Announcing a Breakthrough in Measuring the Impact of Social Media on Sales

on Thursday, 31 May 2012. Posted in Innovation

Based on work by Michael Wolfe, Bottom Line Analytics


Everyone agrees that social media measurement is important, but figuring out how to do it hasn’t been easy. There are many tools to measure social media activity, and some like, are even free. But until now, there hasn’t been a systematic approach to measuring the financial returns from social media efforts, much less a way to evaluate their contributions relative to traditional media.

Marketing-mix modeling applies econometric predictive modeling to link various media and marketing efforts to brand revenue.   The resulting model enables a brand to determine its marketing ROI. Incorporating social media to the model is a logical next step. This step requires a definitive measure that reflects social media activity as well as intangibles like sentiment that make social media so powerful.


The Holy Grail of Social Media Measurement


To determine the best approach, various sentiment metrics from leading social media data vendors were tested. This approach quickly reached an impasse.  To be useful in a predictive model, social media sentiment metrics must correlate to a brand’s retail sales.  None of the sentiment metrics from the six leading social media vendors had very strong correlations to brand sales.


This finding led to the development of a new social media metric, called Social Engagement Index, or SEI™. It is based on the principle that conversations in social media are more than just words. Semantics matter. Words, language and context are key to understanding the level of ‘customer engagement.’


The approach involved five steps:


  1. Social media channels (Twitter, Facebook, blogs, etc.) were scraped for brand specific conversations.
  2. Linguistic and language analysis tools were applied to separate conversations according to positive and negative sentiment.
  3. An algorithm was created to “score” social content based on thirty linguistics rules according to its degree of emotion and personalization.
  4. The scored conversations were aggregated to derive a time-series metric of positive and negative sentiment.
  5. This metric was correlated with brand sales across four different $2+ Billion brands, representing the telecom, food and beverage and hospitality industries.


The result was astonishing: SEI™ has an incredibly robust correlation to brand sales.  


Across all four brands, the correlations were in the range of +76% to +89%, more than adequate for predictive modeling.  Figure 1 below represents correlations of SEM™ and sales for a food and beverage brand. Similar results were obtained for technology and hospitality brands (not shown).


Figure 1: Correlation of SEI™ with Retail Brand Sales 

Figure 1


These correlations are especially impressive when compared to those of off-the-shelf social media sentiment metrics (Figure 2 below). The best of the off the shelf measure shows only a 21% correlation, far below the 83% for SEI™.


Figure 2: SEI™ Correlation with Retail Brand Sales Compared to Leading Sentiment Metrics


Figure 2

Impact of Social Media Relative to Traditional Media: Buzz Matters!


The next step was to see how well SEI™ predicts brand sales relative to other media and marketing efforts, both traditional “offline” and digital.  Again the results were astonishing.  Using SEI™, social media is seen to have both a direct and indirect effect on brand sales. The impact is far from equal.


As can be seen in Figure 3 below, direct spending on social media accounts for only 2% of retail brand sales. In contrast, the indirect effect of social media-generated word-of-mouth is a whopping 37% of total brand sales. This finding suggests that social media enhances the impact of other marketing efforts and significantly improves total marketing spend ROI, and this enhancement is far greater than its direct impact. This was always the theory, now we have proof.


Figure 3


Positive vs. Negative Sentiment


SEI™ also reveals how social media conversations influence sales. Figure 4 clearly illustrates the relative impact of these different sentiments. In absolute terms, negative-toned conversations have a significantly greater net impact than positive ones, by a factor of 4X. This finding indicates that it is more productive to manage and reduce negative-toned brand conversations than to attempt to increase positive ones. Until now, anecdotal evidence supported this finding; now we have proof.


Figure 4


Putting the Model to Work


All this is exciting news for marketers. Everyone wants to understand how social media affects brand performance, for better or worse, and how to increase its effectiveness. With SEI™, we can tell which activities work, how hard they work, and the return for every level of investment. It’s all knowable.


Getting started can be as simple as creating a baseline model using existing information about sales and marketing spending over time. Creating historical measures of SEI™ allows us to fine tune the model to and quantify the relative impact of positive and negative social media conversations.  With this information, we can then create a plan for ensuring social media plays an optimal role in the overall marketing mix.


About Michael Wolfe


Michael Wolfe has 25+ years of experience in marketing analytics that has taken him across firms such as Kellogg’s, Fisher-Price, Kraft Foods, Coca-Cola and his own firm, Bottom-Line Analytics, a Brand Amplitude partner firm. Michael has a Master’s degree in Economics from the University of Iowa. He develops innovative approaches for measuring the impact and ROI of social media on brands based on media mix modeling.


Contact: This email address is being protected from spambots. You need JavaScript enabled to view it.

Leave a comment

You are commenting as guest.