Big Data – From Hype to Reality Dr. Richard Benjamins Group Director BI & Big Data Telefonica Telefonica © 2014 Overview • • • • Telefonica What & Why of Big Data Opportunities of Big Data Privacy challenge Example application: Smart Steps 2 What’s the big deal with Big Data? Big Deals Big Data McKinsey McKinsey Telefonica 3 Big Data is a hype Telefonica 4 But what is Big Data? Dave Feinleib, Forbes blog 1. 2. 3. 4. 5. Big Data is Only About Massive Data Volume Big Data Means Hadoop Big Data Means Unstructured Data Big Data is for Social Media Feeds and Sentiment Analysis NoSQL means No SQL Telefonica 5 Where does all the hype come from? Today, huge marketing budgets are being thrown at those two words, driven by new business… no wonder all the noise! 2004: Google publishes Map Reduce paper (link: here) Telefonica 2006: Yahoo’s Doug Cutting open sources Hadoop out of his older search engine project Nutch. (Link: here) 2011: McKinsey Global Institute publishes report on Big Data’s market potential for business, reaching out of the tech. world (link: here) 6 Where is Big Data coming from? Type of Big Data OTT/Telco Cost of data collection By product/ seeking Batch/realtime Differential? Social media OTT Low Active Both No Web logs Both Low Passive Both No Network data (telco) Telco High Passive Both Yes M2M (sensor) data Both High Active Both Might Open data OTT Low Both Batch No Transact. data Both Medium Passive Both No Telefonica 7 Several business opportunities with (big) data Different “business” models with different maturities and different risks Internal use Improve services Examples of external use Access to insights Advertising PI Economy Become a gatekeeper of personal data Data = improved business Data = better advertising Data = business Data = risk = business Leverage data to understand and improve business (x/up sell, churn) and products Leverage data for targeting users with relevant ads and higher CTR and conversion Insights that help improve businesses and governments Recognize that digital data is delicate (privacy) Turn that into an opportunity M2M Smart cities Telefonica 8 But Big Data is also good for society and environment H1N1 virus pandemic 2012 Earthquake in Mexico Telefonica used mobile data to measure the spread of a global epidemic (“swine flu”) in Mexico DF Dimensioning emergency services in advance for an optimal response to natural disaster situations After the magnitude 7.4 earthquake in Mexico DF, Telefonica researchers captured modile data records that once anonymized and aggregated allowed building visualizations of the density of calls in the differents part of the city, immediately depicting the areas most affected by the earthquake. With Big Data tools like this, it would be possible for authorities to better anticipate contingency plans, dimensioning emergency services and placing them in those points where there is evidence that will be mostly needed in case of catastrophic events. To understand more about human mobility and the spread of epidemics through society, Telefónica Digital’s research team used anonymised and aggregated mobile phone call records to measure numbers of people visiting locations such as airports or universities. The study found successful Mexican Government’s decision to shut down key infrastructures, reducing virus propagation by 10%. (Click images for more) Telefonica (Click images for more) 9 Privacy remains an issue Telefonica 10 There is increasing awareness of what customer data companies store Telefonica 11 The industry is learning by doing Telefonica 12 Are you aware where your data is going? Telefonica 13 To the US … Europe’s leading analytics companies call upon European Institutions to replace Google Analytics Telefonica 14 Smart steps, for retailers Big decisions….. ....made better 1st product – “Smart Steps” for Retailers: • Decide on store location • Understanding store performance vs footfall • Plan local marketing campaigns and track their impact • Optimise resource planning – staffing/open hours Telefonica 15 Retailers worry about … Retailers have questions... Where should I target loyalty or acquisition marketing campaigns? Where are my customers coming from? Where should I locate my new store? I need to manage my resources. When are my peak times? Could I be operationally more effective if I changed my opening times? I am a large supermarket owner and one of my competitors has opened up down the road. I need to identify our battleground. Where is my competitor strongest and weakest? Strategic Decisions Telefonica I know the activity that goes on inside my stores. But what % of my target market is walking past outside? What is the opportunity that I am missing? Performance Management 16 Case study with 4th largest UK food retailer 400 stores nation wide Crawford Davidson: Customer Director at Morrisons Supermarkets: “Unlike some of our competitors, we don’t have a store card to tell us who our customers are, and how they shop our stores, which means we’re at a disadvantage in targeted marketing. Over-rewarding one loyal customer disadvantages us in investing in the next” “This increase in customers was achieved without any reduction in customer spend, “Smart Steps identified many more and with an improved new customer suitable target post code sectors, activation rate. Overall there was a 150% enabling us to send promotional increase in the amount of new or coupons to double the number of reactivated customers who visited households” Morrisons stores. This is a fantastic result.” Telefonica 17 NETWORK DATA The o2 mobile network has hundreds of cells to measure the trends in footfall across the country 2G Network 3G Network 900 MHz 1800 MHz 39 % 2100 MHz 2013 4G Network Telefonica 18 PRIVACY A 3 step process ANONYMISATION Before Telefonica Dynamic Insights (TDI) receives the data, all personal information is removed. The data TDI receives are cryptographically hashed values AGGREGATION The hashed values are aggregated into groups, i.e. gender & age band. At this stage there are only crowds of o2 customers EXTRAPOLATION We take our sample and extrapolate to population totals, using mathematical algorithms. This gives us the grouped values Smart Steps uses. Telefonica 19 200 x 200 GRID Footfall is rendered into 200 x 200 metre grid squares across the country Easier to use Further protecting anonymity Extrapolated to represent local population 39 % Telefonica 20 Example question of a marketer COUNT How does the footfall in our Whatchange are thethroughout profiles of the area the people in theday? area of my store? Telefonica 21 Differential aspect Vast sample base based on observed crowd behaviour Intuitive web tool covering the whole of the UK to draw insights from Today’s data tomorrow. Fastest data delivery in the market Eliminates retailers’ blind spots. The profile of the footfall in their area Insights 24/7/365. Data every hour, day, week and month. You choose. Export data and combine with other sources Telefonica 22 And what about the Semantic Web and Data? Telefonica 24 Semantic web and data trends Telefonica 25 Semantic Web and Gartner’s Hype Cycles Telefonica 26 2006 – 5 to 10 years for reaching mainstream Telefonica 27 2009 – more than 10 years to go Telefonica 28 2012 – more than 10 years to go Telefonica 29