Hi ,
"Fast, cheap and good - pick two."
A former boss of mine used to say this all the time - meaning that when you're building a data science solution, you can build something fast and cheap
if you sacrifice quality; cheap and good if you sacrifice speed; or fast and good if you sacrifice cost; but you can't have all three together.
Except, what if you don't need to have all three together to reap the benefits just the same?
The most valuable
data science solutions aren't built by going away and working for 12 months behind closed doors and then emerging with the perfect model. Rather, they're produced by taking an iterative approach that delivers value right from the start.
By strategically selecting which of fast, cheap and good to ignore as you progress, you can dramatically speed up the impact of your
results.
Here's how:
Stage 1: Storyboard (Fast and Cheap)
When you're starting on a project, your solution doesn't need to be good. What's important is understanding what a good solution looks
like. This is where "storyboards" can help.
By taking the time to understand your stakeholders' business questions, the decisions they're trying to make, and what they to see - and putting these requirements together in a "storyboard" (i.e. a Word doc or a PowerPoint deck) that you can share - you're setting yourself up for success in a way that's fast and cheap.
Stage 2: Development (Fast and Good)
Once you've got the storyboard, the next step is to make it real as quickly as you can. A good solution now may be more expensive than the same solution delivered next month. But paying that premium allows you to start getting some wins - and fail fast if the project is doomed
to never succeed.
Stage 3: Optimisation (Cheap and Good)
Having established the value of the solution, you can then make it cheap - through performance optimisation and the like. As the previous fast and good answer from Stage 2 is providing an interim solution to stakeholders, this buys you the time to
build something that's cheap and good.
