[IgorPartola]

As I mentioned in another reply, this particular data entry job involved doing math at about college linear algebra level and writing SQL among other things. We got to this point by relying on one of the founders to do all the data entry. Suddenly three things happened: our system got more complex so the old tools no longer helped, we signed on a lot more customers, and the founder stepped away for a well deserved break. Thus, we were up a shit creek without a paddle. If the creation of the data entry tools was a highest priority, this would not have happened. However, being strapped for cash, we had to instead add customers, and also launch the more complex system to get paid on one of the bigger contracts. The price to pay was stalling of all development for a period of over a month to do data entry. Everybody that could, did help out but in the end only the dev team could do it all and do it properly. The high level lesson is to not take on this kind of technical debt. What I could have done to avoid it from where I was standing is still something I am figuring out.

[ddebernardy]

Stepping in and highlighting the problem before it happens come to mind.

If you anticipated that such complexity was involved in data entry at any point, putting upper management in front of their responsibility might have helped, i.e. "how do you think dev morale will look like if they need to do data entry for weeks just so we can make a few more quick sales?"

[IgorPartola]

I think this was my biggest mistake. I underestimated the problem and while I explained it to the upper management I did not push it hard enough for them to make it a number one priority.

[Natsu]

Would it really have taken longer to automate it, then feed in the data than to do it by hand?

I ask because I have, when presented with similar tasks, spent the time learning how to do it in WSH (sendkey), doing it, then doing more productive things while the script ran.

[IgorPartola]

So without going into too much detail, yes it would have taken longer. The reason is that we were solving quite a few non-trivial problems in terms of updating geographic road data. Just as an example, a small part of the data entry tools ended up being a road router similar to the one provided by Google Maps, but tuned to our use case. This tuning took quite a bit of work. Another piece was splitting the graph of all roads at each intersection. The algorithm here was simple enough but figuring that this was the right solution took a while.

[edoloughlin]

There's an XKCD for that: http://www.xkcd.com/1319/