From open communication to closed interaction, the ecosystem of social data is constantly changing and evolving. Amidst shifting landscape of consumer behaviors and privacy concerns, how do you adapt and capture sophisticated, multi-dimensional audience insights? This blog is the first in a series of posts to help you build deeper insights products using Facebook topic data.
How social data analytics will adapt to the mass migration of consumers to non-public social spaces
Social data provides the world’s largest real-time source of detailed market research opinion. But as social networks mature, the nature of the spaces where we interact with friends and colleagues is changing.
Four key types of networks have emerged, and each has evolved its own role in the way we share and communicate with each other:
Public spaces. A large minority of American consumers now contact brands directly when they have a product or service issue, either with an @ mention on Twitter or a post on the brand’s Facebook page. Over the past 2 years, more than a third of US online consumers have reached out to brands this way, according to Forrester.
Walled gardens. Discussions about what to buy, where the people we’re connected to influence our choices, now occur largely on non-public networks, away from the prying eyes of brands. That can mean asking which of two cars you should buy within the relative safety of your Facebook friend network, or seeking opinions on two competing software solutions on LinkedIn.
One-to-one spaces. One-to-one and small group messaging platforms are the fastest growing communication services in 2016. Ranging from Snapchat, the ephemeral messaging platform which brands are just beginning to figure out how to leverage, to tools like Facebook Messenger and Whatsapp which are displacing SMS, these spaces offer us something that was previously missing from the social landscape.
Image-oriented spaces. Highly visual platforms like Instagram present a unique challenge for social data analysts in that the text is sparse. Image recognition technologies within social platform have thus far largely focused on facial recognition to aid us in tagging our friends in photos. The landscape of technologies for identifying brand logos and products is comparatively nascent. We think social networks themselves, with their mountains of training data and top-flight data scientists, are best positioned to develop this technology for marketers, as opposed to commercial providers. In the meantime, analysis of networks like Instagram will continue to revolve around surfacing new hashtags applied to images.
As opposed to expecting that large quantities of verbatim text will be available for analysis from each network, the social data ecosystem is beginning to recognize that the approach to analytics for each of these four types of networks is fundamentally different. Technology builders and analysts who deliver insights out of social data to brands must adapt to this shift. Particularly for walled garden networks, where most text cannot be displayed. The good news is, there is a massively beneficial tradeoff for our industry in all of this: when we step away from individual-level data and begin to work with anonymized, aggregated data, social networks can begin to share with us much more insightful information than we have ever had access to before.
Social data is about to go through the same shift that business intelligence data went through about 15 years ago: a multi-dimensional explosion off the screen into 3D. Business analysts, long accustomed to dealing in flat pie charts of sales figures, suddenly had a richly multi-faceted cube of operational business metrics to work with. Building multi-dimensional views, contextualized with baseline comparisons, was a big evolution for BI which took time. Similarly, social data analysis tools with access to richer demographics on aggregated, anonymized data, are now poised for this same shift.
As our industry evolves, counting brand mentions for share-of-voice comparison against competitors will no longer be enough to compete. Highlighting the demographic segments with heavy interest in a product requires slightly more advanced statistical techniques. So does identifying the affinities of a brand’s audience to other brands, activities and events in their lives, all with the purpose of building messages that resonate with the right people.
To compete in social data analysis in the second half of this decade, analytics teams and platforms will need to look across a rich dataset of audience demographics and interests, and surface consumer insights from the activity of those networks in aggregate and at the audience level. This requires a change in mindsets and skillsets which is already underway. Flat, one-dimensional dashboards fail to tell the story of how audiences express their interests. Teams that manually tag social posts with brands and products mentioned struggle to derive actionable insights from the text alone without the context of audience-level metadata. Topline mention counts that passed for social analytics in the first generation of listening products are now more likely to be rejected as vanity metrics.
The next generation of products and services around social data will need to cross-reference brand-focused results with sample data for context. They will need to perform statistical analysis of text to identify a dataset’s uncommonly common words, phrases and themes. They will look for the intersections of demographic characteristics and interests that represent new addressable markets. All of this will require a strong combination of technology-powered analysis and people-powered contemplation of the results.
A new era is dawning in our industry built on deeper insights from better data. It’s an exciting time to be studying our relationships with brands, products and experiences in a privacy-first, multi-dimensional way.