Working in a Big data project in Marketing is constantly surprising. Essentially.
When I first got involved in a big data project, it was in marketing. I knew marketers had struggled for ages to really understand who their customers were in a holistic way. They had tried to reach them across several channels. They had built datalakes. They had implemented CRM and data analytics solutions. They had communicated through social media and new digital channels.
I began to realise that this would be a “do and learn” approach.
Thousands of articles and posts have been written by world-class experts talking about the high-level business drivers, potential use cases and benefits of big data. Vendors have been pushing the technology stack at the same time. It is clear that this hype really distracts us from understanding the business value we can achieve from it.
And it is true that the term big data engenders confusion, conveys all sorts of concepts, leads to multiple understandings on the potential value of the information, the mechanics to put in place, and overall what it means for an organisation.
I had read that most firms had to overcome specific challenges in their journey. They had to move tactically up to a strategic level in the search for use cases. They had to put the operational models in place. They had to build the required technical and business foundational capabilities.
“Big data is not a new thing” I remembered. It had been a reality for organisations for a good number of years. Of course, everyone knows Google, Amazon or Facebook. But there had been experimentations, pilots, and solutions already in operations closer to us in the Nordics. I recall presenting big data concepts to clients when I was at IBM already in 2014 for instance.
The project I was joining was about marketing campaign effectiveness. The bank was achieving very low results from doing some kind of generic marketing campaigns. Data came from different sources and the bank lacked a cleansed and reconciled view of the customer. Data was not accurate either, but it was the best they had. They had data analytical tools in place that analysts could use. Campaign managers could ask those analysts for lists, so that they could send direct mail campaigns. For every campaign, they had to ask other units.
Interestingly, the bank had been capturing customer comments and complaints for years. This information was written in free form and usually contained spelling errors, acronyms, and partial words. The efforts to extract meaning from it were largely manual and had not resulted in higher customer satisfaction or reduced churn.
We believed that this would help the bank gain a greater insight into the customers’ needs and create more timely and targeted offers for them.
For example, the solution would identify a potential high-value customer, based on the analysis of his or her life events, transactions, and interactions. As a result, the customer would receive a personalized offer either through email or upon the next interaction at a branch, in the internet bank, in the mobile banking app, or at the call centre.
Our first objective was to create a single view of the customer that brought together insights from all the different parts of the organisation. An important step I have to say, considering that when you have a 360-view in place, it is far easier for an organization to be agile, pivot, and make changes. You have to work across siloes and systems as well.
Technical questions quickly arose. Questions such as managing the high volume of data or how to extract knowledge. It is usual in that type of project.
But I found different types of challenges. What was important to me was not the technologies themselves. The real focus for me was how to leverage the potential of big data to get business value.
I began asking questions. Not necessarily easy questions, but important ones.
This first question soon came to the table. After the initial excitation, we understood quickly during the first work sessions that coexisted several understandings of big data, all depending on the participants.
In most cases, people thought about the overwhelming wave of data collected across social media, blogs, and the like. Big data was synonym with Twitter, Facebook or YouTube. It could be anything, huge quantities of data coming from the outside world. A jungle. It was quite difficult for them to make sense of the potential use cases beyond the hype.
Few understood the subtleties of structured, semi-structured, and unstructured data. This is technical.
Others commented on how social media could be analysed to understand customers’ sentiments and intent to buy. Of course, one participant mentioned “big brother”. This was expected too.
Hardly any really knew that almost 80% of all internal data in the bank was unstructured: Letters, emails, recordings of conversations in call centers, use of a website, old archived contracts, and the like.
Briefly, most participants lacked the knowledge and experience to understand the potential value of their data. They recognized it was important for them, but most just did not know what to do with it. Indeed, with big data, marketing units are moving from having too little information to potentially having too much data arriving from too many sources.
Our participants started to ask lots of questions: What does that actually mean? What does it mean for the bank? How to use it to gain better insights into our customers? What are the foundations? What are the steps marketing needs to take in order to understand and leverage data? What happens when you start a big data project?
And while those questions are important, answering them the first time is always the hardest. But answering them is always worth the effort.
I knew, from that moment forward, that we had to align and establish a shared view of what big data actually means in the bank’s context. We had to align on the objectives and ambition level too. The danger here was that this limited understanding could give way to the absence of common strategy across the bank. It would result in plans being developed at tactical level only, within specific units. It would also probably lead to an over-estimation of the risks.
In order to obtain that mutual vision, it was important to remove the hype around big data.
Intentionally, we formulated questions to help foster the discussion and debate. We started to create and communicate shared terminologies in order to allow the project team to work on the same foundational basis. We then ensured the alignment of both teams, objectives and metrics.
This lead to the second question.
“What is in all that big data?” We could probably answer it by saying “everything”. Potentially. At least theoretically.
After aligning the stakeholders on a common definition, we moved forward to the question of where to focus.
This part requires a good understanding of the business needs, on many levels, but cannot be avoided even in an agile world. If we were going to translate data into business value, we had to start by understanding the needs, and by analysing the data and data flows available.
The problem is often that data is growing faster than organisations know what to do with it. Our bank was not different and it was extremely difficult for them to make sense of all those data, not to say where to invest in the future.
Our initial effort was to focus on internal data in the systems. The objectives were to integrate the structured data existing in the CRM system with the unstructured data – customer comments and complaints – the bank collected across the years.
Recommended for youExploring Customer Experience and Business Agility with PostNord
First, there was a complex IT landscape and a lack of data integration with siloes and isolated systems involved. This has been a recurrent challenge. No big surprise though.
The bank, as many others, was organised around product lines, each of them owning part of the bank’s data and having an incomplete view of the customer. Sometimes it was structured data, sometimes, it was unstructured data. We found redundancies, errors, and duplicative workflows. As an example, customer service basically lived on an island unto itself.
Secondly, we all know that data quality is crucial. It is a pre-requisite in all projects. Marketing units cannot work on “bad” data. This fostered the question on whether we should focus on “big” data or rather on the “right” data. It also leads to discussions on data quality, reliability and integrity: How many data that is collected is relevant or not valid? How can we keep a data updated and protect its accuracy when it is multi-sourced and in different formats?
Third, it is true to say that quantifying and analysing unstructured data is not straight forward and easy. We experienced content could be ambiguous. Sometimes, customers may not use the proper company name, sometimes there may be a mix of languages or slang. How do we identify irony? negative or positive comments? How do we link with the customer ID?
We did not quit on those challenges.
We decided to conduct a baseline assessment. We wanted to explore the data, assess the feasibility, and understand the overall maturity level. This was done in a series of workshops and interviews. We first identified the current data and analytics capabilities within the bank, and determined the target state. We then performed a gap analysis, prioritizing and aligning with the bank’s business strategy and specific objectives.
The bank’s executives understood that data could no longer be maintained in siloes. It was obvious that the value came from connecting various datasets together, not from accessing new ones
I hope you enjoyed the article! Please leave me a review, I love reading them!