We need to talk about Stanley

Last month it emerged that Stanley Johnson, father of PM Boris Johnson, had gone to his holiday home in Greece. In normal times this would not be newsworthy but these are not normal times. At the point Stanley went to ‘Covid-19 proof’ his property, only essential international travel was allowed from the UK. Further, he went via Bulgaria to get round Greece’s ban on direct flights from the UK. Stanley's trip followed hot on the heels of Dominic Cummings's jaunt to Barnard Castle in County Durham to test his eyesight.


Predictably, Twitter went into meltdown. But was this just the ‘triggered’ venting their spleen at Johnson (Stanley, Boris, tick all that apply) or is there something more nuanced that a deeper analysis of the tweets could reveal? And what does the analysis tell us about the application of text analysis and machine learning in research for brands? 



We analysed over 45,000 tweets using natural language processing and machine learning to undertake sentiment analysis and derive a list of topics discussed.



The analysis generated a wide range of topics but we were able to sort them into five main categories plus a long tail of ‘others’: The main categories are:



 

• Undermining government advice (33% of all tweets). This is where most of the outrage happens. Tweets here reference travel restrictions both in the UK and Greece. The sentiment on Twitter was strong: Stanley’s actions undermines government guidance that only essential travel outside the UK was permitted. It also pointed out that Greece itself had restrictions on people entering from the UK and that Stanley had flown in from Bulgaria to get round these.

 

• Abuse of power (21%). A common theme is the view that there is one rule for the elite and another for everyone else. Tweets relate to how it is fine for the elite to break the rules but everyone else is expected to follow them. Interestingly, there is a strong association made between the actions of Stanley Johnson and those of Dominic Cummings and his trip to Barnard Castle.

 

• Descriptors (16%). These are tweets that set the scene and establish the events. Often these are tweets containing links to news reports and those establishing that Stanley Johnson went to Greece and that he is Boris Johnson’s father. 

 

• Message to the UK public (10%). This category of tweets asks what sort of message is it sending out to the public? Points were raised about the need for everyone to work together to contain Covid-19 by sticking to the guidance. People are giving up on their freedoms but the message seems to be that it is now OK to do what you want. 

 

• Brexit hypocrisy (4%). This category is about what is seen as hypocrisy among Brexiters like Nigel Farage and Nigel Lawson who have applied for EU passports for themselves or their families. Stanley Johnson has applied for a French passport and tweets were quick to label him a hypocrite despite it being well-known that he voted Remain in the EU referendum.

 



It’s quite clear then that the main take away is that Stanley Johnson’s actions were perceived to undermine government advice and were part of a pattern of behaviour of high profile figures not following the rules.



But the analysis also shows a great deal of concern among the public that ignoring safety advice might become the norm and this has profound implications for dealing with Covid-19 over the longer term. 



What then are the lessons for the government on Twitter’s reaction to the Stanley Johnson event? The analysis highlights the importance of a clear message being consistently interpreted and universally applied. Actions that dilute the message risk undermining public trust in the message giver and ultimately the public’s compliance with it. We saw this with Dominic Cummings and we see it again with Stanley Johnson. Research published in the Lancet found that Cummings's trip to Barnard Castle may have reduced compliance with lockdown. The same is true of Stanley Johnson's trip to Greece.



This analysis has been about Stanley Johnson and the clear lessons around compliance. But the analysis also reveals some wider lessons for brands in the use of text analysis and machine learning.  We have identified five that are relevant:



1. Social media along with ratings and review sites represent a huge pool of voice of the customer data. It can tell a brand what people like about their product or experience, what they don’t like and how it compares with rivals. It can identify what drives purchases and what puts people off. Why would a brand not wish to plug into this real time feedback?
2. People often get things wrong. We saw this quite clearly in the analysis as Stanley Johnson, a well-known Remainer, was labelled a hypocritical Brexiter. Text analysis needs someone who knows what they are doing to quality control the data.
3. Category damage can impact reputation. People are quick to make associations and sometimes these can work against a brand. Stanley Johnson and Dominic Cummings is an extreme example but a bad player in the category can tarnish the reputation of others. Monitoring your brand on social media is vital to managing your reputation.
4. Machines are not good at sarcasm and nuance. Machine learning helps make sense of big data but there are some things that cannot be easily learned. It needs a human.
5. Text analysis is very cost effective. Processing the insight contained in 45,000 tweets can be undertaken in a matter of hours rather than the days and sometimes weeks that most surveys take. The answers are already out there. Often there is no need for new primary research.



Text analysis is a really powerful tool. We used it to analyse tweets but it can be applied to any dataset that contains large quantities of unstructured text (such as customer experience data, ratings and review sites, blogs and so on). The insight it provides is maximised when used in combination with expert human decoding of the data.



That's why we call it Human Made Machine Learning.



Michael Wagstaff is a Partner and William Wagstaff Principal Programmer at SPARK Partnership.