Our friends from REDC Research, Ireland, presented a paper at the Esomar FUSION event in Dublin about their use of AI in Social Media image analysis.

Social media is a key part of daily life for Irish consumers – as intrinsically entwined with the day-to-day as the brands we market. With 65% being part of the Facebook community and 32% having an Instagram account, the Irish are using social media to connect with the world and this figure is growing. The reality is that the world of social media runs in parallel to the world in which consumers truly live. Identities, attitudes and behaviours are in flux and often not a reflection of the reality but rather the aspiration – and social media platforms further enable this. Personal content posted and shared is filtered through tools provided, enabling the presentation of an enhanced but desired self and they invest heavily in this projection; time is dedicated to this in while alone and in the company of others.

So, we have a group of consumers presenting as a perfected rather than genuine self on a relatively new platform – what good is that to discipline that consistently seeks to unearth an authentic understanding of attitudes, beliefs and behaviours? In truth, as marketers, we are selling the ideal and this is what consumers purchase from us. Whether you’re thinking about banking products that assist in the securing of an aspirational financial position, to beauty products that promise a revised but better version of self, we too are presenting the dream. There is potential for research to avail of the content available on social media to better understand what it is that consumers aspire to be so as to better understand their dreams, hopes and wishes and to become more relevant on the back of it.

With a curious mind, RED C considered the potential of applying clear thinking qualitative analysis skills to the masses of data available on social media earlier this year. To optimise our chances of extrapolating value from the effort, we worked with software company Beautifeye, who specialise in data analysis solutions incorporating A.I. (Artificial Intelligence) to make sense of significant volumes of images and text, allowing RED C’s focus to lie within our area of expertise – thinking!

The skills utilised are predominantly qualitative in nature and so the opportunity to explore was handed over to the RED C Qual Team. Although the information available is monstrous versus that traditionally seen for this research approach, A.I. is there to help. The software enables the collection of images and text at scale across Instagram and Twitter within a specified brief. A.I technology then codes the collected data based on identifiable commonalities. Qualitative researchers can then analyse the coded output to identify key insights.

Briefs are vital to optimum data collection – clear and specific objectives need to be identified and ringfenced by region, gender and post recency to collect the most relevant and useful data. From here appropriate content is identified based on tags linked to the image and text. Further, ensuring that we respect people’s privacy in a GDPR age is of utmost importance.

Having conducted a number of bespoke projects using the approach, we have learned a lot about social listening and its relevance within marketing:
1. Qual at Scale
How many qualitative researchers have stood up to present findings only to get the question “And how many people said that?”. The volume of images and text collected delivers a sense of comfort that can be absent from more traditional approaches.
2. If it’s There, They Care
Items shared on social media tend to be of high value to the person sharing and is therefore something that they are genuinely interested in (or want to be seen/associated with).
3. It’s the Platform of the Now and Next
Understanding online, social media perspectives – and image in particular – is key to understanding consumers of today.
4. Their Lives Through Their Lens
Although the approach does not enable probing, the content and context of posts provides an unexpected depth in terms of sentiment understanding. They also help to bring reports to life.
5. If not Instagrammable, it’s not Worthy
Due to the nature of the social media platforms, you get a sense of priorities – conversation will not reach the platform unless it’s social-media-worthy.
6. Coding is Not Coding as We Know It
AI coding is not developed from a research perspective; close collaboration with software providers is required to optimise analysis output.
7. Good Qualitative Analysis Skills are Essential
While AI is valuable, for the volume of data available it’s key to have a qualitative analysis skill set that enables us to look past the superficial and understand the underlying choices from all the information provided. Allowing us to decipher the why.
8. Targets are specific
Using social media means data will be collected from younger age profiles – those under 30 years are the key user group. In addition, only publicly available profiles are available for analysis.

In its current guise, we believe the platform is of most value in terms of deepening an understanding of brand context and category and in directing comms. However, the possibilities of social media research as another source of insight in how humans feel and behave are undoubtedly significant, and we’re only at the beginning of experimentation.