Estimates are a constant struggle; we've used all kinds of confidence levels that would multiply our original story points by 50%, 75%, 100% based on how confident we were with the points we assigned.
I'm more like the guy who thinks in Spikes. I take a day, try to figure out, create a small MVP, and get some solid proof. Doesn't always work, but this is my preferred way of estimating further.
As someone who has much less experience, it’s really great to see building a small mvp is actually a solid method from a senior’s eye!! This comment really helps me on this issue and release a lot of burden on giving good estimates, thanks for sharing it!!!
We follow a similar approach, and I believe it’s the most effective way.
We provide two numbers: our estimate and our level of confidence.
The final output is simply their multiplication, which gives a range between optimistic and pessimistic outcomes.
I.e this task will take between 1-2d to be executed.
If the confidence level is particularly low, that’s a clear sign that we need a spike to reduce uncertainty — this is far better than inflating a single estimate just to “play it safe.”
Having two data points is always better than one. When you rely on a single number, you tend to overestimate the unknown to protect yourself and avoid future delays — but in doing so, the real uncertainty remains unaddressed.
Estimates are a constant struggle; we've used all kinds of confidence levels that would multiply our original story points by 50%, 75%, 100% based on how confident we were with the points we assigned.
I'm more like the guy who thinks in Spikes. I take a day, try to figure out, create a small MVP, and get some solid proof. Doesn't always work, but this is my preferred way of estimating further.
As someone who has much less experience, it’s really great to see building a small mvp is actually a solid method from a senior’s eye!! This comment really helps me on this issue and release a lot of burden on giving good estimates, thanks for sharing it!!!
We follow a similar approach, and I believe it’s the most effective way.
We provide two numbers: our estimate and our level of confidence.
The final output is simply their multiplication, which gives a range between optimistic and pessimistic outcomes.
I.e this task will take between 1-2d to be executed.
If the confidence level is particularly low, that’s a clear sign that we need a spike to reduce uncertainty — this is far better than inflating a single estimate just to “play it safe.”
Having two data points is always better than one. When you rely on a single number, you tend to overestimate the unknown to protect yourself and avoid future delays — but in doing so, the real uncertainty remains unaddressed.
Thank you so much Franco Fernando for this article
++ Good Post, Also, start here Compilation of 100+ Most Asked System Design, ML System Design Case Studies and LLM System Design
https://open.substack.com/pub/naina0405/p/important-compilation-of-most-asked?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false