Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Social media metrics: in search of the holy grail


Published on

Social media metrics: in search of the holy grail

Michael Wolfe - Senior Director - BBDO

Discussing the limitations of traditional marketing modeling for measuring the impact of online conversations on brand performance. Semantic Engagement Index: an innovative approach for measuring the impact and ROI of social media on client brands. Demonstrating how positive and negative sentiments can impact brand sales and influence consumer engagement over time. The convergence of social media and mobile: how can mobile be used for measuring online conversations?

Published in: Business, Technology

Social media metrics: in search of the holy grail

  1. 1. 2011 Market Research in a Mobile World July 2011 Michael Wolfe 1
  2. 2. THE HOLY GRAIL?• Don’t believe it for a minute, it doesn’t exist!• But some metrics are better than others.• There are many social media metrics available today. – Many simply count words with rudimentary sentiment differentiators – Others based on Influence, klout scores, number of followers, etc – Still others based on language and Linguistics science• We have developed a special metric called the Semantic Engagement Indexsm(or SEIsm). – It is based on the principle that conversations on social media are more than just words or what is said, but how it is said and the context of the conversation. Semantics matter. – We directly compare our SEI with other social media sentiment metrics from a leading social media aggregator company.
  3. 3. SEI AND SOCIAL SENTIMENT METRICS• The comparison is with our Semantic Engagement Indexsm and a Social Media Sentiment indicator from a leading social media data company. As you can see, the Sentiment Indicator has poor and somewhat contradictory correlations with our selected company’s sales. The SEIsm metric, however, shows strong and robust correlation and logically consistent to the positive and negative forms. Correlations to Sales
  4. 4. SEMANTIC ENGAGEMENT INDEX: NUTS & BOLTS• This metric is based on an algorithm devised with the assistance of Boyd Davis, Professor of Applied Linguistics at the University of North Carolina and a vendor partner named Linguistics Insights. The process we go through here involves four steps and it leverages linguistics science heavily. 1. First we secure large quantities of social media conversations filtered for topics on the specific brand of interest. In the case of the hospitality client, this was focused on online review sites pertaining to hotels, resorts and cruise lines. 2. Next we parse these conversations data into positive and negative toned conversations. 3. Next we apply our proprietary algorithm which quantitatively “scores” each of the positive and negative groups along the two dimensions of “emotional affect” and “personalization”. This scoring algorithm applies the science & rules of Linguistics. 4. Finally, we time-code each conversation into topics and aggregate into a time series metric.
  5. 5. CASE STUDIES• To fully leverage the SEIsm for our clients, the task is to understand its impact on their business.• To do this, we do exploratory analysis to see how relevant the metric is to the customer demand of a number of clients.• Then we utilize the SEI within a full marketing response (aka, mix) model in order to not only understand its impact on the business, but also how it interacts with and is affected by direct marketing.• First, however, we compare the customer SEIsm metric to 2 clients’ customer demand over time. – One client is a food & beverage retailer, while the other is in the hospitality industry – The specific SEIsm metric we use here, and in our model, is the SEIsm ratio of positive to negative tonality conversations.
  6. 6. TOTAL HOSPITALITY BOOKINGS • For our hospitality client, the key focus of our SEIsm metric was “online” review sites for hotels, resorts and cruise lines. This makes the metric a proxy for customer satisfaction. As shown, the correlation is a very robust +76%! 140 160 120 140 120 SEI Positive/Negative Index 100 100Bookings Index 80 80 60 60 40 40 20 20 - - 1/7/2008 1/7/2009 1/7/2010 Bookings.Index SEI Positive/Negative Ratio
  7. 7. TOTAL FOOD & BEVERAGE SALES 130 400 125 350 120 300Retail Sales Index 115 SEI Positive/Negative Index 250 110 105 200 Correlation of Sales to 100 SEI Metric is 87% 150 (Total 156 Weeks) 95 100 90 50 85 80 0 07/09/07 08/04/08 08/31/09 TOTAL Retail.Sales SEI Positive/NegativeRatio 7
