Data Quality Assurance the right way
Inventor Henry Ford, also done Data Quality Assurance, and he once said, “Quality means doing it right even when no one is looking.”
Data Quality Assurance becomes every day more important. Everyone knows the feeling of someone doubting the numbers they’re presenting in a meeting. On the screen is a nice graph showing growth or, less nice, decline and then someone asks, “Are you sure about the numbers? Did you check them?” Every time I hear this question, my stomach moves into high alert.
The steps to a perfect QA
To help ensure that this never happens to you, here are seven steps to collect better data collection and improve the communication inside the team.
- Define your data funnel: Find out where the data is coming from and how it’s gathered, and construct a “data-flow map” showing the path of the data, from the first download until it’s displayed on the screen.
- Define the critical parts: Of all the data sources you’re importing from, define which sources are critical for the day-to-day work. Which sources can you not do without?
- Define important measures: As I often say, less is more. Define which KPIs are important on a daily, weekly, and monthly basis. QA them to make sure the data is correct.
- Create fallbacks: What happens if your critical measures or data sources are broken? Define the steps you’d need to take.
- Communication, communication, communication: Needless to say, communication is the most important element in a successful data-collecting process process. Keep your stakeholders in the loop, and if an issue arises, inform them as soon as possible.
- Have an auto-notification system: In case an Issue arises, make sure you have an auto-notification sent to your stakeholders, so they’ll know not to trust the numbers. It’s better to say the numbers are wrong than to have someone ask you if the numbers are correct.
- Keep learning: Data can pose endless challenges. Don’t let it get you down. Keep learning and improving. Think about new ways to spot issues and resolve them. Machine learning — an algorithm that learns about your data and improves your work — may be helpful.
Takeaway: The key to building trust in data is creating a map of the data-flow process, understanding the issues that may arise, and communicating promptly to stakeholders if an issue is spotted.