[larrik]

The problem with this approach is when they say "okay, we'll have an intern do the manual keying each week and use the prototype as the finished product. Thank you, goodbye."

Or the fact that it looks 90% done, so the idea that you did it in 2 weeks means you are just days away from completion, right?

Sometimes the reasons for not hacking stuff together can be quite political, both for internal projects and for consultant/client relationships.

[lgunsch]

"okay, we'll have an intern do the manual keying each week and use the prototype as the finished product. Thank you, goodbye." is a valid business choice though. Maybe the cost of the intern and the risk of being unable to effectively maintain or add features is worth it to the client. It's their choice if they would prefer it that way.

[braveo]

Not only that, but the intern can make decisions that an automated system isn't going to be able to do safely due to ambiguities, etc.

This reminds me of an automated licensing call I was putting together once. The company owner wanted me to insert some verbiage into the notes field, only the web api didn't support the notes field (but you could do it by logging in and manually updating).

I had started to plan on how I was going to do this via a headless browser. I start asking the owner questions to clarify things and he says "just email XXX with the information and she'll manually log into the site and copy/paste the notes you put in the email".

And I thought... you know what? that's really fucking smart, I'm way overthinking this. They're going to get 4 or 5 license activations/month MAYBE and this approach is a whole hell of a lot more simple and robust than pulling in a headless browser to simulate it.

I then had to stop and think about why I hadn't considered that approach before.

I guess the point is that I agree with you wholeheartedly. Not only in terms of simply cost, but in terms of complexity, stability, and ambiguity as well.

[zgryw]

There is a bigger story about the project for "YayHappyFunTimesCorp" which included that difficult conversation - how to, instead of carrying on with designing new features quickly, slow down and make sure that everything that we hacked together is scalable. It's definitely not an easy one.

[aisofteng]

On my team, we tend to say that after finishing 90% of a project what remains is the other 90%.