Last time in our ‘getting more from Facebook topic data’ blog series we focused on PYLON – DataSift’s social analytics platform for non-public networks. This is designed to give marketers in-depth insights from the data-rich non-public networks and is the first of its type in existence.
When talking about non-public networks, there is one in particular that stands out above all others – we are of course, referring to Facebook. The sheer scale and quality of the data within Facebook is makes the platform one of the best representations of the world’s population. It has demographics spread more evenly across age, gender, location, income, and education than any other public or non-public network.
To really fulfill the promise of PYLON as a privacy-first social analytics platform, we partnered directly with Facebook. DataSift’s technology is embedded within Facebook’s data centers and leverages Facebook’s social graph to provide unparalleled insights into public opinion. Facebook topic data provides access to one of the largest focus groups in the world, without compromising user privacy.
Why Facebook ‘topic’ data?
Every time people post on Facebook, its ‘topic inference engine’ detects topics in the text of posts and attaches normalized topic labels to the data in real-time. Multiple topics can be inferred on each post. Facebook leverages the continuous creation of millions of new Facebook Pages to ensure the latest subjects being engaged with are reflected.
This allows social data analysts to analyze the most common topics in their datasets and really understand the zeitgeist of Facebook activity. Marketers can also study the co-occurrence of topics in the same posts to understand communities of topics of interest to a specific audience.
Facebook assigns topics to a post based on natural language processing. The topic universe presents a ‘fresh Web’ taxonomy which represents activities, products and other subjects of interest that evolve over time. Each topic lives in one of more than 200 high-level categories and in PYLON, analysts can constrain their analysis to topics within one of those categories.
The algorithm behind these category and topic classifications is proprietary to Facebook, and is the perfect starting point for exploring data with little effort and no delay. Analysts also have the choice to complement this functionality with their own domain-specific classifiers and categorizations for more detailed or specific results.
Characteristics of Facebook topic data
More than any other non-public (or indeed public) network, Facebook topic data features an incredibly broad array of metadata, beyond just topics and categories, which is at the core of a truly multi-dimensional dataset. These are the main types:
- Demographics – All interactions in Facebook topic data come with self-declared (rather than derived) demographic information about the author of a story or engagement, including age, gender and region. Self-declared demographic data is considered of higher quality because it is directly reported by the users (and is therefore more accurate) as compared to inferred data, which is based on interpretation of activities.
- Sentiment – Facebook topic data comes with built-in sentiment analysis provided by Facebook’s sentiment analysis engine. Sentiment is classified as positive, negative or neutral.
- Hashtags – Facebook topic data supports the filtering and analyzing of hashtags inserted into stories by users. The most engaged hashtags are automatically returned when queried, allowing for easy finding of the most popular examples.
- Links – The links augmentation in Facebook topic data extracts all links contained in the text of a story, as well as the titles of links shared for a comprehensive analysis on link engagement. The links that are most engaged with are automatically returned when queried.
- Super Public text samples – Facebook topic data recordings come with a sample of Super Public posts that Facebook users have chosen to share publicly. Super Public posts are useful as they allow marketers to validate interaction filters and have specific examples of the data returned.
Facebook topic data can give a marketer the deepest insight possible into an audience, its likes and dislikes, and even into how and when to engage with it. It is arguably the greatest leap forward in the history of marketing, a genuine game-changer.
However, it does require a different mindset and approach to extract this insight. Our next post in the series will look at how analysts can leverage this data to fit their own specific use cases, and the techniques that can be used to maximize the value of Facebook topic data.