking – and about what ? 1 More females talking around „healthy food“ stores, but sizeable number of men as well Overall themes the same, but with different priority
- 16. Understanding what aspects actual shoppers talk about Trader Joe‘s shoppers significantly different See „actual shoppers“ for further engagement 1
- 17. Evolving Topics – identify & react to the unexpected 1 Topics in social media over time, automatically identified by SMA Action Check supply chain to pro-actively avoid this problem
- 18. Agenda The scope of social media analytics Demo IBM Social Media Analytics: Architecture and Framework Filtering the noise: configuring relevant social media analytics A deeper look at some of SMA‘s built-in analytics capabilities 1 A deeper look at some of SMA‘s built-in analytics capabilities No silos: Integrating Social into a 360 degree customer view Our successful journey to „Application as a Service“
- 19. ReportingAdministration Analysis Welcome Page Project Project Project Project Security Multi- tenancy SSO 1 19 Analytics Platform Message boards Blogs News Cognos 10.1.1Text Analytics IBM BigInsights 1.3 Search Index RDBMS g DB Twitter Evolving Topics Facebook Reviews Video Project Project Security E2E Orchestration Influencer Identification
- 20. SMA Analytics Pipeline – deeper dive Fetched Data (Json) Document Analysis Identify Concept mentions Detect Sentiment Extract Author Location, Demographics, Behavior Evolving Topics Topic detection Assignment of Mentions to Topics Analysis Results (CSV + Json) BigInsights Orchestrator Job definition SMA Flowmanager Job definition 2 Document Analysis & Evolving Topics are running in BigInsights Document Analysis: Jaql scripts for document preprocessing + BigInsights text analysis rule engine (actual text analysis rules are part of SMA) Evolving Topics: Jaql scripts for document pre/postprocessing + Java algorithm for topic detection Processing phases triggered through a 2-level workflow: BigInsights Orchestrator for steps within Hadoop, SMA „FlowManager“ for E2E orchestration (e.g., including „data in + data out“ steps) SMA Flowmanager Job definition
- 21. Whatever we do – our users must not need to care 2 Hadoop, DB2, Cognos,... totally transparent to analysts - they just work with SMA Administration
- 22. Results delivered at the right time CCI 1.1 provided the capability to update social media content once per day Some use cases call for a higher update frequency: – Identifying sales leads – Monitoring of an ongoing campaign or event – Identifying 1:1 customer engagement opportunities SMA 1.2 provides additional update frequencies – every 20 minutes – every 2 hours 2 – every 2 hours All trend reports support a breakdown on „hour“ level Create Cognos alerts for „author with Klout score > 80 expresses negative sentiment about us“
- 23. Agenda The scope of social media analytics Demo IBM Social Media Analytics: Architecture and Framework Filtering the noise: configuring relevant social media analytics A deeper look at some of SMA‘s built-in analytics capabilities 2 A deeper look at some of SMA‘s built-in analytics capabilities No silos: Integrating Social into a 360 degree customer view Our successful journey to „Application as a Service“
- 24. The challenge in Social Media.... OBI in the German DIY world OBI in the rest of the world 2 OBI in the German DIY world OBI in the rest of the world Persil for Germans Persil for the French
- 25. All Social Media Data fetcher Keyword QueriesConfigure 1 Our approach: a multi-level „relevance funnel“ Relevant Blog , Board , Video 2 Analysis Flexible Type/Concept hierarchy Text Analysis Patterns Analyze 2 Board , Video entries… Product X Product y Product Z
- 26. Configuration Experience Word lists and regular expressions define a particular concept 2 „Types“ and „Concepts“ define the scope of the analysis – anything from brands to products to services to events to customer touchpoints to... Include terms „make up“ the concept Exclude terms „rule out “ irrelevant meanings Context terms describe relevant contexts
- 27. Agenda The scope of social media analytics Demo IBM Social Media Analytics: Architecture and Framework Filtering the noise: configuring relevant social media analytics A deeper look at some of SMA‘s built-in analytics capabilities 2 A deeper look at some of SMA‘s built-in analytics capabilities No silos: Integrating Social into a 360 degree customer view Our successful journey to „Application as a Service“
- 28. Approach: Multi-lingual, rule-based, deep natural language processing All our analysis capabilities are provided for multiple languages and multiple data sources – There‘s value beyond English tweets SMA uses the BigInsights text analysis engine for base text processing, and runs SMA‘s text analysis rule resources in this engine for „our“ capabilities My Nikon D5000 doesn‘t work well in fuzzy light conditions. Text [My] [Nikon] [D5000] [does] [n‘t] [work] [well] [in] [fuzzy] [light] [conditions] Tokens Part-of-SpeechPP NP NP VAUX NEG VB ADV PREP ADJ NN NN [My Nikon D5000] [doesn‘t work well] [in fuzzy light conditions]. Syntax Analysis [My Nikon D5000] doesn‘t work well in fuzzy light conditions. Behavior [My Nikon D5000 doesn‘t work well] in fuzzy light conditions. Sentiment BigInsights SMA
- 29. Identifying author behavior in English, German, Spanish, French Users of a certain product or service – What product features are relevant for them? Recommenders – E.g., authors mentioning „you should use X“ Detractors – e.g. authors mentioning „stay away from X“ 2 e.g. authors mentioning „stay away from X“ Prospective users – Potential sales leads for 1:1 engagement – Identify sites where prospective users congregate
- 30. SMA as a service Hosted on SmartCloud Enterprise+ Every customer has their own SMA deployment SLA: maximum of 2 days between „custom places order“ and „customer receives URL to their SMA welcome page“ Pricing according to the number of social media documents a customer 3 Pricing according to the number of social media documents a customer retrieves per month Besides the user interfaces, customers can also access the SMA Social Warehouse to integrate SMA data into their local business processes – Through ODBC over SSL
- 31. Summary IBM Social Media Analytics targets multiple use cases in marketing, sales, PR, customer service and product development – Our Social Media Framework helps to map the right KPIs to a particular use case IBM Social Media Analytics is an integrated analytic application, harnessing „best of breed“ IBM products – Using the IBM Big Data platform for content processing – Easy access to insights through Social Media Warehouse and Cognos BI – Optimized data access for predictive analytics 3 – Optimized data access for predictive analytics – Sentiment detection in English, French, German, Spanish, Dutch and Chinese (Simplified and Traditional) – Extracting Geographic, Demographic and Behavioral data for Social Media Authors in English, French, German and Spanish „Becoming SaaS“ did not only change the way we do installation: it changed our mindset on things like „performance“, „ease of maintenance“, „migration to new versions“,... (and will continue to do so)
- 32. To find out more... External: http://www-01.ibm.com/software/analytics/solutions/customer-analytics/social-media-analytics/ Product videos (tech, not marketing) – also external http://ibmtvdemo.edgesuite.net/software/analytics/cognos/videos/HTVs/sma-1-2/index.html „Sales Kit“ with Demos, Resources,... 3 „Sales Kit“ with Demos, Resources,... https://w3-03.sso.ibm.com/software/xl/portal/content?synKey=Q420240R36382M68#overview Integrating SMA data with BI and Predictive Analytics http://www-01.ibm.com/support/docview.wss?uid=swg27038638
- 33. 3 33