How much do you gain by testing early? A lot!
But how much exactly?
I have prepared this estimation ROI model to get a rough idea of what could be the impact of shift-left. It is specific to a product (Traffic Parrot) but you can use it for generic shift-left as well as is.
Is there anything I have missed?
What did I not account for or misinterpret?
Does this model match your experience or is this totally out of this world?
It’s a bit like the team velocity myth, managers assume velocity is constant over the life of a project or feature, it’s not. And the ticket sizing is as a consequence skewed, and is to boot not accurate anyway. The $$$ gain is probably there, but it’s going to depend more on the team being happy and communicating well, and familiar with each other than we might think for example.
A few typos here and there, but generally all the points I’ve seen tally with my experience. A few folk who are more leader/manager types that observe these aspects than I, will be able to provide more objective pointers I’m sure. I do like the way you set up the environment/scenario here though. Context is King (or should we say Queen) in so much of this and “the book” on all of this weighs in at a good few kilograms, so a small ROI calculator is really simplifying the input graph down a lot. But I mean that’s what models are supposed to do.
From a QA perspective, it is however hard to even do any manual testing early in a green-fields project, because the artefact we give to the testers are not actually useful, often we are building a sports car, and the first drop is a motorized skateboard with a Lamborghini shell with only one seat. So we have to be careful how soon we start shifting left and how we shift left. Shifting left is an experiment, and if your team has not got a lot of experience devising experiments, the only benefits of the shift will be early customer feedback because you got a working skateboard to show people a lot sooner than otherwise would have. The friction that an early drop creates works against in sheer Jira ticket wrangling that the testers will cause you, unless everyone is on the same page as to which parts of the Lamborghini are intentionally omitted at each stage. I.e fine grained, prioritized and schedule-of-delivery planning needs to be much much better and earlier too.