Live from the Workshop on Social and Business Analytics

“We have to unite the art and science of marketing. The MadMen and the MathMen.”

This post was created as a live blog from the Workshop on Social and Business Analytics (#McCombsWSBA) at McCombs School of Business on March 28, 2014. I’m not a data scientist, and much of the material presented was way over my head, so I’ll apologize both to the readers and the presenters. I know I got some of this wrong, and the rest is stated in an overly simplistic way.

Perhaps I’ll pull it down later, but for now I’ll leave it up as a resource for others who attended and want to recheck their own notes.

Welcome & Logistics: Rajiv Garg, University of Texas

Rajiv Garg is a new faculty member at McCombs, in the department of Information, Risk, and Operations, and he has played a primary role as an organizer and promoter of this event. For the first workshop of this kind at the university, that I’m aware of, he has managed to assemble a stellar list of guests and attendees.

Welcome Remarks: Janet Dukerich, Sr. Vice Provost, University of Texas

Purpose of this session is to bring industry and academics together. We can produce massive amounts of data, but how we manage to make sense of this data is a critical area to explore. There are enormous policy considerations in the management of these data. 

Session 1: John S. Butler, Chair

Shyam Venugopal, Frito-Lay/Pepsico “Analytics as an Organizational Change Force”

Frito-Lay is no longer on the sidelines on how to use business analytics as a decision tool. 

You don’t have to understand every complexity of analytics. Figure out how to organize the data and the questions. The “sexy tools” aren’t going to get you anywhere. 

When the CFO leaves a company, who do key metrics and processes stay in place, but when a CMO leaves everything changes? The reason is that marketing lacks metrics and processes that reliably tie our activities to financial return in a predictable manner.

This is easier said than done, because the world is very complex. It can be quite overwhelming to get to our consumers. We have to unite the art and science of marketing. The MadMen and the MathMen.

Frito-Lay centralized our marketing decision making for all of our brands, so that what works for the individual brand must also work for the entire portfolio.  This was a difficult transition. “Toto, I don’t think we are in Kansas anymore.”

Great marketing will drive both short term sales and long term preference. 

We know that we have to assess the marketing impact at the most granular level possible. This led us to a culture of analytical rigor. On an average, this has been much better than gut-based decisions. 

Secondly, we moved from an historical assessment mode (score card) to a predictive, forward looking mode, with revenue targets and a focus on continuous optimization.

We now keep sales and marketing driven off a single analytical platform. We have a single set of performance goals, so if one area isn’t performing as well, then we can move assets to another channel.

Regarding social media, we track the metrics but we don’t track success or failure based on these metrics. Instead, we focus on Return on Investment of each program. How does it matter whether my Facebook posts go up? There were no clear answers, so for the last 15 months we have collected the metrics, and then we are trying to link them to program performance.

We use the same Analytical Platform for:

  • Scenario Planning
  • Media Planning and Flighting
  • Annual Brand Planning
  • A&M Investment Prioritization
  • Communication/Creative Guidelines
  • Plan Audits/Course Correction

The single platform can’t answer all of the questions, and we can’t answer all our questions overnight. Take baby steps, because when mistakes are made that are based on analytics that often means the last opportunity to use analytics. No model is perfect, but most are useful, so we stay with a model because over time we can figure out how to better use the data.

We no longer think of our product categories, we think in terms of demand categories, the situations in which our customers have needs that can be met with our products.

“If it doesn’t sell, it is not creative.” David Ogilvy

 Anindya Ghose, New York University “Big Data, Randomized Field Experiments and Mobile Marketing Analytics”

How crowded will be your drive to the beach three weeks from now? We data jockeys know that. No matter how spontaneous we think we are, we are predictable in our movements. 

I love to immerse myself in massive floods of data. Mobile has helped create an explosion of big data. Mobile commerce, app purchases, mobile content, mobile ads, and the interdependency between channels and devices all create data.

Six Projects We Have Done

1. How Travel Patterns Influence Mobile Internet Usage. By mining your contextual data captured from your smart phone usage companies can predict a user’s travel path and send more targeted ads.

Is there a connection between the travel patterns of consumers (day-to-day) and their interaction with advertisers. We found that the days when you deviate from your standard travel patterns you are more likely to respond to advertising messages. These are the consumers we want to target, they are more valuable to an advertiser. 

2. Effect of Location on Consumer Behavior

3. Disentangling Coupon Value vs. Distance vs. Rank. When consumers are closer to the store, it takes less of a discount to lure them into the store. 

4. Mobile Message Targeting. The combination of location and time. Both are important, but the effect varies. So for example, when you are close to the movie theater, a same-day coupon is more effective, but when you are far away, a next-day coupon is more effective.

