Diego Bash has a nice post about the difficulties of estimating when you don't have the right skills, experience, capabilities, or reference needed to make a credible estimate.
The Nice story on Quora (link and login required) about a hike from SFO to LA. It demonstrates that the right people have to be asked about any estimate. This is the famailar reference class forecasting, which would be the best. Subject Matter Expert is acceptable as well.
Separating the need for an estimate from the skill and experience of the estimator is critical to the estimating success.
The cost of a Low Earth Orbit (LOE) machine can be estimated within 10% by knowing the mass of the spacecraft independent of the sensor platform. Von Braun “estimated” the mass of the lunar lander that could return to the orbiter on the back of a napkin because he was a “subject matter expert.”
Colleagues can estimate the cost and schedule to add features to a health insurance claims processing ERP system because they “know” the domain, context, and have reference classes.
Setting out on a walk from SFO to LA without some type of reference class or consulting an SME is a good analogy of how estimating goes bad. But it’s not the estimating process that goes bad it’s the people doing the process.
So now for the "hard part." Diego's conjecture that every new piece of software is a machine that has never been built before, means we have no – ZERO – basis of reference for the development. It's a green field discovery process. We have no – ZERO – understanding of the underlying elements of the work, the problem to be solved, or how it hs been solved in the past, or how it might even be solved in the future. Or we have no – ZERO – understanding of what “done” even looks like in some unit of measure meaningful to the decision makers.In this case, yes “estimating is hard.”
When all or some of these are the case, our ability to estimate anything is also ZERO.