“Can I just spare a few moments of your time?” – People conducting surveys, everywhere <2016
I was on the tube in London last week when a lady wearing a fluorescent safety vest and carrying a clipboard stopped me. “I’m conducting a customer survey for Transport for London (TfL)” she announced; “Can I just spare a few moments of your time”?
I consented willingly (those of you who commute daily on mass-transit systems will know how refreshing it is to make conversation on the commute), and braced myself for the inevitable:
“On a scale of one to five, where one is very dissatisfied and five is very satisfied, how would you rate your journey today…”
The problem with this is now I have to try to compute my level of commuter satisfaction in real time. There are two ways of doing this. Firstly, you could either work out the decimal level of your satisfaction, then translate this to the arbitrary labels on the scale. Or simply you can do what I did and smile pleasantly and answer “4” to all the questions (it isn’t very British to be too enthusiastic about these things, after all).
Behind the scenes, someone is feeding the multiple choice answers into a scanner which in turn is OCR-ing the data into a machine readable format. It’s then getting aggregated, perhaps adding a few extra dimensions such as gender, age, time of day etc. in order to provide senior management insights into the minds of their passengers cut by their respective groupings.
On exiting the station you might pass by a notice that triumphantly proclaims “94% of passengers are either satisfied or very satisfied with their experience with TfL” – confidently capturing the collective consciousness of countless passengers over the past few months.
To me, this exercise is pointless – the aggregate of all the “4” answers such as mine is no more an indication of our level of satisfaction than an indication of our level of apathy towards the process. I bet you do the same too, perhaps marking the mid-point or the extremes exclusively when filling in surveys enticed by the prospect of winning a prize or some other incentive. The fact is, that none of us care to work out our true feelings, we know too well that only a machine will ever read the answer – even the poor lady with the fluorescent safety vest just wants to complete this task and make her own way home.
Thankfully, recent technology advancements provide us with a better solution – but it amazes me how little this has been adopted so far. Like all the best technologies, it transforms a process back from its cold, sterile, tyrannical form (as in the 1-to-5 survey scale) to something much more natural; something much more human.
Imagine you are a market stall trader and you want to figure out how to improve your offering. What would you do? Well, you certainly wouldn’t conduct a survey like this, for you wouldn’t need to. You would simply muse over the past few conversations you had with clients, perhaps over the last few weeks or months, and then try to pull together the themes from the feedback.
In the days where machines couldn’t process language with ease, it perhaps made perfect sense to ask humans to convert our feelings into a decimal scale so the machines can more easily aggregate our inputs. A better way is to simply capture the feelings of customers, clients, passengers, staff, whoever in their own words. Without getting technical (there are plenty of other people who can do this much better than I can) the system will then capture these words and group them into words that are similar, resounding themes will be shouting loudest – occupying the biggest ‘bubbles’. The solution works in the same way as the mind of the market stall trader – with the exception that it can be scaled, almost infinitely.
A word bubble chart, such as this, allows words to be grouped into themes – revealing the patterns of intent from their sources Image source
Companies like TextSort take this one step further. At TextSort they have developed a solution that can ‘compress’ natural language into something shorter using a simple User Interface such as a slider. For instance, you could load in the entire works of Shakespeare and ‘slide’ it down to just 5 words. You would probably end up with “Comedy, Tragedy, Love, Betrayal, and History”. What TextSort is doing is looking at individual words and their pattern of usage throughout the documents and then comparing them to phrases, sentences, paragraphs, chapters, acts etc. The ‘theme’ of a particular Act can therefore be easily captured. What this could be useful for is automatic categorisation of the written word, but the team at TextSort believe their greatest traction with early adopters will be in the Fraud identification space, particularly in Financial Services where the surveillance of keywords and themes (and in particular spotting outliers, as in words or themes that stand out from the ordinary) is essential.
I believe though that this technology should first be deployed in surveys – and save us all from their tyranny. For everyone concerned, the experience would be better and the insights would be truer.
So, in your own words, how would you rate your experience today…?
