Conditional sentiment mapping of live rail services
Department For Transport demonstrator
Live service mapping
Live mapping of tweets to live running services – without #hashtags, keywords or @operator handles – means we’re connecting 3m tweets a month to the train or bus services they originate from.
Every tweet is processed using our natural language processing engine and scored for positive or negative sentiment. The scores and the original tweets are archived and made available in realtime
Everybody knows a late train is an unhappy train – so we statistically normalise the sentiment with over 50 other operational inputs to look for unpredicted anomalies in customer experience.
In order to understand the causes behind poor experience, we provided the DfT with a tool that reports delay and cancellation patterns against customer and operational conditions
Sentiment mapping is no longer a gimmick
Collaborating with the Customer Experience team at the Transport Systems Catapult, we wanted to see if it was possible to use basic social sentiment to diagnose operational issues and patterns. By using the Zipabout IM platform to map 3m tweets a month directly to the live running rail services that they come from, we started to build a picture of when people were happy – and when they really weren’t happy – on the rail network. Using the Zipabout transport data aggregation platform to normalise the data and learn from the patterns, we have started to reveal some interesting (and quite unexpected!) patterns that lead to unhappy customers. This led us to develop a customer service ‘health check’ for rail journeys in the UK, and to start working on a proactive intervention tool in partnership with Keolis and London Midland.
As the programme evolves we have begun working with Transport Focus to continuously refine our customer experience tools. The Transport Systems Catapult will be publishing the findings in an ongoing report in 2016, which will be linked from here when available.
Live operational overview
Every train in the country. Live.
Every tweet from every train and station in the country. Live.
Average conditional performance for every train in the country. Live.
2 years of minute by minute data, visible in real-time.
Beats following operators on twitter and counting hashtags doesn’t it?
With over 2 years of operational and conditional data, the DfT reporting tools allow the user to analyse any combination of customer experience, delay or cancellation attribution, and weather conditions in a simple interface. Anomalies and hotspots soon become apparent.