For a long time the main challenge was the absence of reliable measurement, which simply was not possible. Now, however, machine learning and big data enable large volumes of text data to be mined and reveal, for example, employee sentiments or the strength of an organisation’s culture.
Two leading US universities collaborated on an analysis of a large number of internal e-mails at a tech company. They focused on how often workers used singular pronouns as opposed to plural forms. That reflected the level of how strongly they were tethered with the company and its culture. On this basis, predictions can be made about future resignations, according to an article on the website of the INSEAD business school.
Another piece of research focused on the language used in shareholder letters from banks in Europe as a basis for assumptions about the overall corporate culture and likelihood of risky behaviour.
External analysts can also use data from Glassdoor, where many employees rate and review their companies, to predict whether a possible merger will succeed. They examine especially the cultural alignment of the the entities by means of the following indicators:
Topic modelling (revealing common themes)
Sentiment analysis (positivity or negativity of the text)
Traditional keyword methods
Changes in corporate culture over time can be tracked
Will these new methods outperform traditional surveys and interviews with employees? Possible biases occur in every kind of data. Thus the decisive factor will be the reliability of predicted outcomes.
-jk-