5. How Mobile Devices are Being using for M-Commerce. How does the nature of the product change the type of mobile device used to engage with it. For complex products, people prefer laptops, but if it is easier, then tablets take over. For simple tasks, mobile phones increase in utility. 

6. Impact of Cross-Media Ads. How do cross-media exposures affect changes in click-throughs and conversations? Should I have both ads on smart phone, one on smart phone and one on tablet, etc. Turns out there is a stronger impact when ads are run across different channels, keeping the number of exposures the same. 

We live in a mult-screen world, and consumers take a multi-device path to a purchase. 

A new “science of cities” is emerging from mobile phone data analyses. What can we learn about cities from mobile phone data? You can actually learn a lot. When you study how people use mobile phones you can begin to see how a city “breathes.” 

Blake Chandlee, Facebook

We look at billions of data points every day. Data is used to make every single decision we make, but we’re also about people. We’re trying to reach people. “All the people who matter to you.”

Each of us is different, there are billions of people but every one is different and we use data to operationalize that. Large audiences are not enough, reach is not the most important thing. Data enables reach and precision. 

Seven years ago we didn’t know a lot about our consumer base. Today we have enormous audiences over 1.23B users, reaching 1B on mobile. More than half our revenue comes from mobile. 62% of monthly active users return daily. 

Facebook controls more time of people on mobile than all the other channels combined. 

We want to reach real people, not proxies, with greater accuracy. The average Facebook reach for narrowly targeted campaigns is 89% accurate. Deep knowledge at the granular level is very effective. It allows us to directly target messages to specific interests and needs of a consumer.

Our approach to consumer privacy. Mark Z. lives and breathes data privacy. The moment that is compromised we lose our business. We work with governments and agencies, because people think we sell the data. But we don’t, I want to make that clear.

We don’t think clicks are relevant, we want to tie into real impact, business outcomes, and tie that to whatever behavior drives that. With partnership with Datalogix we can now match consumer behavior to their Facebook activity. We move from being a media platform to being a marketing partner.

We think personalized marketing at scale is a reality today. And we know we have to make that easy, so a 24-year-old media planner can put it together at 4 o’clock on a Friday afternoon when he wants to go get a beer with friends.

Session 2: Rajiv Garg, Chair 

Michael Svatek, Together Mobile “Who’s behind the hashtag? Harnessing everyday micro-moments to build deeper relationships with customers”

My focus is on how to connect directly with consumers. 

Every moment a consumer has is an experience. Their time is broken up into micro-moments. I added up every moment of 5 minutes or less when I didn’t have something I needed to do, and I came up with 81 minutes. Doesn’t include time spent responding to email.

Michael showed a sample campaign for Aveda haircare products, based on consumers sending in “selfie” photos and picking a desired hair style. 86% of consumers answer five questions to participate in such a campaign. We average seven engagement votes on these mobile campaigns. 

Event activation. We launched a SxSelfie at SXSW. Post a selfie to win a GoPro camera, ended up with a variety of photos from hundreds, with many celebrities. Getting people engaged with the brand digitally as well as physically during an event. 

Michael is breezing through a fast-paced overview of campaigns using consumer mobile phone data to strengthen brand affinity, create brand awareness and target consumer messages for specific behaviors.

John S. Butler, University of Texas “Managerial Science, Business Analytics and Competitiveness: The Impact of Social Physics”

Managers need to get outside their own networks, or their companies will die, as did Kodak. We must use data to get outside our networks and understand the world more completely.

Natanya Anderson, Whole Foods “Driving Business Strategies with Small Data”

We’ve been promised rainbows and unicorns with “Big Data,” but how do you actually get there? It’s like eating an elephant, one bite at a time. 

You have to teach your organization to appreciate this whole new world of data. Just because they trust the data they have doesn’t mean they will trust new data that is brought to them. One of my jobs is to teach my organization to appreciate all the new data that is coming to us. 

We identified “high value opportunities.” 

  • Actionable Insights–what can I actually impact?
  • Key Business Strategies–what fits what my organization is trying to accomplish?
  • Ability to Affect Change–how able am I  to affect change, where can I build allies?

It’s important for us as practitioners to be “story tellers” that bring actionable insights. And we have to think about how to share these stories in a way that is most meaningful to them. At one point I was thinking about a dashboard, and one of my staff said, “Make it look like Flipboard,” because that is how we like to consume data.

I got to the point that I had too much data, and no time to look at it. You have to have analysts and teams that have time to study the data. They have to know what questions to ask. 

