Social buzz is a critical, but often volatile factor when driving regular sales within FMCG. A single comment could affect brand reputation and sales unexpectedly.

Beyond regular social figures, the context of the post is what matters to sense the intention a post generate. And beyond sensing the intention, how a post is affecting sales is what really concerns brand. Therefore, we invented Sales Intensity Index, a complex metric to not only “measure” but also “predict” how social posts may affect sales, to give early warnings for social communication.

Costco, accounts for 15% of adult milk powder sales, is a significant sales channel for KLIM Taiwan, but importantly - one with a restricted policy on advertising & in-store merchandising, making it difficult to promote Costco listings elsewhere.

However, Facebook Costco group, a place for Costco buyers to share the product experience, becomes a very fierce battleground for brands. Posts inside the group often raised unexpected attention and sales momentum. These powerful social conversations can soar within days or even within hours, catching brands unprepared. Once the damage, like negative comments or competitors’ brands gained any advantage, it would take KLIM a high amount of effort to reverse the impact, if even possible at all.

KLIM Taiwan found it challenging to manage its influence at an early stage, while traditional social listening’s engagement figures were not directly reflecting sales impact.

There are no objective metrics to evaluate how the social posts affects actual sales. Beyond trackable figures (likes, comments, and shares), it is critical to understand the level of purchase “intention” provoked from the post, and how all these factors combined would predict its intensity on sales impact.

If we could capture the sales intensity with every evolving posts, then we get a chance to identify potential threat or opportunity before it goes big and provide real-time warnings for brands to get ahead of the game.

We revamped traditional social listening on Facebook, with technology of natural language processing and corelation modelling to unlock the relationship between posts and sales.

2020/1/8 Data collecting, cleaning, coding, model building

2020/3/18 SII Phase 1 on-air (intention monitoring ready)

2020/3/30 SII Website launch

2020/5/12 SII Phase 2 on-air (co-relating with sales data)

We invented the Sales Intensity Index(SII), an AI-powered social sales prediction solution.

In phase 1, we analyzed 3-year historical data of FB Costco groups, including likes, comments and leveraged Natural Language Processing to “read” the content of the posts.

The intent of each comments was identified into 7 groups: neutral, aware, interest, recommend, experience, purchase intent, and negative.

Entering phase 2, a correlation analysis between historical Costco sales data and post intentions was done to uncover the relationship between social reactions and sales, forming the Sales Intensity Index.

We scan FB Costco groups every 10 mins to convert the intention into SII and generate a SII curve for each posts.

Once KLIM-related posts (including posts of competitor brands) get outstanding SII growth, indicating high probability that it would affect sales, an alert email is automatically sent to the clients and social teams.

The evolvement of the SII enable us the track the intention of each posts and build active social placement to engage audience better, taking necessary action in real time to avoid any possible loss, or even maximize possible gain, before buzz grows too big to control.

SII highly improved the efficiency of social monitoring, filtering 75% of unimportant posts, saving approximately 300 hours a year, almost 2 months full-time working, to deal with irrelevant posts.

SII provided automatic alerts for possible threats/opportunities within 1 hour of posts being published, enabled KLIM social team to early engage and provoke purchase intention, uplifting the influence of KLIM posts with a 69% growth of SII.

We identified the high momentum posts of competitor brand and actively balanced the talk with alternative opinions, which distracted the desire toward that brand. Influence power (SII) of the competitor brand posts dropped from ave 755 per post (2019 Nov) to ave 339 per post during its recent promotion (2020 Nov).

We successfully enhanced the communication toward Costco buyers while Costco had restricted control. And the accuracy of SII continues to provide an impact as we use machine learning to optimize our algorithm constantly.