One reader commented on yesterday's post about spec-driven development with pushback that amounted to, paraphrasing:
It works just fine. The problem is incompetent business analysts being lazy with the requirements, or business owners not nailing down important decisions beforehand.
The LinkedIn comment section isn't the place to properly hash this out, so here's my take.
The valid
If our current delivery flow is deeply dependent on a big and proper spec done just right upfront, of course we'll do better if it's a good spec. And of course that means business analysts tasked with gathering stakeholder requirements and turning them into the spec do their job properly.
Likewise, refusing to make the important decisions that shape the overall landscape in which the product gets developed will lead to an incomplete spec, which, in turn, leads to a poorly built product. Mostly, that comes down to making principled decisions on where you fall on various trade-offs.
The faulty premise
I had a post from a few weeks ago, about the strong pull of faulty logic. It's informed by a chapter in Marty Cagan's great book, Inspired, and lays out the typical way products get developed, which leads to a lot of dysfunction. It lays out the temptingly neat flow of an idea from Idea to Business Case and Roadmap to Requirements, on to Design, Build, Test and finally Deploy.
One way the dysfunction manifests is in the two-way blame game for any derailed product:
The business people complain that the engineers just didn't build it right
The engineers complain that the business folks gave them bad requirements
The superficially alluring solution here is obvious: The business folks will want to write ever more detailed specs so that the engineers will finally get it. And the engineers will demand ever more detailed specs so that the requirements leave nothing unaddressed. And when, surprise, what gets built isn't what was needed, the cycle of blame and calls for more specification repeat anew.
There's no other way out of this gridlock than recognizing that the whole approach is fraught and replacing the "one-shot spec" approach with frequent conversations about building the next right thing, so it can be inspected and adapted much more often.
