Broadband sentiment analysis used to examine broadband providers
When you browse an ISP website looking at the packages they have on offer it is really difficult to decide how to choose. By and large they are all very similar. Some may offer different TV bundles and you occasionally see the occasional high street store voucher thrown in as an incentive to sign up. Been trawling through various ISP streams on twitter doing a bit of broadband sentiment analysis. I though this might throw up some real world feedback on specific ISPs that might help people decide on the right one for them.
What came out was quite revealing and makes me glad I no longer work for a broadband service provider. The amount of vitriol that gets heaped on ISPs when their service goes wrong is amazing. It’s no surprise. The same probably happens when there is an electrical outage. People now rely on their internet connection as an utility.
In doing the work it is worth noting that an automated sentiment analysis tool isn’t perfect because computers can’t understand the nuances of human language. Sarcasm for example is very difficult to get right. eg
So and so is a great provider
versus
So and so is a great provider !!!???
Same words but the second would go down as negative sentiment if judged by a human. Because of this some human checking has to be performed. This human checking has brought out some interesting anomalies.
For example this tweet:
@drdeakin: Most reliable network @EE ? Rubbish! 4 mobile contracts plus mobile broadband each month and was about to add a business mobi…
was retweeted 112 times (at the time of writing). I found this curious so looked up @drdeakin. He has 177k followers. No wonder he was getting so many retweets. I also wondered how he got hold of so many followers whilst only following 412 accounts. Was he a celebrity? Turns out he just got married to the mother of someone in “One Direction” – a popular music group, apparently 🙂 (@JohannahDarling with 1.15M followers). On this basis I didn’t consider it fair to apportion negative sentiment to tweets other than the first, although a few did get through early on.
Other tweets were showing positive sentiment but clearly posted by someone with a vested interest. These were discounted (eg “@Exposure4All Get 152Mb broadband, 260+ TV channels & unlimited anytime calls to UK landlines → http://t.co/vYY5LuJFUN http://t.co/Br52f3djA0″ is on the face of it just a sales pitch)
One provider in particular, Plusnet, looked like drowning in complaints. Plusnet suffered a major outage during the window in which I was looking at the tweets. This was exacerbated by the fact that Apple had just released iOS8 and all the fanbois were at it in droves. As such Plusnet came out very badly compared with other ISPs. However this is a constantly changing data set. I know from experience that ISPs occasionally have problems that seem to the huge disasters at the time but they are overcome. A historical trend chart of broadband sentiment analysis should reveal who is the most reliable ISP overall.
ISPs use Twitter as a means of engaging with dissatisfied customers. Twitter is used basically as an alternative inbound means of communication. Some seem to handle it better than others.
These two examples illustrate how:
- @BTCare @someukbitch Happy to but need a better description of the problem, whats the problem and is the light on your Hub blue?
- @EE @RhodriOR Hi Rhordi, Afraid we can’t help with home broadband queries, Please call on on 0844 873 8586 from your … http://t.co/OCZ4Z8vDYs
In this case BT is doing a good job compared with EE who aren’t making it as easy as they could for their customers.
The one common thread that came out of the analysis was the number of times an engineer didn’t show up. People had usually taken days off work to wait in for the visit. This is pretty unacceptable but is unfortunately a situation that has prevailed for years now. Maintaining the copper broadband network is a nightmare.
I’ll be making the output of this broadband sentiment analysis available quite soon but thought some of my findings were interesting enough to publish beforehand.
