Measuring Impact on Facebook

An interview with Daniel Slotwiner, the Head of Measurement Solutions Group for Facebook, on measuring impact on Facebook. 

Daniel Slotwinter, Head of Measurement at Facebook

“There are a lot of misconceptions about Facebook and data,” Chris Moody eloquently opened the interview with Daniel Slotwiner, Head of Measurement Solutions Group for Facebook. For Daniel’s team, their job is to build tools and methods of analysis to highlight the value of Facebook’s media business. But as Daniel explained, it’s not a win for a brand to measure a brand campaign by the CTR it gets. Instead, he emphasized the importance of working with an advertiser who is defining objectives and setting the right measurement program alongside. The Measurement Solutions Group not only tries to build the tools the industry can use, but also educate and work with them to get the most out of the ecosystem. The hope is that the ecosystem will be self-sufficient.

Partnerships

Last year Facebook announced their partnership with Datalogix, initially for measurement. However, with Datalogix’s comprehensive roster of US households, Facebook realized the impact of the information they could provide to advertisers. Datalogix and Facebook have been able to append data of frequent shoppers with consumer purchase decisions. This has aided in analyzing the impact of Facebook in driving offline sales. With more than 80 campaigns executed with these tools, Facebook can see which segments are responding to the advertising and make smarter campaign. At the end of the day, the value of this data is just to calculate ROI, but rather the scale allows for in-depth analysis and huge learnings for not only Facebook, but also advertisers.

Scale 

If the unique advantage of Twitter is that everything is public, Facebook’s advantage is knowing who is saying what. The uniqueness of this data is two fold: scale and concept of identity (demographically and geographically).  If advertisers can understand the value of this data, they have a fantastic starting point.

It’s hard to argue Facebook isn’t doing a good job of scaling their users. “Obviously we love new users,” David said, and it’s still a huge focus for Facebook, as it expands internationally. And they’re prioritizing serving everyone in the world, especially through segmenting. When it comes to the level of use, Facebook has found light users are more receptive to advertising in comparison to heavy users. As  advertisers, understanding this user segmentation can help shape campaigns and execution on the social network. Facebook is intent building these insight back into the advertising systems to help advertisers make better decisions.

Value in Multi-Point Attribution

The world of influencing consumers is only getting more complex. In one sense it’s because there’s so many touch points. Facebook is focused on making sure the measurement systems are keeping pace with the world, but this is virtually impossible. There are a lot of approaches, but Facebook is pretty focused on multi-touch attribution systems to measure. One way they can look into this is through mobile.

Because almost all users access Facebook using mobile, they get to observe a lot and measure they information around mobile usage. This is information Facebook eventually wants to share with the industry. The platform allows for see the different paths to purchases because Facebook has so much visibility into the touch points. Facebook is in a excellent position to observe how many devices people have and how content is distributed across them.

At the end of the day, there’s a lot of data that can be utilized from Facebook. However, Daniel urges the proper use cases of the data. Research, for example is a huge opportunity given the quality of the data. Daniel cautions against the use of the data for its prediction. While a brand may use the discussion online to respond to an emergency or to participate in the conversation, it’s not clear if they should use it as an objective to drive more sales online.

Big Boulder is the world’s first social data conference. Follow along at #BigBoulder, on the blog under Big BoulderBig Boulder on Storify and on Gnip’s Facebook page.

Data Story: Jaime Settle on How Social Networks Affect Political Mobilization

This week we have a data story with Dr. Jaime Settle, a Professor at William & Mary, who studies how social networks affect how we think, feel and behave politically. Her work has been published in the Proceedings of the National Academy of Sciences, the Journal of Politics, Political Research Quarterly, and American Politics Research. We’re thrilled to have Professor Settle speaking at Big Boulder on a panel about research happening in the academic world on social data. 

In a recent Nature publication, “A 61-million-person experiment in social influence and political mobilization,” Settle and her co-authors looked at how political mobilization messages on Facebook during the 2010 US congressional elections affected real-world behavior.

