The Definitive Article on Numbers and Performance of Narwhal vs Orca - Obama vs Romney - Datamining and voter activation in 2012 US election

Its a month now from the US 2012 election. Most of the final vote counts are in, and many of the secrets of Team Obama's Project Narwhal and Team Romney's Project Orca have been dripping into the public from various sources. I have now gatherred enough data points, to be able to paint the full picture with enough accuracy to tell the full story. This is the voter activation battle of the election, how voter data was collected and mined, and then used on election day, to activate voters. This is a very long article with detail and numbers. Let me do just an Executive Summary here, before we go to the nitty-gritty for those more interested. All analysis here is by TomiAhonen Consulting, based on public data, and in many cases using optimization algorithms to have 'best fit' for the data to match published survey data as close as possible. So let me show the big picture for those in a hurry who don't want to read 11,000 words on stats and math.. This is what it looked like:

CAMPAIGN EFFECTIVENESS IN BATTLEGROUND STATES - NARWHAL VS ORCA
Item . . . . . . . . . . . . . . . Team Obama . . . . . . Team Romney
Registered Voters . . . . . 21.5 M . . . . . . . . . . . 18.8 M
Unvoted RV . . . . . . . . . 15.3 M  . . . . . . . . . . . 13.3 M
Final votes cast  . . . . . . 15.9 M . . . . . . . . . . . 15.0 M
Vote pct/RV  . . . . . . . . . 74% . . . . . . . . . . . . . 80%

Contacts attempted . . . . 33.6 M . . . . . . . . . . . 14.6 M
RV contacted . . . . . . . . 17.4 M  . . . . . . . . . . . . 6.0 M 
Unvoted RV contacted . . 17.4 M  . . . . . . . . . . . . 4.3 M

Actual voters hit . . . . . . 13.1 M . . . . . . . . . . . . . 6.0 M
Voters activated . . . . . . 10.1 M . . . . . . . . . . . . . 2.0 M
Activation percentage  . . 64% . . . . . . . . . . . . . . 13%
Contact effectiveness  . . 30% . . . . . . . . . . . . . . 14%

That is the difference, where it counted - the Battleground states where the 2012 election was decided. Team Obama powered by its Narwhal machine, made twice as many contacts, hit three times as many of its own registered voters (or voters leaning Democratic), achieved four times as many contacts with its own voters, and the real number where it matters, ended up activating five times more voters than Team Romney, powered by its Orca machine. At the Battleground states, Team Obama was able to make an election day activation contact that reached 64% of its final voters on that day, and of the contacts attempted, 30% resulted in an activation. By contrast, Team Romney was only able to make contacts that rached 13% of Romney voters, and a contact had only a 14% chance of activating that voter to vote on the day.

The Narwhal project of Team Obama is the biggest election-related voter list, supporter database and voting support system ever made. It cost over $100 million to produce, and employed over 120 engineers, programmers and mathematicians who worked on the project for more than a year. Its database covered over 175 million voters and massive amounts of data for real time use. Narwhal was far more than just an election-day system, it was used to collect and analyze and even forecast data on most election activities from fundraising to TV advertisement placement. By contrast, the Romney project called Orca was mostly an outsourced project rushed into production only over a few months, mostly with outside consultants and suppliers. Its maximum capacity was 23 million voters. On election day Orca was supported by several other databases and voter lists.

When analyzing the performance in the end, comparing the two systems by states where they were used vs states where not, and comparing the voter turnout gains compared to the 2008 election between Obama and McCain, a measure of voter turnout gain by Romney's Orca system was about 1.4% of Republican turnout (corresponding to a vote margin gain of 0.7% when comparing two candidates). In real terms, Orca delivered about 200,000 votes to the Battleground states in 2012 for Romney, about 29% of his vote gains.

By contrast, Team Obama's Narwhal system achieved a 5.9% gain in Democratic voter turnout in the Battleground states (corresponding to a 2.95% gain in vote margin between two candidates). In real terms, the gains of Narwhal amounted to about 900,000 actual votes cast in the Battleground states, which is exactly the vote margin Obama had in winning those nine states. It is fair to say, that without running Narwhal, President Obama would have faced a dead heat in the election. One should remember, that a Get-out-the-vote campaign effort is almost impossble to measure before the election actually happens, and the average of election polls just before election day suggested the election at almost even, with Obama having a tiny 0.7% edge according to the last RealClearPolitics average of polls on election eve. That he won so handily can be attributed to project Narwhal, which seem to have turned a nailbiter election into a clear victory and re-election for President Obama. If you want to understand how these numbers were achieved, or would like to understand more of how Narwhal and Orca worked, what they acheived, how their effects can be seen comparing Battleground states and non-Battleground states, etc, please follow the article after the fold here:

Welcome back. Lets dig deeper into this intriguing story. And please, do get yourself a cup of coffee before you start, this article runs about 11,000 words or about one full chapter length in one of my hardcover books. 

