Future data
Posed with the question, ‘what do we do when bad data things happen to good data tech?’, one DataIQ member from the shipping industry discussed a bad data event from their previous employment in air transport. The organisation was going to launch a new product – a sales platform – but “they were so focused on the launch and the engineering of it, that they did not start thinking about what information they wanted out of the products until after it had been launched.” This DataIQ member was the only analytical member of staff in the business at the time, which made the retroactive data work even more difficult to complete as those helping did not necessarily possess the relevant skills to understand the solution and why it was needed.
The member explained, “the head of finance and me spent a frantic two weeks scrambling around, trying to pull the data together the information on how the product was performing.” One data aspect needed was for management insights, but the second was for accounting requirements such as money coming in and out and data sets to meet the fundamental auditing obligations. “I think it was a naïve approach from senior management in order to just ‘get it out the door’,” said the member. “They completely missed thinking about what data they want and the format it needs to be in for both legal requirements and design requirements.”
Frustratingly, the member explained how a similar thing happened with a second product being launched. In this second example, it was the case that the way in which data was to be used was overlooked, meaning the data extracted from the product was stored in one long JSON file. This made using the data incredibly difficult and rendered much of it unusable. The member stated that, “my learnings from these experiences include making sure that you are aware of what is going on with the wider organisation, making sure you are aware of who is in the conversations that can include data and then inserting yourself into these conversations when required. One of the biggest learnings is making sure that people know what it is that they want out products and providing that voice of reason from a data perspective.”
It was highlighted by the members of the roundtable that data teams are frequently overlooked because they can be hidden behind layers of hierarchy and find themselves disconnected from wider conversations. To combat this, the members agreed that there needs to be a constant feedback loop with stakeholders as well as the ability for data leaders to be assertive within an organisation to make their needs and concerns heard.
Clarity and cleanliness
Data can only be as good as those inputting it and designing the ways for it to be analysed – a weak link can jeopardise the whole data chain. A roundtable participant from the banking sector recounted an issue they had faced where the wrong value field had been selected. With more than 13 billion transactions being processed each day by this banking powerhouse, there is a large opportunity for errors to occur, so clarity and understandings of data needs are essential. In this example, simply, one value field was selected when there were in fact eight different ones needed. This issue meant that there was no way of differentiating between all eight of the values and that it no longer met the regulatory requirements needed.
How did this occur? Once analysts had examined the situation, it was concluded that the error occurred because of an absence of definition for the term when the value was selected. Certain terms were defined, but over half were not. The member said, “it is therefore unsurprising that an error occurred, and had it not, it would have been down to sheer good luck.” This error shows the importance of understanding that successful data is not down to the technology being used, but rather a fundamental understanding of what data can be produced and what this data represents, or else you can receive outputs that that do not reflect the perception.
“It highlighted two things: Number one, the need to have exceptional hygiene around the way that we define data and make it visible,” said the member. “And secondly, if it was the wrong value, why didn’t the controls pick it up? Were there any controls there? You cannot assume that just because it comes out of the system and looks okay at first glance that it is correct.” Assumptions are often an issue faced by data and analytics teams, particularly in time-sensitive environments when trying to get a new product to market, but the importance for clarity and compliance is pivotal when dealing with tightly regulated areas.
Furthermore, a member from the cargo community shared their experience where being unable to visually understand data can cause problems. The situation arose when dashboards monitoring the movement of data were providing a negative number, implying that cargo movements had begun moving backwards in time. Of course, this is not possible and was clearly obvious to the data team, but seeing a negative number is common in business and being unable to understand the true meaning of this data is where problems start arising. The member said it took nearly a month for the problem to be rectified with a bottleneck being the case that data was not taken as seriously as other areas and there was a lack of data literacy outside the data team to understand why this was an issue. Further down the line, having incorrect data such as this example can render datasets unusable, so it is important that each number is scrutinised.
These examples reaffirm the need for data professionals to make their voices heard and insert themselves into conversations to make sure that checks take place, and that the way data needs to be used is understood to a certain level by the wider organisation. Reliance on technology is not suitable for an effective and efficient data team, but it should be recognised that the use of technology can enhance the quality of the data exponentially when being designed and used correctly.
Optimists reign supreme
Every participant in the roundtable stated that they were an optimist, which is a very encouraging sign for the wider feeling of data’s place within an organisation. “If something goes wrong, I use it as a learning experience,” said one participant from the cargo industry. “I make sure I have things in place to try and avoid it happening again, but I am always confident that we learn, we iterate, we build on what we have got and we can always make it work.” This sentiment was repeated by each member.
Although each member said that they expected to have things go wrong with future projects, their knowledge that these things do happen to their peers is reassuring because they are not alone. Additionally, everyone agreed that having bad things happen is often a benefit in disguise as it enables teams to perfect their craft, improve their testing procedures and learn from a variety of often unexpected experiences.
Make sure you sign up to the upcoming DataIQ roundtables to get involved in peer-to-peer conversations that will improve your data journey and explore our previous roundtable writeups.