Stay Rooted in the Known–whenever you are introducing a new way to think about something (such as data), it helps to tie it to ways of thinking that are understood by the organization.

  • Accepted measures
  • Business processes
  • Segmentation models
  • Existing frameworks
  • Key priorities
  • Expressed desired
  • Cultural dynamics
  • Trusted storytellers

Samples of “Small Data” Projects at Whole Foods

Top Performing Facebook Posts–we looked at brand building through story telling. The opportunity was to focus local store social posts on customer-centric, engaging content. We started to highlight stores that were doing some of the best work, because our people like to be competitive. We put it in an infographic so that store employees could relate to it. Instantly behavior began to change, and they started to pay more attention to the type of social media posts they were creating.

Infographics Gone Wild–we looked at ways to introduce more visual styles on Pinterest to identify the best performing approach. Our art directors can actually use this information to create a cohesive visual voice that conforms to what our customers are posting about us. Store managers began to ask how they can use visuals that appeal to their local customers.

Why Aren’t They Talking About…–we identified customer conversations, to see how they actually talk about their purchases at Whole Foods. This is impacting how we are planning their campaigns. What is fascinating is what they don’t talk about–organics. We’re the number one organics grocery story, and our customers don’t talk about it. Do they just assume it as something that exists in their lives? We don’t know, but that knowledge has started a whole new initiative. 

Session 2: Maytal Saar-Tsechansky, Chair

Sirkka Jarvenpaa, University of Texas

Lisen Selander, Chalmers University “Digital Activism at Amnesty International”

Alessandro Acquisiti, CMU “An Experiment in Hiring Discrimination Via Online Social Networks”

Alessandro described the specifics of setting up an experiment in which it was studied how social media profiles (fake) impacted the hiring practices of employers. They were careful not to interject other factors, such as a distinctive name, which might have a negative impact on hiring, so that they could reduce the effect to what was seen online.

4,152 employers were included in the study, crossing several job types. We  found no significant difference between gay and straight candidates. But there was a big difference between Muslim and Christian candidates in “Red States,” with a negative result for Muslim candidates who got fewer call backs on job applications. 

The experiment provided “some” evidence of discriminatory biases along the traits we studied. 

Prasanna Tambe, New York University “Private Equity, Technological Investment, and Labor Outcomes”

LUNCH 

Panel: Anitesh Barua, Moderator “Past, present, and future of challenges and solutions of dealing with networks and data”

Hugh Forrest, SXSW

I will probably offer the lowest of low tech solutions. We solicit panel ideas via an online interface, panel picker. Over 800 panels, these are selected via an online vote process. It leverages the power of the community, it creates buzz during the summer, and it serves as a barometer of the industry. It is not sophisticated, but it is very helpful to our process.

When the economy dipped in the mid-2000s, we saw an uptick in the number of panels suggested on entrepreneurship. Without that input we might have thought that we should have more panels on how to find a job.

Tonya McKinney, Tata Consultancy Services

About 80 percent of our work is about customer experience. We look at the business side, and look at the connection between the input and the outcome. We get the IT, analytics, marketing, strategy all in the same room, and executives really like that approach. One of the big challenges is changing the operations of the company to match the speed of the data changes. 

The issue of “noise” in the data is really just about asking the question in the right way. We can look at the top ten sales deals and look at the interactions that happened with each of them, and then can say, these are the people and processes that are involved when we make a sale. It’s not just about getting so many Facebook likes. 

Brandon Burris, SnapTrends

We are now looking now at public safety and location issues. Trying to create a tool that would allow us to look inside crisis areas using social media.

The better the tools, the algorithms that come out of academia, then the better are the solutions we can offer to our customers. 

Location becomes an interesting way to get rid of the “noise” in the data. It allows me to look at a smaller segment. 

Jerry Kane, Boston College

We are finding that real value is delivered to companies as their social media practices mature. They start measuring with anecdotes and stories, but move to more analytical measures, but the stories remain important.

Companies are lagging the general public in their use of social media. The big issue is “how do you do business differently?” Unless businesses make this transition they’re not going to get much out of it. The other issue is that as companies adopt social media they have more challenges to face, such as global accessibility. 

Marketing is the low-hanging fruit for data analytics, but there is much more we can do with social media and analytics to transform busineses.

Ramesh Rajagopalan, Dell

A common thread is a theme around customers. There is only one valid business purpose, and that is to create customers. That is critical to our survival. The rules are changing and the customers are changing the rules. 90% of B-to-B customers are doing their search online. Mobile first when finding information online. Customers are 70% through their purchasing process before they even contact a vendor. That means our sales executives are not even in the conversation 70% of the process. We need to engage ourselves in that journey and find a way to deliver the right message, to the right person, about the right solution, at the right time and the right place.