  8. 8. SEI AND MARKETING RESPONSE MODELS• To fully leverage the SEIsm for our clients, the task is to understand its impact on their business.• By incorporating SEIsm metrics into marketing response (aka, mix) models, we can: – Come to a better & more precise understanding of how social media buzz affects a client’s business performance – Understand the impact and interactions of the client’s marketing and media as it affects social media conversations about their brands. – Provide strategic guidance as to the most effective ways for monitoring and managing social media conversations on brands
  9. 9. THE IMPACT OF SOCIAL MEDIA TONALITY • A key insight we uncovered across clients is the difference between “positive” and “negative” conversations about a brand. In absolute terms, the negative-toned conversation has a significantly greater net impact. It is actually better for a firm to try to manage a reduction in negative toned conversations than to increase positive ones. Total Retail Sales Elasticity of Response Total Retail Impact of 90% Change(change in Sales versus change in Tone components of "Engagement") (assumes 100% Decrease in Negative is not realistic) 20% 20% 10% 15% positive Sales Impact Sales Impact 0% 10% negative +16.5% -10% 5% -20% +4.4% -100% -50% 0% 50% 100% 0% Change in "Engagement" by Tone Increase Positive Decrease Negative 9
  10. 10. THE IMPACT OF SEI, SALES AND MARKETING• We also learned that a there is both a direct and an indirect effect from client media & marketing. As you can see, the impact of Semantic Engagement is quite large. The “indirect” effect of marketing is due to the impact on the SEI metric, itself. Overall, this indirect effect is shown to greatly enhance the impact of the client’s marketing and significantly improves marketing spend ROI. Total Retail Sales Contribution net driven by 6% 2% marketing 8% +6% +13% = 27% Sub-model 6% 13% 37% Engagement 55% 18% Base Sales Direct Alpha Brand Media & Mktg Direct Social Media Mktg Blended Media Other.Media Other Base 10
  11. 11. PREDICTING POSITIVE CONSUMER ENGAGMENT (SEI) We apply CART (Classification & Regression Trees) to score subject groups to determine what conversational topics & issues are driving consumer engagement. Brnd&Place >14.29 The most important positive Brnd&Place =461 drivers are: 0.3% >9.63 1. The Brand & Place =320 Brnd&Place 2.1% <14.29 2. For Meeting People Brnd&Place >5.75 Brnd&Place =298 3. The Beverage 1.8% =219 <9.63 4. The Store Atmosphere 9.9% =192 7.8% POS SEI 3.76 =100 Meeting Brnd&Place People>5.84 <5.75 =229 3.6% Beverage>6.29 =87 =255 Brnd&Place>3. 90.1% Meeting 7.9% 64 = 121 Atmosphere People<5.84 7.0% >5.25 =81 =209.6 Beverage<6.29 86.5% 1.3% =79 Brnd&Place<3. 78.6% 64 = 72 AtmosphereThe tree starts with an average SEI score of 100; and each level 71.6% <5.25indicates a higher or lower SEI based on an SEI score for a topic.. =70The percent represents the percent of the sample in each segment. 70.3%
  12. 12. THE MOBILE RESEARCH ANGLE?• Certain social media data companies can now collect social media data for specific geographies and demographics. – –• Social media as a tool for prediction: – Bernardo Huberman, Hewlett Packard • Predicting elections • Test market analysis –
  13. 13. LESSONS LEARNED• Through a number of case studies, BBDO’s “the Worth” has attempted to push back the frontier on social media metrics and understanding• By linking a metric of “Semantic Engagement” to client sales and demand, we have shown that this approach shows great promise as a diagnostic for understanding social media’s impact on a client’s business by including it as an input into marketing response (aka, mix) models.• Some of the key lessons that we have learned include: – For a service-based hospitality client, the consumer’s rating of the quality of the firm’s service, is in fact, the most important driver of that firm’s business. – That negative brand conversations have a greater absolute impact on a client’s business than corresponding positive reviews, i.e. negative word-of-mouth travels wider & deeper than positive. – That the direct impact of SEIsm on business is large and significant. A brand’s marketing and advertising affects the SEI sm which in turn, affects sales. – We learned that the value and ROI of marketing is greatly enhanced due to the indirect effect it has on sales through its direct impact on Social Media Engagement (SEIsm ) – That our SEIsm metric is no Holy Grail, but it shows much promise in delivering un-matched insights on how social media conversations have a direct and tangible impact on company performance.
  14. 14. Presented at: Market Research in the Mobile World 2nd International Conference | July 19 & 20, 2011 AtlantaOrganized by: Thank you to sponsors:LinkedIn Group: Mobile MRUpcoming Merlien Events: GreenBook Directory: GreenBook.orgMarket Research Blog: New Qual Blog & Directory: NewQualitative.orgGet Inspired. Stay Informed. Cut Through the Clutter. Sign up for our free email newsletter