FIRST THERE WAS - HOUDINI

Briefly, in 2008, Team Obama built on the innovative online campaigning used by the Howard Dean campaign of 2004, vastly expanding and deepening the effort. Team Obama 2008 managed at least 6 separate databases and supporter lists, which had a total of over 27 million names (mostly with overlap) and perhaps about 15 million unique names of Obama supporters/voters. Using emails, SMS text messages and social media, in addition to more traditional home mailings and phone calls, the Obama 2008 campaign achieved massive gains over both in the primary the highly organized and powerful Hillary Clinton election machine (considered at the time the best money organizer and 'ground game' voter activation machine in US politics), and in the general election, the less-efficient John McCain effort. There were many innovative uses of technology from SMS activations on election day to the Obama App for the iPhone to a highly ambitious project called Houdini, for election-day real-time voter activation. Houdini did not perform perfectly due to enormous data overload, but was partially successful. Its main purpose was to learn about each voting precinct's actual vote count and voter turnouts in real time, and to report that back to Obama HQ, which could then organize voter turnout efforts (and any other urgent matters like if there were some legal issues and attorneys needed to be dispatched). Houdini was built around mobile phone use by Obama volunteers at polling places. As the system struggled with the data overload, Team Obama had built a backup plan and sent voter data in by manual means, paper and pencil, runners, and traditional voice calls and emails.

Still, inspite of the final voting day hiccups by Houdini, in the end, the 2008 voter outreach by Obama was so superior, it outperformed McCain's by more than 50%. According to 2008 Exit Polls, Obama's team had reached 21.8 million of his voters, while McCain had only reached 14.1 million of his. Team Obama had reached an unprecedented 31% of his actual voters during the campaign. In the process they also reached 3 million volunteers to donate money (a record) who donated over $500 million in 2008 (also a record). For the first time ever, the Democrats in 2008 had more money than the Republicans - usually the Republicans have the money, and the Democrats only had the volunteers. In 2008 Obama had both. Obama also won by 53% to 46% in the popular vote - considered a landslide in modern election terms.

THEN CAME THE CHALLENGER - ORCA

For 2012, Team Romney knew it would need to do better than what had been done before. The team already managed at least 7 databases and voter/supporter lists, and decided to rush to build a new one in very short time, to be the real time election day monitoring and activation system, similar to Houdini. This, called Orca, was a real time database of voters in the Battleground States, with a total capacity of 23 million names. Romney's team had recruited 34,000 volunteers to serve monitoring all polling places who would then report in using like Houdini did, with mobile phones, what was happening. This data would be collected at Romney's Boston HQ where 800 volunteers manned telephones to direct robocalls where needed. The idea was primarily based around the polling history, certain polling places were Democratic-leaning with a clear history of voting mostly Democratic, while others were Republican-leaning. By calculating the levels of votes from each voting precinct, Boston could estimate which state was performing how, and then decide in real time, where was Romney 'ahead of the game' and not needing extra support at that moment, and where was it behind, and would need a boost of supporters to show up. The volunteers would then push robocalls into those precincts that were heavily Republican-leaning, within those states where Romney would need help, to boost his turnout and thus the vote.

At the same time Orca was supposed to also collect real time info of who had voted, so that calls could be more accurately targeted ie not sending robocalls to those who had already voted. Rather bizarrely, along the election cycle in 2012, while reporters quizzed the Romney campaign about collecting voter data, and pointed out that the the Obama campaign was doing that - such as which of voters at a rally had already voted - the Romney campaign said the ''Juice is not worth the squeeze'' in other words, the trouble of bothering supporters asking such personal questions - and collecting and sending that info to Boston - was not worth the effort. All that became academic, as Orca failed just like Houdini had, under the enormous data load. Romney's Digital Director Zach Boffatt told ABC News that Orca had reacted as if it had been under cyber attack, crashing many times during the day and the whole system shut down for 90 minutes during Tuesday election day. There were many problems by volunteers who only got their user guides to the system the night before via email attachment and the passwords and log-ins were so messed up, that all Romney supporters in two battleground states, Colorado and North Carolina were never able to log into Orca at all, during Tuesday. That meant that Orca was unable to be used for real-time robocall direction, but Daily Caller reports that the Romney campaign then resorted to 'carpet bombing' instead of precision targeting with robocalls.

In all, after Orca was activated, the Romney 2012 voter and supporter databases consisted of at least 8 separate lists and databases, with a combined name count of 130 million names and contacts, mostly with overlap. It included the robocall phone target list for battleground states, that had phone numbers prioritized by Republican-leaning voting district, which had about 40 million phone numbers. The largest actual voter/supporter list in the Romney arsenal was the voter address list with 32 million names. The largest interactive database has 22 million email addresses, while the largest voting day activation database had 15 million mobile phone numbers for SMS activation text messages on voting day. A week before the election, Team Romney released the number, that they had passed 50 million total voter contacs (including duplicates) which covered all their methods and formats, from leaflets and mailings to homes, to calls and emails, to Facebook friends and home visits. Clearly the total Romney voter lists and files collection was a whole order of magnitude greater than what Obama did in 2008, and it was justifyable for Team Romney to feel proud of a highly powerful collection of contact tools to help them win the 'Ground war' ie the Get-out-the-vote efforts of the 2012 campaign.

