When there is a sunk cost of any development, be it software, hardware, concrete, a gene-editing anti-virus drug - the value of the resulting product or service is directly related to that sunk cost.
Let's look at a commercial software development project. "We need to save $0.03 to $0.07 per transaction on the provider network of our health care providers," says the CIO of a major health insurance company. "How much is it going to cost to replace the legacy systems with new integrated ERP system that can start to provide those kinds of savings?" "I don't know says the Chief Architect." "But let me start to estimate our efforts, COTS products, conversion processes ad get back to you with a Rough Order of Magnitude (ROM)."
9 months later in this ERP, the CIO sees progress is being made and asks "what's it going to take to add 3 million no subscribers to the network if we bought another insurance company?" Chief Architect says "well we've bee capturing our capacity for work, doing estimates of that capacity before and after and have a calibrated (Cardinal) estimate model." "Let me get with my developers and work up and estimate give what we know now (our Reference Class Forecast) and get back to you."
Now imagine the response to the CIO is - "we use the #NoEstimates appraoch to development, so I can't really tell you what it will cost. Give me $80MM and we'll start work on your top priority features and see where this takes us."
Really? Our you out of the !@#$in mind?
I want to see those who conjecture #NoEstimates is the next wave of agile develop survive that conversation.
If you're writing software using other peoples money, and they say they need an estimate of the sunk cost, it's not because they are dysfunctional. It's because its their fiduciary responsibility to spend that money wisely. Not only wisely, but in accordance with the Governance, SOX, and GAAP rules of the firm. This who notion of searching for dysfunction, exploring on other peoples money, taking out our brains and playing with them is a wonderful experience for the individual doing the searching, exploring, playing. But that doesn't pass the smell test anywhere we work.
If it works for you - providing no estimate of the future cost of future value - enjoy your job. You're in a unique situation. If on the other another hand those with the money ask what can I get for $125MM or what will it cost to some confidence level - say 75% - for these set of capabilities, then being able to estimate that cost with a confidence, error band on the confidence, and the underlying classes of estimating processes to arrive at that number is probably a good skill to have these days of tight money, short delivery depends, and wickedly competitive business environments.