20 december 2019
Data seems the ultimate solution to make the right, unbiased, objective decision. But even a solution can become a problem if you do it wrong. So, what do you have to take into account when using data? Gero Pickert, VP HR Operations at Nokia, gives you the answer.
Enter vs. exit
Centralising all data is challenging in many ways. The first challenge is the decision itself to centralise data. Many companies see the process of centralising data as imposing central governance. So, to counter this feeling, human resources should establish rules. After all, data is not just about figures. It’s about people.
Human resources should decide: What can we analyse? How detailed can our analysis be? How do we protect everyone’s privacy? Who can access our data? And how can we fend off the wrong interest? You want everybody to share their data, but you can’t share access with everybody.
Now vs. later
Stakeholders expect an analysis of data at the push of the button and while the idea is enticing, it’s only that: just an idea. You have to convince your stakeholders you can’t get the results right away and won’t get there without problems.
‘Usually, the analytics part is not the issue, the data part is the issue. It’s cleaning up the lack of discipline of the last 40 years’, says Gero. You need time to collect, centralise and clean data. Plus: you need to answer these questions:
- Who will ask questions?
- Who gets access to what?
- What is your ambition level?
- How do you set up the system?
The right team to answer these questions is not an HR team. It’s a mixed team of 5 to 10 people. HR has to ask the right questions and will achieve greater success with the help of experts who analyse the data. You better think twice: ‘if you don’t target that ambition level correctly in the beginning, if you create the database too small, you can almost not recover from that later’, says Gero.
Quantity vs. quality
Data quality. Data quality. Data quality. Gero repeats it three times because the quality of data is so important. So how do you know your data is qualitative? You can have all data, centralise it, run some tests and the outcome can still be wrong.
That’s because you can still see inconsistencies that you didn’t see during the testing phase. Writing one word in different ways can already make it hard to analyse data correctly. Imagine, for example, working in an international company and collecting data from people with a different mother tongue. Let them write the name of the city they’re working in. Will they enter ‘Munich’ as ‘Munich’, ‘Münich’ or ‘MUNICH’ in the system?
Gero’s advice: ‘You could test it forever in test cases but using it is the safest way of finding out quickly if you get ideate outcome. This highlights a dark aspect of data: you never know if the outcome of your analytics is true. Because there is no reference point.’ A big data solution for HR is a promise of paradise with problems.