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.

Improving employee engagement and performance through social analytics sadat shami

588 views

Published on

The use of social media in the workplace is growing.  One anticipates this trend to continue as more and more individuals comfortable with social media join the workforce.  Social media provides an opportunity for organizations to obtain a real-time understanding of various aspects of the employee experience.  For organizations to truly benefit from this aspect of social media, there is a need to build tools that allow an organization to make sense of the large scale unstructured data generated by social media.  In this talk, I will first introduce a tool named Social Pulse that enables an organization to understand the social media chatter and sentiment of its employees.  Social Pulse provides text and sentiment analysis, search and filtering, and several visualization features while respecting employee data privacy.  I will then discuss how the data generated through Social Pulse can lead to insights that can improve outcomes of interest to an organization, such as employee engagement and performance.

Published in: Business

Improving employee engagement and performance through social analytics sadat shami

  1. 1. Achieving improved employee engagement and performance through Social Analytics N. Sadat Shami IBM Corporate Headquarters June 26, 2014
  2. 2. Organization Science Social Media Analytics © 2014 International Business Machines Corporation 2
  3. 3. Outline § Motivation § Social Pulse – Design principles – Features – Ensuring privacy § Challenges – Representativeness – Limitations of the data § Application of social analytics – Social media participation and performance – Inferring employee engagement from social media © 2014 International Business Machines Corporation 3
  4. 4. Opportunities for employee feedback can improve organizational performance § Direct feedback with leadership § Suggestion boxes § Employee surveys – Most common way of soliciting employee feedback Survey Advantages Survey Challenges • Anonymous • Administer at scale • Once a year surveys cannot capture emergent issues • Takes time to design, deploy and analyze © 2014 International Business Machines Corporation 4
  5. 5. Social media can augment employee surveys § Provides a more real-time understanding of the employee experience § Take advantage of an organization’s internal and external social media footprint IBM’s social media footprint IBM Connections Wikis ____________ 300K wiki pages IBM Connections Bookmarks ____________ 1.6M bookmarks, 3.5M tags IBM Connections Micro-blogs ____________ Over 65K status updates / month IBM Connections Communities ____________ Over 50K public communities IBM Connections Blogs ____________ 19K Blogs, 87K bloggers IBM Connections Files ____________ Over 1M files shared Twitter ____________ 32,000 IBMers active / month LinkedIn ____________ 333,000 IBMers Facebook ____________ 171,000 list IBM as a workplace © 2014 International Business Machines Corporation 5
  6. 6. Social Pulse § Basic purpose – What are my employees talking about? – What is the sentiment of the content? § Audience – Employees with business need to understand the employee experience (e.g. Human Resources, Workforce Communication) § Design principles – Augment surveys, not replace – Establish authentic employee content – Only mine public data – Respect employee privacy – Respect terms of use of sites mined * More details available in Shami et al. CSCW 2014. © 2014 International Business Machines Corporation 6
  7. 7. Social Pulse combines internal and external data sources IBM Connections Twitter opt-in * IBM Communities Blogs Forums Status updates * Lee et al. ICWSM 2013 © 2014 International Business Machines Corporation 7
  8. 8. SP features Organic results Keyword search © 2014 International Business Machines Corporation 8
  9. 9. SP features: Sentimap Data segmentation Filtering © 2014 International Business Machines Corporation 9
  10. 10. SP features: World Map Analytics Visualization © 2014 International Business Machines Corporation 10
  11. 11. SP features: Timelines Analytics Visualization © 2014 International Business Machines Corporation 11 ! ! Figure 5. ‘Timelines’ tab showing sentiment over time.
  12. 12. Ensuring data privacy Data from IBM Connec0ons Iden%fier: Email address Data from HR data warehouse Iden%fier: Email address Apply one way encryp0on Scrambled iden%fier Join data where scrambled IDs match Apply one way encryp0on Scrambled iden%fier © 2014 International Business Machines Corporation 12
  13. 13. Example use cases Overall health of a company Senior leaders roll out different communications, change initiatives, programs and policies in the company, and they are interested in understanding if these generate any employee chatter that can provide them with useful input. © 2014 International Business Machines Corporation 13
  14. 14. Example use cases Reputation as an employer Search for keywords related to career, recognition, strategy, and job/work allowed the team to surface content that provided additional context and richness to the quantitative data extracted from the HR databases © 2014 International Business Machines Corporation 14
  15. 15. Example use cases Fine tuning and aligning internal and external messaging Understand language used in relation to specific products/ services to provide improved and consistent brand messaging © 2014 International Business Machines Corporation 15
