If a D&A initiative fails to create the expected business value, what do you do? How do you frame these outcomes when presenting to the board?
Sort By:
Oldest
Chief Data Officer in Media6 months ago
Use a 3 communications objectives framework. 1) Here's what we expected to happen and what the actual results were. 2) Here's our analysis of why that happened. 3) Here's what we've done to ensure that challenge won't cause the same failure again.Most of all, be accountable for the previous outcomes and resolving the problem that caused it. No matter how mature the business and data team are, there will always be areas for improvement. Focus on finding and resolving issues faster, not building a perfect process.
President & Chief Data Officer in Services (non-Government)6 months ago
Ideally you will have first conducted a proof of concept. You should start with a POC first, followed by a pilot (A/B test in the wild if possible) and then go to production assuming it proves out. When you first propose a D&A project, you have to present it as an experiment where you are testing a hypothesis. There should be reasonable assumptions supporting the hypothesis. When you position D&A initiatives in this way, you are already setting the stage for the possibility that it won't create value. But, you are "failing fast" and not spending a significant amount of money to find out. If you took this approach, you just need to report the findings of your hypothesis test and have some reasonable explanations for why either your assumptions were not correct (e.g., the business landscape has changed) or why the model/algorithm did not perform as expected (e.g., data quality issues and/or certain features were not as predictive as you hypothesized). If you did not take the above described approach, then this discussion becomes a lot more difficult. I would still focus on assumptions and hypotheses, and you will need to have a logical and convincing justification for why they were wrong and why you wouldn't have necessarily known that in advance. Again, some reasons could be changes in the business landscape and/or data issues that were not known in advance.
I hope this helps!
Data Science & AI Expert in Miscellaneous6 months ago
It depends. The reason for failure could be a finding by itself. More often than not, wrong assumptions about your business causes that and it can be valuable to understand the reality.
A partial success could get you closer to better outcomes in the next step.
Last but not least, this is not unique to analytics projects and the answer to this question can be similar to any other project failure.
Senior Director D&A in Manufacturing6 months ago
I will go with the assumption there was a POC done before getting too far into it...always advised if possible... either way, sometimes things don't work out. A lot depends on the "why" you did not get the expected business value. No matter whether it is user adoption, change management, technical issues, bad assumptions on use cases, or whatever... focus on the learnings, what you will do differently next time, and hopefully you were able to build some kind of foundation that will help speed up delivery for future projects. The key is to be open and honest and don't blame anyone, own it, and focus on the learnings (both things that worked and didn't work) Good luck.
For purposes of discussion, I'll assume this is a predictive analytics use case that in the end didn't generate accurate enough predictions to be useful.
Having said that, many D&A initiatives may have accomplished useful work on the way to final failure, and I would start by highlighting that portion. For instance, if data was cleansed, combined in new ways, and insights learned in ways that may be useful for other initiatives, then highlight that as you present a post-mortem.
Then clearly and objectively present the reasons that it didn't work, demonstrating your management of the project, the risks along the way, mitigations taken, etc.
The more you can come across as an executive in charge that had things under control and took a calculated risk that didn't work out, the better shape you'll be in.
Of course, ideally, you would have communicated well the risks of failure upfront, so that you've "pre-warned" them that failure is a possibility regardless of management acumen in some cases.