ENTER THE COLOSSUS - NARWHAL

The Obama team of 2008, led by David Axelrod, had built a powerful set of tools and databases and achieved a huge competitive advantage for Obama over McCain. They also knew, in 2008, after Obama had won, that they would be needing to upgrade this system for 2012, when Obama would run again. With new campaign boss Jim Messina in charge, team Obama set out to build the most powerful election tool ever devised. A system so massive, it would dwarf anything seen or even conceived. They could easily imagine, what a potential rival for 2012 (eventually Romney) might do, and went a whole order of magnitude beyond that. The budget for this colossus was a cool $100 million. Remember, Obama's fund-raising in 2008 attracted $500 million. Now for 2012, they would take one fifth that level of money, just to build a database? Project Narwhal would hire 120 software engineers, dataming geeks and statistics nerds to build the ultimate election tool. To start with, it would be only one comprehensive database - thus with no overlap. Where the largest database for Obama in 2008 had 15 million names as its capacity, and Romney's best had 32 million voter names, Narwhal would manage the data for.. 175 million voters. Just in terms of global customer and consumer databases, Project Narwhal - mind you, this is a temporary project arguably for one election - ranked in the top 20 largest consumer/customer databases ever made, far dwarfing most typical banking, mailing list, subscriber, credit card etc customer databases. Typically the only other databases this big tend to be mobile operator subscriber databases, and even of those there are precious few that are this big or bigger. Plus a few taxpayer files, the IRS in the USA probably has a file slightly bigger, but yes, Narwhal ranks in the top 20 largest databases ever made. And yes, it was a whole order of magnitude bigger than Romney's largest, just by its voter count.

Thats before we consider how many 'relational database' dimensions this behemoth would contain. The problem with most large organizations and their customer databases is, that there typically are many with incompatible data that causes inconvenient gaps in knowledge. For example, you call up your phone company, they offer you a special offer of bundling your cable TV and home landline and mobile phone - and you have to tell them, that you already have that package. That kind of thing. So the first thing, Project Narwhal was going to achieve, was to produce a centralized database that absorbed all data Team Obama had from 2008, and updating all that. No more duplicate databases. So, one file would have the volunteer lists, the contributor lists, the voter lists, the email list, the home address snailmail mailing list, the landline phone number list, the mobile phone number list and the list of Democratic Party members. If you had a landline phone and email and mobile phone, Team Obama in Chicago now knew that, and would not needlessly contact you on the email, landline and mobile to tell you the same thing. This kind of project at this scale is a massive headache and takes at least months, often years. Narwhal achieved this well in time for this election cycle.

Then it had those 175 million names. What is that list? Its essentially the total USA registered voter list, incorporating the RV data from all states (almost all, apparently North Dakota doesn't save its registered voter data). Every single person who is registered to vote in the USA, is in Narwhal. Not only that, but voting is a public record (your vote is secret, but the fact that you voted, is a public record). So Narwhal also collected your voting history at least for the past 5 years. This was used for a clever voter turnout gimmick, by sending via mail to home addresses a 'voter report card' where each Obama-friendly registered voter's recent voting behavior was given a 'school grade' and compared to the neighborhood average. Were you as good a voter as your neighbors, in effect helping 'gamify' the election turnout.

In many states but not all, the voting registration is by party (or you can also register as Independent). If you have this data, and you have the voting history of the past 5 elections, that alone makes this database in Narwhal vastly more powerful than any collected phone or email list by itself. The number of times you voted in the past, is a good indicator of how 'reliable' you are as a voter, and if the party affiliation is included, for those states that collect that data, you then have every voter in the state categorized easily as 'Obama-likely' (Democrats) as 'Persuadeable' (Independents) and 'Romney-friendly' (Republicans). This is FAR more powerful in helping direct your calls, emails and text messages, than what Romney used to send robocalls to districts where Republicans outnumber Democrats.

Narwhal went far further than that. It added all sorts of public and private data, buying databases and customer lists to add to the data. Politico said that in addition to voting history, Narwhal's database included such other consumer info as television viewing tastes, propensity to volunteer, car registration, passions, email address, cellphone numbers, and social media contacts. So for example, magazine subscription lists - if you subscribe to a hunting magazine, you probably are interested in guns and hunting and any gun restriction legislation. If you have a car registered (and what model car), you might be more interested in car-related matters such as fuel effiency laws or the current price of gasoline. And so forth. Narwhal is probably one of the world's most thorough consumer behavior databases. But we're not done.

Narwhal learned and added to its insights. Volunteers kept adding more information, so on election day, Narwhal could not only direct a volunteer to call a nurse, or a student, or a retired army sergeant etc, but also collected all sorts of information that the registered voter had done or said or reacted to, regarding the campaign and its issues. The Narwhal project had established a direct dialogue, literally visting homes and/or via personal phone calls, that had made 125 million personal contacts to Obama supporters by the week before the election, and totalling 150 million by the end. I have calculated that the total number of registered voters reached this way was 26.5 million so Team Obama had made between 5 and 6 actual personal contacts repeatedly and in person, to the same voter, in about a four month period. This allowed an unprecedented level of intimate voter insights of what that given voter thouht about any issues. This info was all fed into Narwhal. How big is 26.5 million? Its 28% of all registered voters identifying with Democrats, and a massive 40% of those who actually came up to vote for Obama on November 6.

Chicago was not afraid to squeeze to get more juice. Narwhal collected all voter behavior it could get, including of course, who had already voted. By that alone, Team Obama could safely ignore the 29% or so, who had voted early in Battleground states before election day. Team Romney would habitually hit not only Democrats with its robocalls, but of Republicans it would reach, nearly a third would have already voted. Team Obama's voting day effort would be directed to those who did, or might, support Obama, and who had not voted. Very powerful targeting insight, when you are in a very great hurry and every resource is precious on election day. No calls wasted on those who had already voted.

One of the many ways Chicago used the voter contacts and Narwhal, was for example to find new unregistered Americans of voting age. Big rock concerts featuring major stars from Bruce Springsteen to Jay Z were used to collect young prospective voters, who were then also provided easy access to register immediately. The Obama campaign registered 1.8 million new voters this way. But the team also used more clever technical means. For example, Team Obama asked its Facebook friends to give Team Obama access to the friends lists. 1 million of Obama's Facebook friends did this (8% of all who had become Obama's friend on FB) and from those names, using the Narwhal database, Team Obama identified 5 million more prospective unregistered voters of whom 1.1 million were subsequently registered as voters. Meanwhile, the Romney campaign used a previously dubious Republican supporter, Nathan Sproul to register voters in several battleground states. His previous company was investigated for fraud and problems with voter registration in past elections. He had created a new company for 2012, and would you know it, his new company also was soon being investigaged in several states for throwing away legitimate registration attempts by Democratic-leaning new voters. Romney had to fire his company. While team Obama was very active and successful in registering new voters (mostly Democrats, of course) team Romney had fired its only new voter registration entity, and too late to get anything to replace it. Just one of the very many cases of campaign incompetence by Team Romney that helped seal his fate.

Back to Narwhal and its database. Narwhal collected, by each registered voter, every one of the various methods of contact - home address, email, landline phone number, mobile phone number, facebook account, Twitter account, etc; plus every type of activity the supporter may have participated in, from volunteering and donating money - incredibly valuable information but also something that is vital to know when assessing voters - remember, 7% of Obama's voters had contributed money, and 3% had personally volunteered to the campaign. These are die-hard supporters who would not need election day 'reminders' or other such encouragement - but equally, who should be separately thanked and recognized, of course.

VOTER RATINGS

So Narwhal has one of the planet's largest databases ever made, consisting of essentially every registered voter in the USA, and has all voter-relevant data 'cleaned' so that there is no overlap and no gaps. It knows if you have been active in politics not just by voting behavior in the past, but if you are a party member and if you've come to a rally or contributed or even liked Obama on your Facebook page or retweeted on Twitter or favorited a YouTube video. Then for battleground states Chicago went truly deep and detailed. They went through every one of the 40.3 million registered voters in the nine battleground states of Colorado, Florida, Iowa, Nevada, New Hampshire, North Carolina, Ohio, Virginia and Wisconsin - and scored every registered voter with two scores, based on all the data they collected. What is your level of support for President Obama - on a score of 1 for none at all, to 100 for total. And then, a second voter score, on your likelihood of voting this election, from 1 of none at all, to 100 for definitely.

Now the voter insight in Narwhal has jumped from the 20th century rudimentary consumer research based on demographics, to the 21st century where advanced consumer data is based on behavior. You might say that young people are more likely to support Democrats, and thus a 19 year old is more likely to be an Obama supporter than a 67 year old, but obviously, actual behavior is far better insights. Now Narwhal knows, who are truly strong Obama supporters - not needing any reminders. Also they know, who are strong Romney supporters (who might, for example, slam the door when you knock on the door and identify yourself as an Obama volunteer). These will never be converted. But those who lukewarm supporters on either side, can be encouraged, with the right messages and arguments, to become Obama supporters. They need the attention from Narwhal and the volunteers.

Same with the propensity to vote. In America you have to separately register before you can vote (an archaic system, but what else is new). So the first thing Chicago wanted to know, of any prospective Obama voters, was if they were already registered - any who were not, were encouraged and helped to register. Then in those states where early voting was allowed, the Obama team would encourage registered Obama supporters to go vote early. And as election day came, they were left with Narwhal database showing who was a registered voter, who had not yet voted. And for each of those prospective voters, there was that voting propensity score. If your likelihood of voting was nearly 100, there was no need for Narwhal to direct any encouragement on election day. If your likelihood was near 1, no matter how great a supporer you were, there was no point in trying to get you to the polling booth. But again, those in the middle, they could use with some encouragement, and for those nearer 80, perhaps just one SMS reminder or email or Facebook reminder would be enough, but for those nearer 20, it might be good to get an actual volunteer to come to your door to help you go vote, or for example to get some others to also encourage you like having a friend call you, or a Facebook friend remind you etc.

With the targeting ability of Narwhal on election day, on those two critical propensity scores, and eliminating any who had already voted; compared to Romney's Orca, which even if it had worked perfectly, would only help direct robocalls to voting districts where Republicans outnumber Democrats - this was like a modern army with machine guns fighting a Zulu tribe with spears. There was no contest at all - and this, before we remember, Orca crashed on election day. And while Team Romney had to abandon their very rudimentary targeting because Orca was crashing, Team Obama had the opposite problem. They were able to recruit so many volunteers, that in some critical battleground states, they didn't need to target. They had too much capacity and simply contacted everybody. As Clint Ecker, Senior Engineer at the Obama Narwhal team said in an article at Techstation, "in some cases we ran out of people to call. We had so many volunteers using it, in some states we just called everybody." 

Team Romney had to abandon precision attempts, because their aiming system was broken. Team Obama could abandon the limitations of precision targeting due to an overabundance of volunteer time and even hit those very unlikely to show up and vote for Obama, and contact even those, just in case. But Narwhal was not just an election-day activation machine. No no no no noooooo.... 

DREAMCATCHER, VERTICA, USHAHIDI, TAARGUS...

Narwhal was way more than just the election day engine. It included many more elements and probably I haven't even found all. But at leats it had Dreamcatcher which was for targeting of messages; Dashboard the smartphone app and mobile website for field workers ie volunteers; the Call Tool which provided highly targeted call groups with individually aimed scripts to help every individual phone call; Vertica the actual database; Ushahidi a poll watching & incident reporting system of for example where to send lawyers for example if there were some irregularities; Optimizer the TV ad targeting system to get most efficient use of TV ad time to reach the best possible viewing demographics at lowest cost; Taargus the social media tool including a Facebook Blaster and a Twitter Blaster to mass-message via FB or TW plus other social media; and QuickDonate the tool to collect SMS donations on a single click.

