In today’s connected world, much of how we communicate is visual. From emojis to selfies, images are everywhere and have become part of the modern lexicon – especially on social media. Yet to date, most social media analysis has been purely based on the written word – not on images.

Imagine the power of being able to analyse brand and product imagery on social media

For furniture designer Arper, we did just that – and by doing so we made chairs (and tables and sofas) talk.

Arper is a fast-growing Italian company that manufactures and distributes designer chairs, sofas and tables for contract, home and work sectors worldwide. Arper wanted to understand the amount of interest in its brand through digital mediums in English speaking countries – such as the USA – and extend the analysis to German and Sweden as relevant markets. In addition, Arper wanted to decipher relative strength of its brand vis-à-vis competitor brands.

The power of AI

The power of AI (Artificial Intelligence) is dependent on the input being fed into it by humans. Hence, we started the project with extensive desk research and a few depth interviews within Arper and with architects. Based on those, we started extracting data from social networks (Twitter, FaceBook, LinkedIn, YouTube, Instagram, Pinterest, etc.), media portals, designer websites and forums. In this sense, we used a data collection method that extracted people’s opinions in the format and style of their preference, not through surveys and not through focus groups, but rather through their own spontaneous conversations.

We quickly realised that this sector was heavily image led with less usage of text, i.e., people were sharing images or talking via images with minimal usage of text. This posed a serious problem as most of the data extraction methods available today are text based. Hence, apart from relying solely on textual keywords to extract data, we started using images (9000+) as seeds to feed into our image crawlers which subsequently extracted 1.1M+ images. These crawlers are similar to Google crawlers but instead of text they work on images; they not only covered all major design websites and forums but also all image dominant social networks like Pinterest and Instagram These images were also tagged to better understand the content inside the image. And, we found ways to submerge themes from images to themes from our text mining, giving the client a complete picture of the markets across text and image.

The Importance of Image analysis

Because of this image analysis, which aimed to extract insights from free flowing conversations, we unearthed some very important insights around the strength of certain brands which was not getting reflected via our text-mined data. Images changed the entire understanding of the sector and brand strength. Combined with our text mined insights, we were able to deep dive into the strengths and weaknesses of brands, their products, the impact of influencers, sector trends and much more. The analysis was well received by the mid-management and C-level management which has started acting on the recommendations, moving to an image led expansive strategy and relevant investments while seeing improving results with every passing day.

Paper: Can Chairs Talk was co-authored with Chiara Davanzo Zamarian (Brand Manager, Arper) and awarded Best Paper (Overall) at ESOMAR World Congress 2017. The full paper can be obtained from WARC (http://bit.ly/2xarnMm) and is also available to all ESOMAR members. Presentation slides can be found online: http://bit.ly/2fDAYpc

Preriit_SoudaPreriit Souda

Director: Data Science, Kantar TNS UK

Twitter: @preriit2131 Linkedin: https://www.linkedin.com/in/preritsouda

LEAVE A REPLY