Oscar Wilde said “the only thing worse than being talked about is not being talked about” and everyone knows that age-old phrase “any publicity is good publicity”.

Seeing as tonight is the final of Celebrity Big Brother, we thought we’d use this live case study to test this theory using our social media data.

The hypothesis we’re working with is this: celebrities who are talked about with the least emotion (good or bad) would be the ones to leave the competition early.

The Methodology

We wrote some queries to bring back only high emotion words and phrases (e.g. love, “can’t stand” etc.). To ensure these terms reflected a personal opinion in the mention (rather than for example “Rylan hates Speidi”), first person pronouns were included in the query. Also included were terms specific to Big Brother, such as “get him/her/them out”.

Volume vs. Emotion
The first analysis we ran compared volume to emotion, to see if they’d give us a different ranking. For this we looked at the entire period that each housemate has spent in the house. For those who have been evicted already, we looked at the data for the period until the day before they left the house, to avoid the skew of any reaction to that eviction.

To make it comparable, volume is shown as average mentions per day spent in the house. The crosses denote the evicted housemates and which order they went out of the house.

We then looked at what percentage of that conversation contained strong negative or positive emotion.

Both the volume and the emotive words charts show more evicted contestants at the lower end of the ranking, with one anomalous result for each (Paula for volume, Sam for emotion).

However if we look at the order in which they were evicted, the emotive terms methodology was more accurate in its ranking. This doesn’t mean that volume of mentions is meaningless, just that it should not be considered without also investigating the tone and context of the conversation.

What happened to poor ol’ Sam?

Before we looked in any more detail at our remaining finalists, we were interested to understand what happened to poor old Sam.

Sam Robertson was the resident hunk; a closer inspection of his emotive mentions revealed that they related almost exclusively to his good looks. We also took a look at some more neutral comments about Sam in social media, which often refer to him as boring.

So it seems that a pretty face alone won’t take you to the final in reality TV, and some more work to take into consideration what kind of emotive language is used could reveal a more accurate prediction (we’ll leave that for another day, perhaps).

Using emotional conversation trends to make predictions

Tricia and Frankie were both evicted on Wednesday night, with Frankie’s eviction causing shock among the housemates. However, if the housemates had access to Brandwatch, they’d have been more surprised that Tricia was sent packing.

The chart above shows how the emotional conversation for our mid-runners was distributed over the past three weeks. The majority of the emotive chat about Frankie happened in week one and fell sharply in the last two weeks, supporting the hypothesis that a lack of emotional response from the audience could lead to eviction.

But according to the buzz, Tricia should have stayed in and Ryan (Toadfish from Neighbours) should have gone. Interestingly, it’s not just us who were surprised with this turn of events, as earlier this week the bookies slashed the odds on Tricia to win.

Choosing a winner

So, we can already see our theory isn’t perfect, but we’re still feeling fairly confident that we can pick tonight’s winner based on our data. If we were the betting type, we’d take a punt on Rylan (although we’d get pretty poor odds, as he’s also the bookies’ favourite to take the title).

So, why Rylan? Well, he has the highest volume of mentions, as well as the highest percentage of emotional chat, which has grown steadily across the three weeks of the competition.

In the past week, emotional discussions about Rylan were boosted by controversy around news that he has been allowed out of the house for X Factor rehearsals. However this does not seem to have hurt his net positive to negative ratio too badly and, according to the theory, any emotion is good emotion.

In Conclusion…
Believe it or not, we didn’t run this analysis because we’re desperate to know who will win Celebrity Big Brother. But this is just a funky way of showing how you can use social media data to predict and examine trends. The general public are a fickle lot, but analysing their conversation serves some meaningful insights into their opinions, habits and behaviour.

We hope this article has given you some ideas for ways to use some advanced queries and benchmarking/analysis techniques to make Brandwatch data interesting and actionable for your business, as well as giving you a tip on who to blow the last of your cash on at the betting shop before payday.