Because Narwhal was a single comprehensive database and centralized tool to collect all possible campaign-oriented data points, it was also the logical place to drive simulations. And boy did the Obama team run simulations. Chicago ran 2.. MILLION simulations through Narwhal !!!!! They could test anything and everything. They tested which email header would result in a higher rate of opening an email, and what text inside a button would result in the web link being accessed more etc. They ran A/B tests and ran all sorts of simulations, for example discovering that a given variation of several email body texts on a given day, would yield 5 times more total donations than another variation. They used the simulations to test the tech team including total system crash testing in simulations. They had a consumer survey panel of nearly astronomical scale compared to traditional consumer surveys - 29,000 respondents in the state of Ohio alone (nearly 1% of the total voting population in that battleground state). For comparison, a usually reliable national consumer survey might have a sample size of 1,000 and many state-wide surveys might have sizes of around 600 interviews. 

Chicago could run every conceivable test and variation with its target groups and refine and learn and fine-tune. It knew what messages worked best in whatever media, in what states, and to what voter demographics. That is why, it often seemed like Obama was slow to react to some breaking news stories. They did not want to make mistakes and mis-steps, what Team Romney was very prone to do with breaking news, from the Libya Consulate mess to the Jeep jobs to China TV ads. So what happened? Team Obama would run a simulation first - of whether that given story would help or hurt their election messages - and very importantly - using A/B testing, they could compare what worked best! So even if something Romney did this week seemed like a good story to jump on like the 47% comment or wanting to kill Big Bird, team Obama always reacted in a measured way, and with a lag in timing. They tested it first, to make sure, they didn't accidentially substitute a popular current story of modest impact to the voters, replacing a better-performing argument that was winning over more voters.

I wrote more about Narwhal and Orca in the earlier piece about these campaign tools so I won't repeat all that here. We want to get to the numbers. Here is where the rubber meets the road. How effective were the two campaign in 2012. May we have the envelope, please?

NUMBERS, FIRST TEAM ROMNEY

I have taken all possible public sources of information and constructed the following tables of the types of contacts by both sides. I have divided it into two main tables, one is on the early contacts before election day, and the second table is the 'Get-out-the-vote' table for election day itself. Because the election was won and lost actually in the nine 'Battleground states' of Colorado, Florida, Iowa, Nevada, New Hampshire, North Carolina, Ohio, Virginia and Wisconsin, I have separated those out, where for Romney I separate the Battleground States and the Rest of USA (plus the total USA) contact activities. For Team Obama I actually have even the split between the four 'critical states' Colorado, Florida, Ohio and Virginia - separated out, as those are the ones where Team Obama was so rich in volunteer support, they could simply flood those four critical states and hit essentially everybody. But lets start with Boston.  

REPUBLICAN VOTERS CONTACTED BY ROMNEY BEFORE ELECTION DAY
Republican Voter Base 2012
Registered Voters . 18.8 M . . . . . . . . 57.9 M . . . . . . 76.7M
Actual Votes  . . . . 15.0 M . . . . . . . . 45.7 M . . . . . . 60.7 M
Sources: NBC Election Briefing Book, Pew voter identification, Wikipedia, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

Contact type . . . . . Battlegrounds . . Other States  . . Total Contacts
Any kind of . . . . . . 11.9 M . . . . . . . . 35.6 M . . . . . . 47.7 M
Leaflet  . . . . . . . . . 10.7 M . . . . . . . . 35.6 M . . . . . . 46.3 M
Mail . . . . . . . . . . . . 9.6 M . . . . . . . . 13.7 M . . . . . . 23.3 M
Robocalls . . . . . . . . 9.6 M . . . . . . . . 13.7 M . . . . . . 23.3 M
eMails (forwarded) . . 5.0 M . . . . . . . . 16.5 M . . . . . . 21.5 M
Calls  . . . . . . . . . . . 7.9 M . . . . . . . . 13.2 M . . . . . . 21.1 M
SMS texts  . . . . . . . 3.8 M . . . . . . . . 11.4 M . . . . . . 15.2 M
Social media . . . . . . 3.6 M . . . . . . . . 11.9 M . . . . . . 15.4 M
Home visits . . . . . . . 2.8 M . . . . . . . . . 2.6 M . . . . . . . 5.4 M
Sources: Pew, YouGov, CBS/NYT, NBC/WSJ, ABC/WaPo, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

Those were the various ways that Boston contacted Romney supporters and voters in 2012 before election day, estimated as best, removing any contacts that went to supporters of Obama. You can see, that in Battleground states and overall to all voters, the major means by which most voters were contacted, were traditional unidirectional media and communications, like mailers, robocalls and traditional mail. Less than half were contacted by any modern digital interactive means like eMails, SMS text messages and social media. Now lets look at operation Orca and election day for Boston:

REPUBLICAN VOTERS CONTACTED BY ROMNEY ON ELECTION DAY
Republican Voter Base 2012
Registered Voters . 18.8 M . . . . . . . . 57.9 M . . . . . . 76.7M
Actual Votes  . . . . 15.0 M . . . . . . . . 45.7 M . . . . . . 60.7 M
Sources: NBC Election Briefing Book, Pew voter identification, Wikipedia, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