Jaime Settle, William & Mary

1. How was your team able to get access to Facebook data?

Appropriately enough, the connection was made through the power of social networks. The collaboration was formed when I was a graduate student at University of California, San Diego. My advisor, James Fowler, was introduced to Cameron Marlow, the head of data science research at Facebook, through a mutual friend. This friend thought that James and Cameron had a similar vision for the power of online social network data to help us understand fundamental social processes that structure human behavior. Cameron and his team of scientists at Facebook are interested in many of the same underlying questions that academic social scientists want to study, and the collaboration is an effective way to tackle those tough research problems.

2. Do you think your findings would have been fundamentally different if you had scraped data?

What access to the universe of data provides is the ability to detect very small influences that are important because of the massive scale at which they operate. For example, the paper we published in Nature shows how behavior can spread through a social network and that our behavior is affected by the influences our friends receive. The amplitude of each individual influence is small, but the overall effect is massive because of the hundreds of friends we have and the millions of active Facebook users. These small effects would have been impossible to identify without “Big Data” even though the processes at work would be present.  We are also able to make broader generalizations about our results due to the access we have, whereas conclusions must be more circumscribed from research using scraped data because of the multitude of ways in which a small, potentially non-random sample of users may not be representative of the larger population of users on a site.

3. Your personal research has focused on how people behaved differently in battleground states? What were your findings?

I find that people living in battleground, or politically competitive, states are more likely to discuss politics online, and are more likely to do so emotionally than are people who live in less competitive states. These effects appear only in the most intense part of the campaign season, in the weeks leading up to the campaign. I also show that this increased propensity to talk about politics on Facebook explains part of the effect we observe that a higher proportion of people living in battleground states clicked on the “I Voted” button that Facebook displayed on election day in 2008.

4. What do you think social data can tell us about how people are influenced when it comes to politics?

The pace at which we are confirming processes we’ve observed in the offline world—as well as learning new things–about social influence from studies using online social data is really incredible. We are demonstrating that people are more influenced by people with whom they have closer “real world” relationships, and we’re identifying the most influential people in networks. We’re figuring out why some memes are more likely to spread than others, and thus what kind of memes are likely to have the largest influence on our attitudes. We’re able to characterize people’s political ideologies based on their patterns of behavior on social media, and will be able to look at the differences in influence from those we agree with versus those we don’t.

5. What are you interested in researching next?

I’m very interested in the process of how contention and disagreement affect people’s attitudes toward—and participation in—the political realm. My research moving forward is looking at particularly controversial policy debates, such as that over the Affordable Care Act, to see how the context in which people talk about the policy affects their rhetoric and attitudes toward it. I’m also interested in iterated online discussions instead of single expressions of attitudes in status messages.

Thanks to Jaime for participating in the interview! You can check out her fellow Big Boulder panelist’s data story with Sherry Emery of UIC, studying social data and smoking cessation. Click more to see previous data stories! 

Continue reading

What Facebook Data is Available from Gnip's Social Media API?

Facebook is among the most in-demand (but also among the most challenging) social media sources to access. Most Facebook conversation data is private and so it’s not accessible via Facebook’s API or any of Gnip’s feeds. Facebook data availability is also pretty confusing to understand and the rules keep changing. So, let’s clarify what kinds of Facebook information we can offer through our social media API today. 

What Facebook Data is Available from Gnip?
Within the realm of publicly accessible data only, we provide:

  • User page content: status updates, wall posts, comments
  • Fan page content: wall posts, comments (probably more than you’ll find from any service), “Like” count, historical data up to 90 days


How Can You Get the Data?

Instead of a firehose of Facebook data, you enter parameters indicating what you want to find:

  • Keyword search
    You provide a list of keywords. We’ll return public mentions of those keywords.
  • Username search
    You provide a list of usernames. We’ll return publicly accessible posts generated by those users.
  • Fan page search
    You provide a list of fan pages. We’ll return publicly available posts and comments on those fan pages.

    While these lists are vastly simplified, we hope they’ll clarify what kinds of Facebook data most businesses can access legally, and exactly what Facebook data Gnip provides.

    Oh, and one last thing. We’re sometimes asked how we feel about Facebook’s privacy policies. At Gnip, we don’t make the rules — we just play by them. Our job is to facilitate access to the social data that publishers (like Facebook) officially make accessible to our customers.

    Best wishes to you with your Facebook data collection! If you think we might be able to help, please drop us a note.