About Me

My photo
Kings Mountain, North Carolina, United States
"A mind lively and at ease" is a blog by a first-generation Russian-Ukrainian immigrant Maria K. (Maria Igorevna Kuroshchepova). An engineer by education, an analyst by trade, as well as a writer, photographer, artist and amateur model, Maria brings her talent for weaving an engaging narrative to stories of life, fashion and style advice, book and movie reviews, and common-sense and to-the-point essays on politics and economy.

Thursday, March 11, 2010

Nerd Central - a day in the life of an operations research analyst

When people ask what I do for a living I tell them I predict the future. It's not a joke - my team and I predict how much volume will arrive at various bank sites for processing. We call ourselves "Nostradami", "the oracles" or "nerd central". Of course, the process we use to do so is far less exotic-looking than the scene with the undulating priestess from the movie 300. My teammates and I often joke that we wish it was as easy as inhaling some fumes or ingesting some hallucinogenic substances.

This is an older post I wrote in December after a particularly exciting day on the job. I shared it today with someone considering joining our team and decided to repost just for the fun of it. Most people following the blog know me as a writer and an artist, so this is the other side of my life.

Today is Tuesday. We received a call on Friday that one of the major volume migrations has not gone anywhere near the way it was supposed to, which caused one of our three sites to become completely overwhelmed with extra volume. This means that everything we did for December and January is wrong and needs to be redone. Normally we have a two-months window between getting a forecast approved and the time when a forecast applies. So, the original version of the December forecast has been approved in October. This being an emergency we are lucky if we get two days.

We get the details on Monday. Someone at the executive level demands that we produce a new forecast ... on Monday. We manage to negotiate the deadline out to Tuesday, because analysis requires time.

Tuesday 7:30 a.m. I am at my computer in my pajamas. I work from home, because the nearest office is an hour away. My only office day is Wednesday. Anyone who says that people who work from home are slackers has no idea what they are talking about. In fact, most companies expect those people to work more, because they don't have to drive.

8:00 a.m. My boss pings me via Communicator. We spend the next 45 minutes going over what needs to be done.

8:45 a.m. I ask my boss if I could step away for 45 minutes - I still haven't eaten. In those 45 minutes I manage to squeeze in an abs workout, a shower, a bowl of cereal and a cup of tea, plus getting dressed.

9:00 a.m. My boss calls me and we iron out the monthly numbers. The whole thing is like a numeric jigsaw puzzle: the totals for each product (credit card payments, mortgage payments, etc.) remain relatively similar from month to month. It's the portions of those totals that go to the three different sites that change. With the migration gone awry and migration plans no longer applicable, we have to deduce what those portions are going to be without anything to back us up other than logic.

10:00 a.m. With the totals more or less in place, the three of us - site analysts - go to work on the daily volume distributions. This is an even greater extrapolation than our monthly numbers, because during migrations, daily numbers do not behave the same way as they would under the normal circumstances.

11:00 a.m. - 1:00 p.m. We go back and forth, continuing to adjust the totals and translating them into the dailies. At 1:00 p.m. I finally publish the monthly numbers to the sites.

2:00 p.m. Show time! The strategic forecast governance meeting. This is where we have to sell what we deduced at the monthly level and get the sites' approval. If the numbers don't make sense to them, we have to go back to the drawing board. I am the meeting's facilitator. As one of my fellow analysts says, this particular one frequently resembles a UN war room. Today it's even more intense, because everyone is stressed about the off-kilter migration.

It takes us an hour to go over the new monthly totals for December and January. As it frequently happens, one particularly fastidious participant nearly hijacks the meeting by bringing up an issue that has nothing to do with the overall concern we are discussing today. It's a valid issue, but this is not a good time to bring it up. One of the site managers helps me wrestle the meeting back on track, and I almost love him for it, despite the fact that he was the one who wanted all the updates on Monday. All in all, it's not nearly the bloodbath I expected.

3:00 p.m. Sold! The monthly numbers pass the site test, but the dailies need to be tweaked. We can't leverage the credit card payment patterns to forecast the mortgage payments - they are not the same at all. Ok, fine - glad we caught it. We go back to revamp the dailies.

4:30 p.m.
The dailies are finally in place, and we can publish. My problem? I am also responsible for a forecast that is a derivative of the one we just updated. Which means it has to be re-done too. Fortunately, it is for a time zone that is two hours behind us.

5:00 p.m. - 5:30 p.m. I debate with myself whether or not to go to aikido. I just got back into training last week after a 2-month break. I have been a brown belt for two damn months, and I have to train consistently if I am ever to make black belt. Besides, if I don't step away from my computer NOW, I am going to start throwing things and weeping uncontrollably. I decide to go.

5:30 p.m. - 6:20 p.m. I drive to the dojo, which is 45 miles away. The weather is the worst imaginable: it's dark, foggy and pouring rain. I make it. I have 10 minutes to change into my uniform, but I make it.

6:30 p.m. - 8:00 p.m. Frustration is good for the training. Despite some pretty intense stuff we work on, I never run out of breath, never botch a fall or a roll, never lose my balance. During an attack line, I only mess up two attacks out of fifteen. I beg off at 8:00 instead of the class end at 8:30 because I feel the distant flickering of fatigue, and I still have to drive home and finish that one last piece.

9:00 p.m. I nuke some tea and get back to the grind. I pull the numbers together for the forecast that derives from the two we had to adjust today. Get them ready for upload... and nearly have a heart attack - the upload is not working and is giving me an error message I have never encountered before. I frantically check the Communicator and see that the guy in charge of the upload is - miraculously - still on line. I ping him and we get this thing straightened out. I try uploading again for December, and it works. I tell the upload guy I love him. I upload January and send out a note to let the site know that both forecasts have been updated and are now available.

9:40 p.m. I am done. Tomorrow is office day, which means getting up at 6:30 a.m. I still need to clean up, pack up my laptop, and pick out tomorrow's clothes. I think I'll have a glass of wine before bed. Studies show that the brain scan of a person who has been working for ten hours straight is identical to the brain scan of a person who is unconscious. I think I am unconscious. Tomorrow is another forecasting marathon - better be ready.


Julie said...

A working day in the life of Maria K! Fascinating. Thanks for sharing.

Quiet said...

It sounds really interesting, Maria. Is it mostly a matter of noticing past patterns and extending them into the future?

Maria K. said...

Dear Quiet,

It's somewhat more than that, and it really depends on the site and what type of transactions we are talking about. When we build a forecast we look at some standard things: seasonality (changes between months), day-of-week dependencies (weekly patterns very quite a lot between places and transaction types), holiday impacts, payday impacts, and some local factors (big events, like football games or NASCAR races or other stuff that's typical to the area).

Then there are things that come along as we work with a site: migrations, local economic climate, local industries, etc. Some of this stuff can be pre-programmed into the code that generates kind of a raw version of the forecast - stuff that is unlikely to change. And then some of it we have to do by hand - where the program is just not smart enough to figure it out.

So, have I confused you yet? :-)

Quiet said...

No, actually. :)