  16. 16. Challenge: Representativeness © 2014 International Business Machines Corporation 16
  17. 17. Challenge: Limitations of the data Even most satisfied employees ‘read the wind’ before speaking Organizations need to build trust, and foster open communication Will allow employees to influence decisions that affect them Self censorship Focus on the idea of ‘helpfulness’ (Palen et al. 2011) ‘Helpfulness’ is relevant to what is needed Provide clue to analyst to dig deeper to determine if concern exists Noise Witnessing social content grow over time Augment passive listening with active polling on questions where sparseness exists Sparseness of data © 2014 International Business Machines Corporation 17
  18. 18. SOCIAL MEDIA PARTICIPATION AND EMPLOYEE PERFORMANCE For more details see - Shami, Nichols, Chen. Social Media Participation and Performance at Work: A Longitudinal Study. In Proc. CSCW 2014. © 2014 International Business Machines Corporation 18
  19. 19. Application of Social Analytics What is the relationship between social media usage and employee performance? © 2014 International Business Machines Corporation 19
  20. 20. Studied the same 75,747 3 Employees Years © 2014 International Business Machines Corporation 20
  21. 21. © 2014 International Business Machines Corporation 21
  22. 22. © 2014 International Business Machines Corporation 22
  23. 23. Number of status updates over time 52,588 74,047 111,721 120,000 100,000 80,000 60,000 40,000 20,000 0 2010 2011 2012 © 2014 International Business Machines Corporation 23
  24. 24. © 2014 International Business Machines Corporation 24
  25. 25. Number of forum posts over time 44,730 60,800 83,054 90,000 80,000 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0 2010 2011 2012 © 2014 International Business Machines Corporation 25
  26. 26. © 2014 International Business Machines Corporation 26
  27. 27. Number of blog posts over time 7,256 9,311 11,939 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 2010 2011 2012 © 2014 International Business Machines Corporation 27
  28. 28. Controlled for Years at company Job Category Manager / Non Manager Age © 2014 International Business Machines Corporation 28
  29. 29. Generalized Estimating Equations Ordinal logistic regression © 2014 International Business Machines Corporation 29
  30. 30. Results Status Updates CommuIBnMi ty Forums Blog Posts Posts: ✗ Length: ✔ Posts: Length: ✔ Posts: ✗ ✔ Length: ✗ © 2014 International Business Machines Corporation 30
  31. 31. Positive correlation between certain social media use and employee performance Social media use at work NOT considered harmful Social media is not monolithic Visibility is influenced by other’s usage and competing forms of social © 2014 International Business Machines Corporation 31
  32. 32. INFERRING EMPLOYEE ENGAGEMENT FROM SOCIAL MEDIA © 2014 International Business Machines Corporation 32
  33. 33. Application of Social Analytics Can social media text improve the explanatory power of a model predicting employee engagement? © 2014 International Business Machines Corporation 33
  34. 34. Employee Engagement “The extent to which employees are motivated to contribute to organizational success, and are willing to apply discretionary effort to accomplishing tasks important to the achievement of organizational goals” Wiley, Herman, Kowske, 2010 © 2014 International Business Machines Corporation 34
  35. 35. Traditional Engagement Survey http://www.itsinc.net/retention-health.htm http://www.baudville.com/pc/employee-engagement-survey http://ist.mit.edu/for-ist/2012survey © 2014 International Business Machines Corporation 35
  36. 36. Taking advantage of ‘Systems of Engagement’ © 2014 International Business Machines Corporation 36
  37. 37. Word choice in social media can be predictive of engagement © 2014 International Business Machines Corporation 37
  38. 38. Augmenting traditional engagement research to surface more dynamic, fluid views of engagement Inferring employee engagement from social media allows for more real-time views into engagement patterns and changes © 2014 International Business Machines Corporation 38
  39. 39. Step-wise regression model Country Exec / Non Exec Client facing Language dictionaries Demographics IBM 16% 17% 18% 60% of the variability of engagement scores can be explained by the model Social Media © 2014 International Business Machines Corporation 39
  40. 40. A minimum of 25 posts provides the best prediction. There is also a recency effect. IBM variability in engagement scores explained k = minimum number of social media posts 70% 60% 50% 40% 30% 20% 10% 0% The month closest to the Engagement survey (October) had the highest predictive power, suggesting ‘state’ rather than ‘trait’ characteristics. © 2014 International Business Machines Corporation 40
  41. 41. Implications § Systems of engagement provide more predictive power – ‘Honest’ signal – Limitation of small samples § Engagement may be more of a ‘state’ than ‘trait’ – Appropriate action planning can move the needle § An organizational dashboard of engagement – Fluid view of engagement – View engagement across different demographic segments © 2014 International Business Machines Corporation 41
  42. 42. Questions? Contact @sadatshami sadat@us.ibm.com © 2014 International Business Machines Corporation 42

×