Contact type . . . . . Battlegrounds . . Other States  . . Total Contacts

ROBOCALLS
Robocalls . . . . . . . 6.0 M . . . . . . . . 0.0 M . . . . . . . . 6.0 M
Miss (Dem/Voted) . 2.8 M . . . . . . . . 0.0 M . . . . . . . . 2.8 M
Hit unvoteds . . . . . 3.2 M . . . . . . . . 0.0 M . . . . . . . . 3.2 M
Ignored robocall  . . 2.1 M . . . . . . . . 0.0 M . . . . . . . . 2.1 M
Listened to . . . . . . 1.1 M . . . . . . . . 0.0 M . . . . . . . . 1.1 M
Voter activation . . . 0.2 M . . . . . . . . 0.0 M . . . . . . . . 0.2 M

FACEBOOK FRIENDS (passive)
Voter Friends FB . . 3.6 M . . . . . . . 11.9 M . . . . . . . 15.5 M
20% friend reach . . 0.7 M . . . . . . . . 2.4 M . . . . . . . . 3.1 M
Miss (voted/robo) . . 0.2 M . . . . . . . . 0.8 M . . . . . . . . 1.0 M
To activate . . . . . . .0.5 M . . . . . . . . 1.6 M . . . . . . . . 2.1 M
Voter activation . . . 0.3 M . . . . . . . . 0.9 M . . . . . . . . 1.2 M

SMS
Text message . . . . 3.8 M . . . . . . . 11.4 M . . . . . . . 15.2 M
Miss (voted/FB/.. ) . 1.2 M . . . . . . . . 3.5 M . . . . . . . . 4.7 M
To activate  . . . . . . 2.6 M . . . . . . . . 7.9 M . . . . . . . 10.5 M
Voter activation . . . 1.3 M . . . . . . . . 4.0 M . . . . . . . . 5.3 M

EMAIL 
eMail . . . . . . . . . . . 5.0 M . . . . . . . . 16.5 M . . . . . . . 21.5 M
Not opened  . . . . . . 4.0 M . . . . . . . . 13.2 M . . . . . . . 17.2 M
Opened . . . . . . . . . 1.0 M . . . . . . . . . 3.3 M . . . . . . . . 4.3 M
Miss (voted/SMS.. ) . 0.4 M . . . . . . . . .1.3 M . . . . . . . . 1.7 M
To activate . . . . . . . 0.6 M . . . . . . . . . 2.0 M . . . . . . . . 2.6 M
Voter activation . . . . 0.2 M . . . . . . . . . 0.7 M . . . . . . . . 0.9 M

TOTAL VOTING DAY ROMNEY CAMPAIGN
Robocalls . . . . . . . 6.0 M . . . . . . . . 0.0 M . . . . . . . . 6.0 M
Robo activation . . . 0.2 M . . . . . . . . 0.0 M . . . . . . . . 0.2 M
Voter Friends FB . . 3.6 M . . . . . . . 11.9 M . . . . . . . 15.5 M
FB activation . . . . . 0.5 M . . . . . . . . 1.7 M . . . . . . . . 2.2 M
SMS Texting . . . . . 3.8 M . . . . . . . 11.4 M . . . . . . . 15.2 M
SMS activation . . . 1.3 M  . . . . . . . . 4.0 M . . . . . . . . 5.3 M
eMails sent . . . . . . 5.0 M . . . . . . . 16.5 M . . . . . . . 21.5 M
eMail activation . . . 0.2 M  . . . . . . . . 0.7 M . . . . . . . . 0.9 M

TOTAL Contacts . . . 14.8 M . . . . . . . 27.9 M . . . . . . . 42.7 M
As % of voters . . . 99%  . . . . . . . . 61% . . . . . . . . . 70%
TOTAL Activation . . 1.9 M . . . . . . . . 5.5 M . . . . . . . . 7.5 M
Activation pct . . . . 13% . . . . . . . . 20% . . . . . . . . . . 18%

Total GOP Contact . 6.0 M . . . . . . . 13.8 M . . . . . . . 19.9 M
As % of voters . . . . 40%  . . . . . . . . 30% . . . . . . . . . 33%

Unvoted GOP . . . . 4.3 M . . . . . . . . 9.8 M . . . . . . . 14.1 M
As % of voters . . . 29%  . . . . . . . . 22% . . . . . . . . . 23%

RV activated . . . . 11% . . . . . . . . . 10% . . . . . . . . . 10%
Voters activated . . 13% . . . . . . . . . 12% . . . . . . . . . 12%
Sources: Pew, YouGov, Exit Polls North Carolina and Iowa, CBS/NYT, NBC/WSJ, ABC/WaPo, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

That was Romney's effort. In the Battleground staes, they made a total of 14.8 million outbound contacts (plus had 3.6 million more Facebook friends to help) which together achieved 1.9 million voters activated on election day. The activation efficiency in Battleground states by Team Romney, due to the carpet-boming nature of mass saturation was only 13%. They hit 40% of total Romney voters including those who had already voted and 29% of Romney voters who hadn't yet voted before election day but would vote on the day. Of all registered Republican (and leaning) voters 11% were activated by team Romney, and of his actual voters, 13% received an activation contact on election day. This was a big task well done, in the context of previous elections, achieved mostly with classic use of 'carpet bombing' tactics and obviously with a lot of overlap and hitting Democrats etc.

Nationally, the Romney campaign made 42.7 million outbound contacts on election day (supported by 15.5 million Facebook friends) which achieved 7.5 million voter activations for an 18% activation efficiency percent. The team hit 33% of all Romney voters but removing those who had voted early, of actual voting day voters, Romney's team hit 23% of its supporters on the day. Of registered Republican (and leaning) voters 10% received a reminder from Team Romney on election day, and of his actual voters, 12% did.

NOW THE MASTERCLASS - HOW TEAM OBAMA CONTACTED ITS VOTERS - NARWHAL

The whole methodology of Team Obama was different, thanks to Narwhal. They could avoid the overlap on all media they controlled directly, and they could target precisely. Lets start first, by looking at the types of contacts Chicago had made prior to election day:

DEMOCRATIC VOTERS CONTACTED BY OBAMA BEFORE ELECTION DAY
Democratic Voter Base 2012
Registered Voters . 21.5 M . . . . . . . . 71.6 M . . . . . . 93.1 M
Actual Votes  . . . . 15.9 M . . . . . . . . 49.5 M . . . . . . 65.4 M
Sources: NBC Election Briefing Book, Pew voter identification, Wikipedia, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

Contact type . . . . . Battlegrounds . . Other States  . . Total Contacts
Any kind of . . . . . . 14.9 M . . . . . . . . 28.1 M . . . . . . 43.0 M
Leaflet  . . . . . . . . . . 8.0 M . . . . . . . . 26.5 M . . . . . . 34.4 M
Mail . . . . . . . . . . . 10.9 M . . . . . . . . 19.6 M . . . . . . 30.5 M
Robocalls . . . . . . . 10.9 M . . . . . . . . 19.6 M . . . . . . 30.5 M
Calls  . . . . . . . . . . . 8.6 M . . . . . . . . 17.9 M . . . . . . 26.5 M
SMS texts  . . . . . . . 5.9 M . . . . . . . . 18.3 M . . . . . . 24.2 M
Social media . . . . . . 4.3 M . . . . . . . . 14.3 M . . . . . . 18.6 M
eMails (forwarded) . . 2.6 M . . . . . . . . . 8.6 M . . . . . . 11.2 M
Home visits . . . . . . . 5.4 M . . . . . . . . . 5.7 M . . . . . . 11.1 M
Sources: Pew, YouGov, CBS/NYT, NBC/WSJ, ABC/WaPo, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

You can see how much more diversified and evenly split the contact types are by Team Obama than Team Romney. In the Battleground states, more than a third of all people contacted were by personal home visits and more than half were personal phone calls. Leaflets ranked only fourth in these critical states as the ways voters were reached.

Nationally, Team Obama again used diversity and targeting more efficiently due to insights from Narwhal. Leaflets, robocalls and mailings were used to reach most homes, but more than half of all contacts were made at some point in person, by phone, and more than half were reached by interactive digital media - SMS text messaging. Now lets see how Chicago went into overdrive for Election Day. Note, I have separated now the data into one more column, adding the Critical 4 states of Colorado, Florida, Ohio and Virginia. On election day that is where the election was decided. Romney had to win all four, Obama needed to win only one of the four, but had to win at least one. Obama ended up winning all four of these critical states. But look at the intensity of Obama Team's contacts in those Critical States even when compared to the other five Battleground States:


DEMOCRATIC VOTERS CONTACTED BY OBAMA ON ELECTION DAY
Democratic Voter Base 2012
Registered Voters . 21.5 M . . . . . . . . 71.6 M . . . . . . 93.1 M
Actual Votes  . . . . 15.9 M . . . . . . . . 49.5 M . . . . . . 65.4 M
Sources: NBC Election Briefing Book, Pew voter identification, Wikipedia, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

Contact type . . . . . Criticals . . Other BGS . . Tot BGS . . Other States  . Total Contacts

HOME VISITS
Home visits . . . . . . 6.0 M . . . . 0.0 M . . . . . . 6.0 M . . . . 0.0 M . . . . . . . 6.0 M
Voters reached . . . 7.5 M  . . . . 0.0 M . . . . . . 7.5 M . . . . 0.0 M . . . . . . . 7.5 M
Activated  . . . . . . . 4.5 M  . . . . 0.0 M . . . . . . 4.5 M . . . . 0.0 M . . . . . . . 4.5 M

CALLS
Personal calls . . . . 9.5 M . . . . 1.5 M . . . . . . 11.0 M . . . . 0.0 M . . . . . . 11.0 M
Hit Republicans . . . 2.8 M . . . . 0.0 M . . . . . . . 2.8 M . . . . 0.0 M . . . . . . . 2.8 M
Reach Democrats . 6.7 M  . . . . 1.5 M . . . . . . . 8.2 M . . . . 0.0 M . . . . . . . 8.2 M
Half answer phone . 3.4 M . . . . 0.8 M . . . . . . . 4.1 M . . . . 0.0 M . . . . . . . 4.1 M
Activated . . . . . . . . 0.6 M . . . . 0.1 M . . . . . . . 0.7 M . . . . 0.0 M . . . . . . . 0.7 M

FACEBOOK FRIENDS (passive)
Voters as Friends . . 2.8 M . . . . 1.5 M . . . . . . . 4.3 M . . . 14.3 M . . . . . . 18.6 M
20% friend reach . . . 0.6 M . . . . 0.3 M . . . . . . . 0.9 M . . . . 2.9 M . . . . . . . 3.8 M
Miss (voted/called)  . 0.2 M . . . . 0.0 M . . . . . . . 0.2 M . . . . 0.0 M . . . . . . . 0.2 M
To activate . . . . . . . 0.4 M . . . . 0.3 M . . . . . .. . 0.6 M . . . . 2.8 M . . . . . . . 3.5 M
Voter activation  . . . 0.2 M . . . . 0.2 M . . . . . .. . 0.4 M . . . . 1.6 M . . . . . . . 2.0 M

