31 March 2011

Highlights from SD Forum's 'Analytics the Next Wave'

SD Forum Logo

By Mary Vincent - Follow on Twitter @MaryVincent
I've always been impressed with SD Forum's events
, and they recently held an Analytics: The Next Wave conference last week which focused on addressing the following questions and goals: 

- How do businesses transform the wow of analytics and big data into the bottom line? 
- The focus has shifted from using analytics for better data access and reporting to using insights collected from analytics to deliver better outcomes to the bottom line and drive high performance. 
- This conference will explore increasing the bottom line through new sources of data and uses, new deployment and data architecture models, and consumer targeting derived from big data intelligence.

Here are some of the highlights I thought I'd share from the event. I hope this helps you in your Analytics Design and Planning.
1. New Sources of Data and Usage Panel with Zia Yusuf, Streetline, Ralph Clark, ShotSpotter,  Ilen Zazueta-Hall, Enphase Energy, and Moderator: John Fraser, Accenture
a. Ralph Clark explained that Shotspotter provides data and insight to government officials to help them make decisions, including identifying gunshot locations. He states the 1st most important Use Case is to identify the exact source of a gunshot so the dispatcher can dispatch resources more effectively, instead of a caller to 911 not knowing exactly where the gunshot originated, and Officers then driving around 3 to 4 blocks to find the source of the gunshot. The 2nd Use Case is to use Heat Maps to determine most of the occurrences so action can be taken.
b. Zia Yusef of Streetline places sensors in parking meters, and iPhone and iPad apps help folks find parking spots. He says 30 percent of traffic is caused by people trying to find a parking spot, and he says his company is partnering with IBM.
c. Ilen Zazueta Hall of Enphase Energy is a solar energy company making micro-inverters that allow for troubleshooting and data visualization and brings the power of analytics to physical data. She gave a color map example which displays solar panel wattage production; the solar homeowner sees dark spot on the display and finds a frisbee on the panel.
d. John Fraser of Accenture states the goal of 'looking through a car's front window instead of rear view mirror' and using a systems and data centered approach.

2. Analytics - Fireside Chat with Tom Peck, Levi Strauss & Co and Sanjay Poonen, SAP.
a. Tom Peck from Levi Strauss started out by joking "we pay people to put holes in jeans...years
ago people used to throw them away".  Joking aside, Tom mentioned t
here is a Partnership with IT and Business and look at early warning indicators to change practices. They have embedded Social Commerce into their Facebook application. Tom used the expression that it's really a Science and Art and use Regression Testing. He does Customer Research Walks with customers.  Since Consumer sentiment changes in 12 to 16 months, they consistently use Voice of the Customer practices and Segment and Tier their customers. India, China, and Mexico are their biggest emerging markets. "Analytics is about how to influence investments."
b. Sanjay Poonen from SAP showed a balanced scorecard and mentioned: customers come first and they
invest in people: "when you invest in customers and people you have a good business" and used the slogan "we make customers successful."  He mentioned Retail and Banking are exploding and gave an example that Amazon creates loyalty by predicting the next product people will like in addition to "Don't look in rear view mirror; look through front windshield."

