Data Professionals
Design, build and deploy data models, analytic frameworks, data storage and transformation processes and systems, and apply analytical techniques and tools to gather, process, analyse, interpret and present data for strategic planning and decision-making.
The Monday morning dashboard
You are a data analyst at an online grocery business, responsible for the weekly numbers that managers use to make decisions. It is Monday, and the weekly report goes to the leadership team at ten. You have a pile of requests, one odd number you do not trust and not much time to sort it all out.
Choose what you would do. Then see what happens. There are often good reasons for each choice.
The new delivery route chart looks fantastic, it's exactly what I told the board would happen. Just leave it as is, please. I don't want any questions about the data before the meeting.
Leave it as it is, because it supports his plan and there is very little time left to check it before ten
The chart goes to the board. A week later, someone notices the lines are measuring different things. The error is now public, and the data team's reputation is damaged.
Check the chart first, show Hugo what you find and fix it together before the meeting
You find a mislabelled axis that exaggerated the result. Hugo is annoyed for a moment, then relieved. The corrected chart still supports his main point, and it holds up when questioned.
Tell him the chart is misleading and you will not send it until it is rebuilt
You may be right, but Hugo has no new chart for the meeting and feels ambushed. Being direct helps most when you also offer a way to fix it by the deadline.
I'm planning a mailout to customers who bought baby products. Could you send me their names, addresses and what they bought? It would save me heaps of time, and it's only for a good offer.
Send the full list with names and addresses, because she is a colleague, it is for a good purpose and it will save her hours of work
You have passed personal information to someone who may not be allowed to see it. Purchase histories can reveal sensitive things about people. Being helpful is not a reason to ignore privacy rules.
Check what customers agreed to, and offer an anonymous or grouped version, or route the request through the privacy process
Kayla finds out that only customers who agreed to marketing may be contacted. She gets what she can use, and the customers' details stay protected. Good data work includes knowing what not to share.
Say no, because you cannot share personal data, and leave it at that
You protect the data, but Kayla leaves without a solution. A short explanation of what could be done would have kept her on side and made a safe path possible.
Last week's report said late deliveries were falling, but my team is still getting loads of complaints. Are you sure the report is right? Or are we just measuring something different?
Check how lateness is defined and measured, compare it with the complaints, and tell Pat what you find
You find the report counts a delivery as on time if it arrives anywhere in a long window, while customers expect the time they chose. The answer is useful, though it would be even better with a suggestion for a better measure.
Defend the report, because the data is correct and complaints are only anecdotes
The data may be correct for what it measures, but Pat's team sees real customers. If you dismiss them, you will miss what the numbers hide, and trust in your work will fall.
Look at how lateness is defined, compare it with complaints, and propose a measure that matches what customers expect
You find the gap between the measure and the customer's experience. Pat's team and yours agree on a better measure. Numbers are only useful when they match reality.
How it went
The report goes out with numbers you can stand behind, a customer's data stays protected and a gap between data and reality is on its way to being fixed. Data work is about being curious, careful and honest, and then helping others see what the numbers really say. Does that sound like detective work you would enjoy?
This is a simplified, fictional situation to help you explore. It is not professional instruction or a test of your ability.
Case study: The chart that was too good to be true
A fictional case set in the data team of a fictional online grocery delivery business.
The situation
Tomasz was a data analyst at Basketful, an online grocery delivery business. The weekly dashboard showed that orders from a new suburb had doubled in a week, and the chief executive was already planning a big expansion on the strength of it. Tomasz had a nagging feeling about the number, because nothing had changed in that suburb. His report was due in an hour, and querying the figure would make him the person who spoiled the good news in front of the whole leadership team.
What they did
Tomasz traced the number back through each step of the pipeline. He found that a change to the way orders were recorded had caused some orders to be counted twice, and the doubling appeared only in the suburb where the new system had been tested. He fixed the report, kept a note of what had happened and wrote a short, plain explanation for the chief executive. He gave the real figure, which was a healthy rise, and suggested a check to stop the problem recurring.
How it turned out
The real growth was still good news, though much smaller. The expansion was planned more carefully, and the data team added an automatic check for double counting. The chief executive thanked Tomasz for finding it before the plans went further.
What would you do?
- Why did Tomasz feel uneasy about a number that looked so positive?
- How would you explain a mistake in the data to someone who was very excited?
- What would you check first if a number suddenly doubled?