It’s been around 19 months since GDPR was introduced and the world of personal data was given its biggest shake-up since my analytics career began in 2005. This was a moment that many of us had been talking about for months in advance – yet, come 25th May 2018, we were still buzzing with nervousness.
I’m not convinced much has changed since.
As a data scientist working with consumer brands (and therefore knee-deep in their customer datasets), I still get questions on what information one is allowed to collect, store and use when it comes to personal data. While I am in no way legally qualified enough to give the definitive answer I am mathematically qualified to extract valuable insight from this data, and that’s what I’d like to focus on now.
Data is richer
One thing that is for certain is that scrutiny is getting tighter; the fines handed to both Marriott and British Airways in the UK earlier this year are testament to that. The expectation is that we should all be more knowledgeable in how we can use personal data held within our databases. In turn, this has promoted fear of using personal data and a perception that the quality of customer information has declined as a result.
Is this true though? In my opinion, no. Indeed, I’d say quite the opposite – customer datasets are getting stronger and are therefore deep in rich, actionable insight that can help you personalise the experience for your customers.
Data is more relevent
So why do I believe that datasets are getting stronger? It all comes down to consent. In the pre-GDPR world, we were probably all guilty of being lazy with our customer data. It didn’t really matter if we never deleted records from our databases once customers lapsed. It didn’t really matter if some of the data on our current customers was slightly out of date. GDPR has changed this.
Customers must opt-in to us storing and using their personal data and we have a much shorter time span in which we can do this. Yes, this does mean our databases are smaller, but it also means they are more relevant for analysis.
We no longer need to get excited about generating ground-breaking insight, only to be told that the information on half of the customers is out of date – or worse still - that they ‘lapsed’ six months ago. Instead, we start with a dataset of records rich in the most up-to-date behavioural and transactional information on our customers.
This is also a bit of a data scientist’s dream as it cuts out a lot of the “data preparation”; a task many of us hate as we’d rather be knee-deep in all sorts of crazy mathematical equations that I won’t go into today. What I do want to touch on is the output of those equations and the valuable insight we can get from the data.
Data is powerful
You have probably come across the phrase ‘single customer view’ at some point in the past year. We all want a simple and digestible view of who our customer is, how they behave (i.e. transact), and how we find more of them. A lot of these questions can be answered using traditional analysis, looking at each attribute to understand how your customers are distributed across them for example: the ratio of male customers to female customers, or how many customers have spent more or less than a given amount over the past year.
This leads to the creation of customer segments within your database – smaller pockets of individuals that display similar characteristics and can therefore be targeted in a similar way.
Rich; GDPR compliant behavioural data allows us to take this a step further, and provide a more personalised experience for our customers through machine learning modeling to predict what they are likely to do next.
In simple terms, machine learning means pattern recognition – and as we are all creatures of habit, there really is no better place for a pattern than in behavioural data. The model looks at all the behavioural information we have given it, finds any patterns and then uses these patterns to make predictions about new pieces of data we feed it.
For example, if we build a model that analyses transactional patterns, we can use it to tell us who is likely to purchase this month, even down to a product level if we desire, and then target them with the right message to ensure we secure that purchase. This effectively puts everyone on their own journey – only contacting them when they are likely to be “in market”, using the content most suited towards their preferences.
Of course, with power comes complexity – these aren’t simple models to build or implement, but our team at RocketMill are experienced at developing these for brands and I can assure you that the enhancement to the customer journey far outweighs the effort involved in creating these models.
I know you’re all thinking, ‘but we could have done this before GDPR anyway’, and of course you are right. A model is however only as good as the data you feed it – and as we’ve seen before GDPR, that data could have been incomplete or out-of-date, thus giving you some very different and misleading conclusions.
For all its uncertainties, GDPR really has helped create richer and more relevant datasets – and with this comes the power to understand your customer and personalise their experience.
This is crucial to total digital performance. As people, our attention spans are shorter than ever and this means we expect content to be tailored to our needs – and if not, we’ll happily switch our attention elsewhere, most likely to a brand that have used their data more intelligently.
Jeremy Horne, head of data science, RocketMill