3. New Deployment and Architecture Models Panel withScott Burke, Yahoo, Anant Jhingran, IBM, Oliver Ratzesberger, eBay, and Moderator: Eileen Boerger, Agilis Solutions
a. Scott Burke says Yahoo uses data to drive consumer experience and advertising. Everyone gets a different home page and looks at how many times someone clicks on an item. Yahoo uses the data to target relevant display ads that drive revenue for advertisers and publishers. Regarding what has worked and what has changed, the early solutions were home-grown and file based. Yahoo always had an ethos of home-grown and then pioneered Hadoop: a grid based solution, where they have tons of petabytes on Hadoop. Yahoo is an Oracle shop, a large OLAP cube, and Hive and Yahoo created web apps. The trend is data management and providing value to all the players. Bidirectional feeds. Hadoop is not used for realtime.  We built real time loops ourselves..lots of optimization. .runs on serving layer don't have time to go into data center or batching.
b. Anant Jhingran from IBM says Analytics will be one of the top 3 growth areas (cloud and smarter
planet are the other two both of which use analytics). Software and hardware optimization is a big deal = Systems Optimization.
c. Oliver Ratzesberger mentioned they discuss hundreds of use cases and "try out things on you to maximize learnings"; AB testing and multivariate testing are used and they serve 150 billion search impressions. It starts with data centers and IT operations that look at user behavioral data, and he called it: "Analytics as a Service."
Oliver says that eBay has traditionally looked for partners. They use Hadoop and bring learning from the large data warehouse, and keep amount of data replication to a minimum. They're using Tableau as rapid visualization tool. "It's unstructured but it's really structured; need ability to combine structured and unstructured, and get data updated every 10 minutes. From an Analytics as a Service perspective, initially every group wanted their own data build, but it was too much work and they wanted to reduce replication. Data Mart users were saying new data didn't match old data and consistency was an issue. Physical data marts were too complex and we have to discourage them. How do we offer a Virtual Data Mart? The Business self-provisions own Data Mart and it's up in a few minutes.  When the Data Marts access Customer Table, everyone uses the same data and latency goes away. 'Virtualized Analytics as a Service'. Ebay Built realtime in process memory systems so within a few seconds we know if something is fradulent.
Q. Is there an Issue with many people accessing the same table.
A. Good
 question..how are we going to make that happen..on our main system hundreds of people using 92 percent busy. In order to do that you need superb workload management.  Systems run millions of requests every
day. Our systems are built to take the most ridiculous request that could crash the system; there are many first time requests. The Request goes into a rules engine and very tight workload management.  The System knows if the CFO needs data, more resources are given. Marketing gets a certain percent. Through workload management we maximize efficiency. Hadoop has massive benefits on data processing but doesn't handle workload management very well. The Goal is to design for the unknown..don't prevent future use cases..make it as flexible as possible.

4. Fireside Chat with Simon Khalaf, Flurry and Sharon Wienbar, Scale Venture Partners
- Flurry is many times the size of twitter.
- Amazon has people who bought this bought that. How do I segment my audience and how do I provide a better experience. Experience is different on phone and ipad. People take time on the iPad..people on the phone are interrupted.
- It's not about the download, it's the engagement.
- Retargeting is the most high growth thing on the web, i.e. abt electronics..1st party cookie given from my website..now gets ads from ABT electronics wherever she goes on the web.
rightmedia and doubleclick exchange can find her.
- Flurry allows publishers to define a segment and that's what we want back.
- Facebook is showing friends a person hasn't interacted with me for a while.
- How do brand advertisers ensure they get ROI.
- Nike launches campaign and they look at analytics, measure engagement by doing it automatically without violating privacy.
- Movie and Automotive ..Toyota going into gaming experiences by putting in engaging racing videos. Wait before ad shows.
- Flurry is free with analytics view. Need to put code in mobile app. Doesn't impact performance.

5. Analytics. The Investor Perspective Panel withAsheem Chandna, Greylock Partners, Vispi Daver, Sierra Ventures, Lars Leckie, Hummer Winblad Venture Partners, Ping Li, Accel Partners and Moderator: Harold Yu, Orrick
a. Ping from Accel...Trend to Real Time Predictive Analytics. Cloudera is used for large scale data management. Analytics is now part of the development process. Analytics is now part of the marketing team.
b. Lars from Hummer Winblad ... Reduced cost of storage is enabling the big data revolution. Invested in mobile..analytics is built into the business. Hope Cloudera and Karma become big companies.
c. Vispi from Sierra...  Analytics is part of Intellectual Property i.e. Zynga is looking for what moves make the most money. Analytics is a Revenue Generating Product.