We listen to our customers in 10 different languages, 24/7, but how can we bring in that data so that we can listen to it? 

We don’t have clear solutions to all of this, but it is a process of listening and learning. 

Paul Pavlou, Temple University

We are looking at data analytics from a global perspective, not just a business issue, or a medical school issue. How can we solve the big problems, genome data, and other ways to solve big problems. I understand the customers, but I like to look at this from a bigger perspective. 

Session 3: Garrett Sonnier, Chair

Claudia Perlich, Dstillery “Tales from the Data Trenches of Display Advertising”

[Perlich began by explaining how cookies and pixels work in tracking consumer behavior relative to advertisers.]

The primary issue is who to target. Second question is where should we advertise at what price. Third question is whether it matters, does an ad have a causal effect? And how do we know which ad gets the credit. 

Browser data is agnostic, “I do not want to understand who you are.” 

Does it matter whether I show you our ad on The New York Times or on Kayak? 

Ashish Agarwal, The University of Texas at Austin

Data from Field Experiments on Facebook regarding social and regular ads

Ran ad campaigns for two advertisers. Each with multiple posts, targeted to different keywords, by gender and age group. If users see ads in the regular news feed, they are less likely to click on social ads. Perhaps it is because when they see a regular ad and then see an ad in the regular news feed there is a sense that it is too impersonal. 

The more likes an ad gets, eventually the performance declines over time. 

Conclusions:

  • Efficacy of social advertising is influenced by the ad placement.
  • Higher strength of social signal leads to lower CTR performance

Mike Bailey, Facebook “How Effective is Targeted Advertising?”

Mike spoke on a series of studies that were done using Yahoo, not Facebook.

Determining ad effectiveness is incredibly difficult. There is a huge selection bias–are our ads only reaching people who were going to buy from us anyway?

The advertising effects are often really small, with a lot of variance in the sales decision, so it is difficult to measure the effects.

It’s incredibly expensive to drive clicks with ads. Advertising is not a good driver of search. 

Research study (Yahoo) looked at factors that impacted online ad response:

  • Location–consumers living within one mile of the story responded at a higher rate to online ads.
  • Recency–recent shoppers also saw a boost
  • Big Spenders–big spenders saw a boost
  • Income–income level saw a similar effect

From that it appears we should focus on your best customers, those most likely to respond. However another study showed that you should focus on those less likely to buy, because ads had little impact on the closer, more recent buyers; they are going to buy from you anyway.

So we are left with a big question mark. Measuring ad effectiveness is difficult, and often requires a LOT of spending and a controlled experiment. Future research should focus more on the hetergeneous response and less on overall effectivenss. 

Session 3: Sirkka Jarvenpaa, Chair

Rahul Telang, Carnegie Mellon University “Contagious Churn in Mobile Networks”

Scott Shriver, Columbia “Identifying Causal Effects in Social Networks”

Juanjuan Zhang, MIT “Tweets and Sales”

Session 4: Frenkel TerHofstede, Chair

Sharad Goel, Microsoft “Predicting Individual Behavior with Social Networks”

Rajiv Garg, University of Texas “Social Data Analytics: Measurements and Methods”

Nodes, connections and content. Who is who, who is connected to who, and who talks about what? We want to understand how, when, and why information is pushed or pulled on a social platform. 

Push vs. Pull

We want to either increase the push by increasing source volume or frequency, or increasing the pull by increasing follower volume or content value.

Does having more connections improve my job search? We did a study and found that LinkedIn is the most effective platform for job search (lowest cost and most productive).

Weak ties on LinkedIn lead to more job leads, but strong ties are the ones that actually convert leads into interviews and offers. Someone with many weak ties appear to be so well networked that they don’t need my help. 

How do we distinguish between strong and weak ties in a network? 

Peers on online social networks do play a role in diffusing new information to their network. Pull has a larger impact compared to push. 

Joe Rohrlich, Bazaarvoice “The power of social data: Relevant experiences, stronger relationships”

Session 4: Ashish Agarwal, Chair

Beibei Li, Carnegie Mellon University “Surviving Social Media Overload: Predicting Consumer Footprints on Product Search Engines”

Neel Sundareshan, eBay “From Lemons to Peaches” Reputation Evolution in Network Markets – Studies from eBay”

Harikesh Nair, Stanford “Big Data and Marketing Analytics in Gaming”

Closing: Anitesh Barua, Chair 

Andrew Whinston, University of Texas

Prabhudev Konana, University of Texas

Anitesh Barua, University of Texas

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s