In a connected and data-driven world, customer analytics practices have never held greater opportunities to improve customer experience (and, by extension, sales and profits) – but they’ve also never faced such a vast and bewildering set of challenges.
This balancing act was recently articulated in a McKinsey survey of over 260 US-based leaders in the world of CX. The survey found that while 93 per cent of respondents use survey-based metrics for measuring customer experience (CX), only 15 per cent were “fully satisfied” with that system – and, strikingly, just 6 per cent were confident that this kind of measurement system leads to strong decision making.
It’s not surprising that these respondents weren’t confident that surveys – which, according to McKinsey, typically sample a mere 7 per cent of a company’s customers – represent the best way to improve CX.
Surveys – as, perhaps, these respondents sense – simply can’t capture the nuances and quantities available via the enormous amounts of data amply supplied across today’s dizzying mix of social media platforms, emails, call audio recordings, chat messages, and so on.
On reflection, however, it’s not difficult to see why so many organisations might view this abundance of unstructured data (UD) in terms of its intimidating complexities rather than its exciting possibilities.
It’s vital that the world of customer experience management rises to the challenge of harnessing UD and overcomes the hurdle of collecting quality data from a variety of sources – the rewards for doing so are as limitless as the data itself.
Unstructured data – hidden in plain sight
Of course, it’s understandable for organisations to lean on reassuringly structured and limited survey data – after all, when smaller sample sizes are used, humans are able to grasp that data.
It’s also possible that the sheer vastness of UD makes it almost invisible – UD exists all around us to such a degree that we can’t always see the wood for the data-heavy trees.
Thankfully, however, we now have AI capable of providing us with the gist of enormous quantities of UD, allowing us to receive detailed and actionable insights without needing to read through millions of Tweets at a time.
This kind of AI is so necessary because UD refers to any kind of data which isn’t stored or arranged via any kind of organised model – you wouldn’t find UD in a table, for example.
This means, of course, that UD is everywhere. Not limited to text, UD can equally include images, audio, video, and – in essence – anything that produces any kind of information.
It’s easy to see how data in these quantities and manifestations could be disregarded or overlooked – especially when CX measurement is, as McKinsey indicates, so often associated with surveys, questionnaires, and feedback forms (whether they be physical or virtual).
Obviously, customer data collected through these means – no matter how sophisticated the technology involved – are mostly structured due to the nature of the formats used (barring the answers to open ended questions).
The problem here is that those listening anxiously to hear the voice of the customer (VoC) via VoC technologies or initiatives still rely on tiny samples of structured data, denying themselves the millions upon millions of data points (expressed in various languages) now available with the right kinds of customer analytics technology.
Quality data, superior insights
The McKinsey report notes – unsurprisingly, given the limited customer samples reported by the surveyed organisations – that “data quality” is a challenge for many companies. How, then, can properly harnessed UD be considered ‘quality’ data in the world of customer analytics?
“Quality”, like many terms in the world of data and analytics, is something of a buzzword – it means different things to different people.
For the purposes of this discussion, however, quality data can be understood as data that has precision, recall and F score over 80%, and is fit for purpose.
In customer analytics, fit-for-purpose data is accurately annotated data that offers an honest and accurate impression of how customers think and feel about the organisation in question.
While the aforementioned surveys, feedback forms, and questionnaires might appear to be the optimal way of procuring this kind of data, these limited options with small sample sizes represent neither a variety of sources nor high data quality – because there’s no reason to believe answers supplied by customers are sufficiently nuanced, honest, or reflective of their sentiments.
By leveraging UD, meanwhile, via the analysis of vast numbers of – for example – Tweets, Instagram posts, and verbal feedback from customer service calls, customer analysis will not only use multiple data sources, but that data will be far more fit for purpose – in that it will offer unsolicited feedback from each and every customer in a manner that intrinsically stunted questionnaires can’t reliably produce.
Consequently, using UD in customer analytics isn’t necessarily about achieving something beyond the usual raft of sentiment analysis, personalisation, lead scores, and other goals of customer analysis.
Rather, UD is of a far higher quality and exists in larger volumes, imbuing these use cases with a degree of accuracy and nuance that a survey just can’t match.
For those respondents to the McKinsey report, and the many organisations who share their challenges, this makes UD a hugely worthwhile investment.