Project Orca isn’t the first run-in the Romney camp had with software testing. A photo-macro app released by the campaign also misspelled “America”

It’s no secret that I’m a bit of a political junkie.  While I enjoy the challenge of software development, there’s something fascinating about the gamesmanship and strategy that goes into political contests.  Ordinarily I have to keep these two aspects of my life separate but every so often they happen to overlap and, when they do, I can’t help but get a little giddy.

So while the political junkie in me was glued to a TV set on Tuesday night, obsessively checking reporting percentages for urban vs rural counties and referencing those against minority population figures and previous election results, the software developer in me didn’t care much about the election until today when I found a story entitled “The Romney Campaign Sure Had Some Bad Smartphone Apps.“

If 2008 was the rise of the data-driven election — and boy-howdy was it — then 2012 was to be the rise of cloud-driven democracy.  The Obama camp invested simply unimaginable amounts of money in data infrastructure and analysis, driving almost every aspect of their campaign off of data services which all fed into the same monolithic data warehouse.  The Romney campaign didn’t go quite so far but was proud to unveil what Republicans called “Project Orca,” a 21st century update to the ages-old practice of poll watching.

Poll watching, as it turns out, is important stuff.  Since campaigns know who their supporters are and where their supporters live they also know where they’re supposed to be voting.  A poll watcher can go to a polling place, sit there with a list, and check off the folks who’ve voted.*  That list can then be passed to volunteers who can target get-out-the-vote (GOTV) efforts on those who haven’t voted.  GOTV efforts are expensive, time consuming, and staff intensive so they’re best not squandered on people who’ve already voted or may not be terribly interested in voting.

Project Orca sought to eliminate a key inefficiency in this process.  If you’re doing your poll watching using a pad and paper there is an interval between when a volunteer crosses off a person’s name and when that list makes its way into the hands of your GOTV staff.  During that interval — which could be hours — your GOTV folks are potentially targeting people who’ve already voted when they could be targeting people who haven’t.  Orca provided a real-time poll-watching capacity so that poll watchers could cross off names on a smartphone which tied in to a national database.  GOTV efforts would thus be updated in real-time as voters checked in.

Except it didn’t work out that way because, well, let’s just quote from the New York Magazine:

They didn’t test it. Politico’s Maggie Haberman and Alexander Burns reported: “Among other issues, the system was never beta-tested or checked for functionality without going live before Election Day, two sources said. It went live that morning but was never checked for bugs or efficiencies internally.”

Let me say that again.  In fact, let me put it in bold and all caps just in case there’s any question at all about what went horribly, terribly, inexcuseably, unimaginably wrong with the Romney campaign’s get out the vote software:

THEY DIDN’T TEST IT.

All total, something on the order of $6 Billion — that’s billion with a “B” — were spent on the 2012 election.  Presumably about half of that — some $3 Billion — was spent to get Mitt Romney elected President.  Every penny of that $3 Billion was spent to convince people to show up on Tuesday and push a button, pull a lever, or check a box for Mitt Romney.  The GOTV effort is, as a result, the single most mission-critical aspect of a political campaign.  Without it, everything else is for naught… and they didn’t test their software.

Why is an answer we may never have.  By most accounts the campaign was confidant that the product would work, that it would hold up under load, and that it would be easy to use but there seems to have been no regression testing, no load testing, and no user testing to validate those assumptions.  As a result, on election day, the vast resources of the Romney campaign were left undirected.  GOTV efforts were inefficient — more inefficient than if they’d simply stuck with the tried and true paper methodology — and volunteers left frustrated and in all probability disillusioned.

This is not to say that the GOTV failure actually cost the Romney campaign the election — determining that will require analysis well beyond the capabilities of this blog — but it almost certainly cost him Florida.

Americans have long been told that “every vote matters” and that remains as true today as it ever was but we can add another political axiom to the pile, one peculiar to our digital age: every test matters too.

* My wife expressed some incredulity that this is, indeed, a thing.  It is.  Here’s an article mentioning a similar application (that worked) from the other side of the aisle.