Saturday, November 10, 2012

Inside Team Romney's whale of an IT meltdown - Ars Technica

Inside Team Romney's whale of an IT meltdown: ORCA, the Romney campaign's "killer" app, skips beta and pays the price


It was supposed to be a "killer app," but a system deployed to volunteers by Mitt Romney's presidential campaign may have done more harm to Romney's chances on Election Day—largely because of a failure to follow basic best practices for IT projects.

. . .

IT projects are easy scapegoats for organizational failures. There's no way to know if Romney could have made up the margins in Ohio if Orca had worked. But the catastrophic failure of the system, purchased at large expense, squandered the campaign's most valuable resource—people—and was symptomatic of a much bigger leadership problem.

This is quickly turning into a how to on how NOT to run an IT project.  Project ORCA was the Romney campaign Get out the Vote platform, developed by Microsoft and an unnamed consultant. (ORCA was a play on the Obama version, NARWHALE, because killer whales are a predator of narwhales)  It was developed in seven months, but more importantly, it was developed in relative obscurity, with volunteers who have computer development experience stating that their briefings were more like rah-rah sessions and technical questions that they raised (like 'was this stress-tested?') went un-answered.

1.  Security through obscurity.  The reason given for not having more open testing (e.g. expose it to the internet and put a load on it) was to avoid having the system be hacked.  Most open source advocates state that you want to have your system be exposed and tested by friendly adversaries (e.g. a Red Team made up of campaign volunteers) because they will find faults that you can then fix, security or otherwise.

2.  Users are part of the system.   The users of ORCA had nearly a literal 0 involvement in development of the system. Volunteers who were expected to use ORCA were expected to obtain the app the Monday before elections.  The instructions for the app (which people got Monday night) turned out to be a 60 page PDF.  Those that received and read it complained that it included numerous mistakes that could have been caught with proof reading.  And another surprise for people looking at Android and Blackberry App Stores was that this was a web app.

3.  Users are part of the system part 2.  The deployment of those who were part of the ORCA managed GOTV campaign did not account for any realities on the ground.  Issues like multiple precincts with co-located voting locations, distance from volunteer to the assigned polling place, and the instructions given (or not given) to poll watchers (e.g. they neglected to mention that you need to have a poll watcher certificate to hang out at a polling place or you will be suspected of voter intimidation).

4.  Models, systems, and data are nice, but the value is on what decisions you can make out of it.  When they Romney campaign gave an interview proudly presenting ORCA as their response to the Obama campaign data analysts, they said that the main use was to redirect resources from areas where they determined they had good turnout to areas that had low turnout.  But what are you going to do?  That would mean moving people across the state on short notice?  If you had resources you could have used in an area, why were you not using them before?  i.e. high tech is flashy, but what are you going to do with it?

From what I can tell from various articles and interviews (and blogs by technically adept people who were Romney campaign volunteers) was that Project ORCA may have really been a technical marvel, as those involved still claim, but it was a project management failure.  Rather ironic coming from a campaign and a party that prides itself on business sense.

No comments: