First couple of days of work

I didn't have a chance to blog yesterday about my first day, but I'll include both of my first two days in this post.

Both days I've gone in with Jon at about 9. Everyone else gets there and starts by 10, although they trickle in before then. Everyone who works there is very nice, it's an enjoyable work environment. In all, I've interacted more with the Appfrica developers than with the Question Box operators, I fear, though.

So yesterday I started out upgrading and redistributing the memory between the main (live) server and two other similar machines. One is going to be a development server, and one is to run the eGranary drive, which is a big repository of educational and academic, etc, documents on an external hard drive.

Stupidly, the eGranary only comes with Windows software which is irritating and closed (basically everything here is Linux-based for cost, expandability, openness, and so forth -- all of the reasons it's a better platform for this sort of thing). But it turns out that the contents are easy to access and stored in a standardized fashion (having been pulled from the web in the first place with an open source tool, HTTrack). We're going to try to integrate it into the Question Box software backend using the existing indexing system.

So, upgrading those machines and working with the eGranary were big parts of what I did. It was a pretty slow day for Question Box, but I saw a couple of calls taken. Today has also been slow (the caller liasons, Grameen's so-called Community Knowledge Workers, have been at a training and doing a field survey.

Still, I've been introduced to the software they use, written in-house, which is actually pretty impressive. It integrates searching the approved information sources, beginning with cached previously used answers and then moving on to cached webpages and documents, and eventually the web, if need be. It also includes all of the record-keeping functionality that allows us to keep detailed call records of the sort I've been analyzing.

My task is to figure how to improve what the operators are doing, and as yet I'm not sure. The software's pretty slick, and the operators seem pretty good with it, though there's always room for improvement.

There are a few steps I don't understand, such as logging much of the same information on paper immediately after entering it to the computer. While it isn't harmful when things are slow, I don't understand the purpose and it seems like it would become a major slowdown when volume is higher. So far as I can see (and anyone can tell me) the paper records are never used for anything. I think I will recommend against continuing their use (which has already been mentioned -- that's something Rose [Shuman] and Jon [Gosier] asked me to look at).

Anyway, the other major task I've been working on is an analysis of call log data from the Question Box records. Unfortunately, the behavior of the exporting tools on the server is variable at best; I started with copied and pasted data from the web-based tables into Excel. Since then an Excel-export tool purportedly came back online but still isn't right. Anyway, as of today I've started over on a table pulled from a slightly different query with more complete and useful information, but still missing some handy info (that another set of otherwise less useful queries do include). Also weirdly, the total number of records generated by these different sets of queries don't seem to quite add up, which is distressing.

The data is definitely useful for general trends, regardless of its exactness. We are talking about 1000+ queries for the month of May, so slight imprecision should not affect the overall picture too much. Redoing what I already did is unfortunate, but a lot quicker as now I know the Excel tricks necessary both to strip out some of the data, such as reducing very long answers that bog down Excel to a set of statuses (answered, not answered, referred to expert, etc), as well as generating counts for various categories (and marginal distributions) using Pivot Tables.

It's still not the most enjoyable/interesting/stimulating work, but I feel like I should do it first since it's got to go into a midterm report to Grameen (damn, and I though I was done with midterms...). Also, there's less to observe while Question Box activity is slow like it has been. Still, I know I need to get more actively involved with the operators. Once I get this stuff out of the way and no longer have an excuse not to, I'll have to. That is what I'm here to do, after all. I can wrangle Excel data from anywhere (and could learn a thing or two from Mark at TSW).

Anyway, I've gotten a certain amount done, and a lot of it's going to take time (although identifying software issues to have the developers work on is something to do early, so there's time to develop, debug, and deploy it). This week I still need to try and have a meeting with Barbara (the office supervisor, oversees the operators and handles escalated questions) and the operators themselves to try and better identify problems they see with the system and whatnot. I wanted to see more and better understand what they're doing as is first. Hopefully tomorrow there'll be a bit more activity. And I know I've missed a few calls today working on the data and stuff; I'll try to wrap the report up soon as well.

I'll attach some pictures that I took of my desk (in Jon's office) and if I get a chance to take some, the bigger office area, which is where I'm working now, since there are some seats, and it's more interesting. Plus I can see/hear the operators in action even if I'm not focused on them. The office is high up on a hill, too, and has a really nice view.

EDIT: pictures probably to follow in another post, as I think I'm going to leave my computer here at the office to download iPhone OS 3.0 (as there's now a software unlock for it on the 3G, so I can use it with MTN, T-Mobile, etc). The unlock came impressively quickly after the software release, especially given that there wasn't a working unlock for the previous newest (2.2.1) baseband, only the one before that (causing 3G owners some headache). 2G owners have it so easy by comparison...and it sounds like for the moment 3GS owners (anyone?) are high and dry, but in all likelihood, anyone who has one this early in the game must be on AT&T.

If I'd brought my phone cable, I could have put them up now, but I didn't. Sorry! Tomorrow!