The Dashboard Nobody Asked For
A Case Study in Failed Data Projects
Projects can fail for a myriad of reasons, from lack of funding to unreasonable expectations. Now when we talk about data and analysis, it just gets even more complicated. Failure might get obfuscated by the analysis. Teams might implement a solution that wasn’t needed, and still think it was a success. Let’s explain this common scenario with an example.
Case Study
You work at a fortune 500 company. You just got hired for a data role (data scientist, data engineer, whatever just pick your poison). For your first project you meet up with the project manager and the marketing team.
The problem at hand is the following:
The company recently received a lot of negative attention because of a marketing campaign that was perceived as “insensitive”.
While discussing on how to solve the issue, you suggest using sentiment analysis. You’re not an expert, but you know enough about it and you’ve done some projects with it. Everyone else kind of understand what that is, and they seem to like the idea. At the very least, they understand that you can label comments with “positive” or “negative”.
You get a green light and the project starts. After some time, you finish your project. You even make a dashboard that shows the percentage of positive vs negative comments in real-time. Everybody loves it. A TV is added next to the marketing team with your dashboard on it.
After some time, the office is reorganized. The marketing team left behind the TV and you notice nobody actually cared about the dashboard. When investigating why, you find out that the marketing team wasn’t really OK with the analysis in the first place. They felt that you couldn’t really measure perception just with online interaction.
What went wrong
Apparently everything was going well. The problem was that nobody ever dared to question if the project was important. Everyone quickly moved to see if things could be done and how, but never if they should. Everyone quickly glossed over who would be directly affected by the project. The basic perception was that it was for the benefit of the company (since it was in damage control mode). Also, nobody questioned how to use the results. It just sounded cool getting a sentiment analysis, but not how to make use of it.
Problems like these can be easily avoided when you know how to catch them early on. Usually data professionals ask 5 key questions that can help identify red flags in a project. To learn more about these key 5 questions, check out my other article:
