Hi ,
Picture this...
It's Friday afternoon. The CEO drops by to tell you the news.
That year-long data science project? It's now due next week.
Your stomach drops. That's impossible. Right?
This is exactly what happened to me, and what I learned has changed how I've approached every data science project
since.
My team had been developing a suite of models from scratch to classify incoming audio, video, text and image files based on relevance. And although we all had high hopes for the project, we hadn't yet produced anything worth deploying.
Then suddenly,
everything changed.
The volume of data received by the business had increased by 10X - and the human monitors responsible for reviewing the files were threatening to quit if they didn't get some extra help fast.
We had two weeks to get our models out the door or our
jobs were on the line.
Overnight, the scope of the project completely changed.
Nice to have features were cut and manually trained models were replaced with pre-built APIs for transcription, translation and other data processing tasks.
In 8 days, the suite of models was done. Or at least, to a point that was good enough to produce results.
Our jobs were saved and the human monitors were thrilled. They now had a tool that could classify the easiest 95% of the files, leaving them only the remaining 5% to review.
Then, over the 12 months, we iteratively improved the tool based on the feedback we received - fixing bugs and customising it to address edge cases.
Before the big change happened, the scope of our project had determined the time to value of our project. After the change,
the time to value determined our scope.
One of the biggest myths of data science is that it takes time to create value.
But you don't need more time. What you need is to flip your thinking.
Let time determine scope - NOT scope determine time.
Set an aggressive deadline first, then ruthlessly prioritize what fits within it.
Get feedback from the results. Learn. And then
repeat.
Value comes from solving real problems fast, not perfecting solutions slow.
Talk again soon,
Dr Genevieve Hayes.
