The business of social data – are we using it respectfully?
There’s plenty of fizz around outbound applications of social media to business issues, but it’s important to take a small step back and see how to answer business issues by looking at social media data, says Rob Kramer.
Social media in broad strokes is the same as ever – but are we using it respectfully? By posting online – text, video or photo – you’re providing information about what matters to you. Our job in building marketing and brand intelligence is figuring out how to use it respectfully. This can include brand or category experiences investigated by close content analysis, or applying a somewhat more superficial approach by just looking at changes in volume. Depending on the brand or category, volumes can be proxies for or even predict certain brand measures.
And that’s just the tip of the iceberg – where data permits, meta analysis can reveal communities of interest by connecting individuals or topics to reveal contextual nuances that exist across rudimentary demographic demarcations. Quite simply, Kantar has learned after years in the field that there is no single answer except to continually think critically about what information is currently accessible, and how that matches against what you’re trying to achieve – I use ‘currently’ deliberately, as analytical options come and go according to platform prerogatives.
Three key considerations
The absence of evidence isn’t the evidence of absence
This is the perfect segue to acknowledge the myriad challenges in leveraging social media data. Apart from easily mitigable risks around personally identifiable information (PII) and intellectual property (IP), its crucial to remember that the people we’re listening to may have dramatically different priorities than discussing a brand’s equity or image compared to practitioners in our industry.
We must accept that the absence of evidence isn’t the evidence of absence. It just means people aren’t as interested in saying if they like the brand or not. What we get is insight as to what motivates people to say something.
How do we know what’s real and what’s noise?
Whether from bots or other influencers who are juicing a brand’s voice without calling out the association, what’s reality? Colleagues looking at political campaigns and issues seem to have more interaction with the bot brigades – all we can do is keep abreast of likely signals to clean out of our data set. The question of influencers is fascinating – again, worth a separate discussion around how we either assess their contribution to a brand campaign, validate their audience, or otherwise ensure that we’re engaging with them as effectively as possible.
Context is king
Colleagues from the survey side of the insights industry frequently point out that we have little recourse to adjust for sample bias within a collection of social media posts. They’re quite right, and rather missing the point. Using acquired data isn’t just point and shoot. It’s critical to understand the data context: different networks have different audience profiles, different brands and issues elicit different levels of involvement.
We need to be clear that social is a sample of opinion, which is why modelling can be a good idea and taking social data in isolation of other audience views is a very bad idea.
The approach has changed
Broadly following the archetypal hype curve, using social data has become more pragmatic as we move on from the naive ‘wild west’ of a cornucopia of freely shared data. At present, particularly from an insights and strategy perspective reliant on licensing data from platforms under various privacy regimes, the trend in the last 12 months is less about individuals than themes and audience interests. This trajectory is likely to continue as we keep adapting to what the platforms will permit access to.
As data security and privacy implications become clearer, the trend is releasing less as opposed to the earlier approach of sharing as default. But that’s okay – if we keep a symbiotic relationship with the data sources and use our experience in getting value from multiple aggregated data sources versus the prior model of leveraging individuals, the approach should remain fruitful.
That’s on the ‘challenge’ side, but the huge upside is that data processing is getting more sophisticated as we apply improved taxonomies, more automated classifiers for text and image and improved models to move from high-level trends to usable details more quickly and surely.
What they want to know
Of course, we could undertake an ambitious first-party data collection exercise to collect everything that a sample audience does online, then analyse a gazillion data points to figure out how they connect with brands online… if at all. Yet, there’s a data source that effectively represents a huge sample of the online population, trended over time, capturing every privately expressed need, question or desire.
We know that search engine marketing can be an exceptionally effective means of driving direct response – but can we glean any insight as to what people need? Yes! Just ask Seth Stephens-Davidowitz, the economist and former Google data scientist whose book Everybody Lies shows that this could just be the most important dataset ever collected.
How does search data help?
It’s a same-but-different approach to seeing what’s important to people online – where social is what they feel like expressing, search is what they want to know or do. Translated, it’s more salience than equity or awareness that social corresponds to. And we know that search is almost a universal touch point – it’s obsoleted entire information organisation industries so understanding when and how people search, as well as what they’re after, really builds out that touch point understanding.
It’s also fantastic in that as one search leads to another, we can use that to, say, build a category picture of need or build out a map of themes related to a brand. Additionally, we can apply similar processing for social to recombine queries to match things we’re needing to understand, as opposed to that which the data providers organise into – it’s a delightfully rich source as you delve into the data.
And the best part? The data comes pre-aggregated, so there are little, if any, personally identifiable information (PII) considerations in acquiring and analysing the data.
What do you know? How do you know?
Periodically and critically revisit your assumptions about the data channels you are exploiting. It’s clear that social data is not a pure reflection of individuals’ opinions; yet it remains more than a channel to direct communications to those individuals. With a clear-eyed analytical perspective, it is a valuable lens on the business issues we already address via other means. But by itself it leaves blind spots, so a thoughtful marketer will also include the view from consumer intent, such as that embodied in search data. The information is there, we simply need to consider it thoughtfully and in context with the rest of our business intelligence.
Rob Kramer is search, social and innovations lead at Kantar
- Five things that will shape social media in 2019 »
- Get antisocial – what Lush UK’s social media departure says about the landscape »
Image credit:Eddy Billard