I feel that to some extent this should be similar to estimation questions that consultants face in their interviews. Questions like:
- How many gas stations are there in Paris?
- How much savings does the leading bank of the USA have?
- Estimate the population of Indonesia (to someone who is not that familiar with Asia).
It's probably a great deal more complicated than these type of questions. But if there is a course like this and a student still feels as overwhelmed by these type of questions, then I feel the title should be updated accordingly.
The reason I'm also mentioning this is because I'm curious if anyone did both and can attest to whether one does get better at estimation questions like the one I outlined. If so, I just might want to take this course, as I'd like a deeper exploration on the topic than just consultant interview estimation questions.
>The reason I'm also mentioning this is because I'm curious if anyone did both and can attest to whether one does get better at estimation questions like the one I outlined.
You can get better at them over time.
Questions like these spawn a handful of new questions, and those questions might recurse into more questions.
You can find multiple paths forward given what kinds of information you can collect and what kinds of metrics can be used. You can estimate what the accuracy will be from each hypothetical path and you can estimate how much work it will be to collect such information.
It isn't always necessary to solve these kinds of questions this second. Sometimes an interviewer might want a really rough estimate solved then instead of being presented with different paths of research, but imho it's always best to talk about paths forward first, and it is best to hash out and clarify exactly what they want and mean. Only then once it is clear they do not want high accuracy and a simple estimate, then you can do just that.
In the real world you almost never want a low accuracy instant estimate, so I imagine you would look good by showing you can build a path to figuring it out at a higher degree of accuracy than just some basic estimate.
I agree with a lot of this but I find instant estimated of low accuracy are very useful to me regularly. It's when you actually only need to know if something will fall within a certain range.
For example, I need to reprocess some documents and don't know how many will be affected. The scope of the change means that it can't affect more than a few hundred thousand, and I know I can easily reprocess a few million before it becomes an issue. I can leave the estimate there, even though it's "between one and 500k" because all I really need to answer is "is it under 3M?". Similarly there are cases the other way where I know quickly that the scale means certain approaches aren't possible despite not really knowing what the actual value is.
Learning how to identify what questions are important and easy to answer quickly has been something very useful.
I agree with this, which is why I found it a good moment to ask a bit more about the discipline itself with regards to this thread. I'm suspecting that more people disagree than they are commenting here as I'm seeing some downvotes. Unfortunately, they're not participating in the discussion as it would be quite interesting to tease out more why this is or isn't a useful skill. Apparently, opinions on the topic are a bit polarized.
Personally, another thing I find it useful for is for salary negotiations when you're talking face to face. Simply being able to guestimate a company's revenue, amount of clients, cost and their mindset gives you an idea if a low ball or high ball offer is due to mindset or due to finances.
I find in general that estimations of these types are amazingly good for when you really only have a few seconds, or to warrant further inspection. Like you said, if something is orders of magnitude away, then you already know the answer.
How can you see downvotes? I haven't gotten any and that's all I can see.
Taking a stab in the dark here but it might be because YC is mostly software engineers and these kinds of problems rarely fit into the domain. It's a useful skill set, absolutely, but isn't something taught in computer science.
On the data science side this has to be used all the time for most problems, but there are not many data scientists or other kinds of analysts on YC, so the kinds who use these tools are more than likely not going to be represented.
An article by Douglas Hofstadter, On Number Numbness, eloquently described these types of estimation problem and the inability of human to comprehend the scale of large number.
- How many gas stations are there in Paris?
- How much savings does the leading bank of the USA have?
- Estimate the population of Indonesia (to someone who is not that familiar with Asia).
It's probably a great deal more complicated than these type of questions. But if there is a course like this and a student still feels as overwhelmed by these type of questions, then I feel the title should be updated accordingly.
The reason I'm also mentioning this is because I'm curious if anyone did both and can attest to whether one does get better at estimation questions like the one I outlined. If so, I just might want to take this course, as I'd like a deeper exploration on the topic than just consultant interview estimation questions.