'Whatever Our Product Is, It Must Be Built On Trust And Accuracy.' How Does NASA Divide Good From Bad?

Stephen Chesley, NASA's Senior Workforce Planning Specialist will help us to understand - is there such thing as good or bad data?


NASA is one of the largest organizations in the world, with over 50,000 people working across their centres and facilities. This huge workforce appears in forms of on and near-site contractors, federal employees, as well as commercial and academic partners.

They have four main mission areas: Human Exploration and Operations, Science, Aeronautics Research, and Space Technology. Considering the size of the organization, its not-for-profit nature, and the fact that it's funded by Congress and the Whitehouse, the workforce analytics department deals with a lot of sensitive details and data when setting up workforce plans.

Stephen Chesley, NASA's Senior Workforce Planning Specialist who spoke at the HR & Analytics Summit in Chicago in 2015, says that similar to other large organizations, NASA has their personnel data warehouse, where personal data is collected, stored and processed, and where the entire system gets updated every two weeks. Within the warehouse, there are things like the recruiting database, learning management database, federal personnel payroll system, and other databases that somehow touch on their people. Stephen has spent considerable time searching for an answer to what elements can shape good data in workforce analytics, and came up with some interesting findings.

Firstly, he believes that in order to understand good data, you need to have context:

  • Does this data make sense?
  • Does it agree with other data?
  • What is the source of it?
  • Can all types of data be good data?
  • Is it important for data to be replicable?
  • Do you have a large sample size?

Answers to those questions can lead organizations to very different conclusions and viewpoints, where there is no such thing as right or wrong, the information reveals more about what exactly you may want to achieve from data.

During Stephen's crowdsourcing experiment, where people were asked 'why do we need good data?', the answers were mostly given in contrast to 'bad data': Bad data can lead to inconsistent reporting or analysis, unlike good data which is reliable. Bad data can take more time to make a use of and understand, whereas, with good data, the analysis is organic. And finally, bad data can cause mistrust. However, there was never a universal definition of what good or bad data is.

People may have different interpretations of data, this is because it can be of different types and use points, depending on each case. The critical factor is, Stephen says: 'Whatever our product is, it must be built on trust and accuracy.' And if ever analytics experts cannot deliver these two basic but very important principles, they may end in being out of the business.

More isn't necessarily better and it can be useful to define the scope and only use data that is needed. Additionally, not all data is equally available, but that's not the reason to start the analysis based on data which may not be enough to achieve the required level of accuracy. Stephen recommends waiting until requisite data emerges (as it always does with time), so the analysis can be completed. And also, he says, not every insight adds enough value to justify the resources needed to extract data, so think and plan thoroughly your data sources.

Companies can and should innovate with data. From NASA's experience, there is no such thing as one single method or approach to analyze data. Experimenting and testing different techniques is a good approach. But most importantly, we need to remember that data can give you 'an answer' but never 'the answer', so it can be worth trying to make a good forecast out of it, but these predictions will never be perfect. So, maybe it shows that it doesn’t matter how much data you have, it’s what you do with it that counts.

Big data hype small

Read next:

Is Big Data Still Overhyped?