Do Singaporeans hate Tesla?

According to an analysis done by Rave Reviews, apparently so. 

The site recently studied an amassment of tweets mentioning a number of large global names, using the percentage of negative ones to calculate a "hate rate". Tesla came out tops in Singapore, with 58.3%. 

But we're not the only place where Tesla is apparently the most hated brand - it's also Number One for the wrong reason in Taiwan, Croatia, Poland, Portugal, Serbia and Spain, where it has a whopping 75% hate rate. Compare that with Taiwan's - at a meagre 14.5%. Imagine having only 14.5% of tweets skewing towards negative, and already being the most hated brand! Taiwanese Twitter users must be quite the amiable and forgiving bunch. 

But of course, with every data set, caution must be exercised.

First of all, Singapore's Twitterverse is not terribly representative of the entire country, probably skewing towards a younger demographic. We're not sure which way the hate rate would swing if we were to take all 5.7million of us into account (if we had added our ministers in back in 2019 though, the number would probably creep above the current 58.3%). 

Rave was also very transparent about the fact that they had used a sentiment analytics tool, SentiStrength, to evaluate how Twitter users felt about particular global brands. But here's where things get a bit murky.

There are multiple ways in which SentiStrength takes apart any given statement. On the most uncomplicated level, it assigns a binary score of -1 (negative) or 1 (positive). A trinary score allows the possibility of 0 (a neutral zone). But it also has a couple of other measurements - dual and scale (we won't bore you with the mechanics) - to take into how account how negative or positive a statement is. 

For example, "I love Singapore's expensive cars lol" has a trinary score of 1. But under a dual approach, the statement has a positive strength of 3 (from 1 to 5, 5 being very positive) and a negative strength of -2 (from -1 to -5, -5 being very negative). 

See where the problem lies already? SentiStrength is an undeniably formidable and amazing tool, but it's not 100 percent accurate at picking out nuances (sarcasm, for example). Rave states that it considered tweets with a negative component from -2 to -5 as negative, meaning that it actually included quite a huge range of negative sentiments.

The other gaping flaw it has is that it's trained most strongly in English. We tried statements like "Walao the weather sibeh hot" and "istg Teslas rly cmi leh" and they both returned a binary result of 1 and a trinary result of 0 (with dual, the result was effectively neutral as well). Maybe both Singlish and Twitter slang sound very chirpy to SentiStrength. 

Finally, Tesla is still quite the youngling in Singapore, so we'd be very hesitant to cast any aspersions about how it's perceived here until it establishes a proper presence. A large chunk of the tweets analysed may have been skewed towards Tesla events in other foreign contexts, rather than reflecting sentiment on the ground -our ground - towards the Model 3s in Singapore, for example. We can't exactly work with what we haven't had. 

Of course, this is just an unserious attempt to be more critical about the data pooled together by the good folks at Rave. They've nonetheless gifted us an interesting insight that demonstrates the inescapability of Tesla as a brand these days.