FACEBOOK APP (active)
Friends with App . . 0.2 M . . . . 0.1 M . . . . . . . 0.3 M . . . . . 0.0 M . . . . . . . 0.3 M
8x friend reach  . . . 1.8 M . . . . 0.9 M . . . . . . . 2.7 M . . . . . 0.0 M . . . . . . . 2.7 M
Miss (voted/called)  . 0.6 M . . . . 0.0 M . . . . . . . 0.2 M . . . . 0.0 M . . . . . . . 0.2 M
To activate . . . . . . . 1.2 M . . . . 0.9 M . . . . . .. . 2.1 M . . . . 0.0 M . . . . . . . 2.1 M
Voter activation  . . . 0.7 M . . . . 0.5 M . . . . . .. . 1.2 M . . . . 0.0 M . . . . . . . 1.2 M

SMS
Mobile numbers . . . 8.5 M . . . . . 1.0 M . . . . . . . 9.5 M . . . 14.5 M . . . . . . 24.0 M
Actual SMS sent . . 5.3 M . . . . . 0.9 M . . . . . . . 6.2 M . . . 14.2 M . . . . . . 20.4 M
Activation . . . . . . . .2.6 M . . . . . 0.4 M . . . . . . . 3.1 M . . . . 7.1 M . . . . . . 10.2 M

EMAIL
eMail List . . . . . . . 1.7 M . . . . . 0.8 M . . . . . . . 2.6 M . . . . 8.6 M . . . . . . 11.2 M
Actual eMails sent . 0.8 M . . . . . 0.7 M . . . . . . . 1.5 M . . . . 7.7 M . . . . . .   9.2 M
Not opened . . . . . . 0.6 M . . . . . 0.6 M . . . . . . . 1.2 M . . . . 6.1 M . . . . . . . 7.3 M
Opened  . . . . . . . . 0.2 M . . . . . 0.1 M . . . . . . . 0.3 M . . . . 1.5 M . . . . . . . 1.8 M
Voter activation . . . 0.1 M . . . . . 0.0 M . . . . . . . 0.1 M . . . . 0.5 M . . . . . . . 0.6 M

SMS-TO-CALL
Volunteer to call . . 0.4 M . . . . . 1.5 M . . . . . . . 1.9 M . . . . . 0.0 M . . . . . . 1.9 M
Half answer call . . 0.2 M  . . . . . 0.8 M . . . . . . . 1.0 M . . . . . 0.0 M . . . . . . 1.0 M
Voter activation  . . 0.0 M . . . . . 0.1 M . . . . . . . 0.2 M . . . . . 0.0 M . . . . . .  0.2 M

TOTAL VOTING DAY OBAMA CAMPAIGN
Home visits . . . . . . 6.0 M . . . . 0.0 M . . . . . . 6.0 M . . . . 0.0 M . . . . . . . 6.0 M
Activated  . . . . . . . 4.5 M  . . . . 0.0 M . . . . . . 4.5 M . . . . 0.0 M . . . . . . . 4.5 M
Personal calls . . . . 9.5 M . . . . 1.5 M . . . . . . 11.0 M . . . . 0.0 M . . . . . . 11.0 M
Activated . . . . . . . . 0.6 M . . . . 0.1 M . . . . . . . 0.7 M . . . . 0.0 M . . . . . . . 0.7 M
Voter Friends FB. . . 2.8 M . . . . 1.5 M . . . . . . . 4.3 M . . . 14.3 M . . . . . . 18.6 M
Voter activation  . . . 0.2 M . . . . 0.2 M . . . . . .. . 0.4 M . . . . 1.6 M . . . . . . . 2.0 M
FB App via Friends . 0.2 M . . . . 0.1 M . . . . . . . 0.3 M . . . . . 0.0 M . . . . . . . 0.3 M
Voter activation  . . . 0.7 M . . . . 0.5 M . . . . . .. . 1.2 M . . . . 0.0 M . . . . . . . 1.2 M
SMS reminders . . . 5.3 M . . . . . 0.9 M . . . . . . . 6.2 M . . . 14.2 M . . . . . . 20.4 M
Activation . . . . . . . .2.6 M . . . . . 0.4 M . . . . . . . 3.1 M . . . . 7.1 M . . . . . . 10.2 M
eMails sent . . . . . . 0.8 M . . . . . 0.7 M . . . . . . . 1.5 M . . . . 7.7 M . . . . . .   9.2 M
Voter activation . . . 0.1 M . . . . . 0.0 M . . . . . . . 0.1 M . . . . 0.5 M . . . . . . . 0.6 M
SMS to Call . . .  . . 0.4 M . . . . . 1.5 M . . . . . . . 1.9 M . . . . . 0.0 M . . . . . . 1.9 M
Voter activation . . . 0.0 M . . . . . 0.1 M . . . . . . . 0.2 M . . . . . 0.0 M . . . . . .  0.2 M

TOTAL Contacts  . 26.6 M . . . . . 7.0 M . . . . . . . 33.6 M . . . . 36.2 M . . . . . 69.8 M
As % of voters . . . 256%  . . . . . 128% . . . . . . . 212% . . . . . 73% . . . . . . . 107%
TOTAL Activation . .  8.7 M . . . . .1.4 M . . . . . . . 10.2 M . . . . 9.2 M . . . . . . 19.4 M
Activation pct . . . . 33% . . . . . . . 21% . . . . . . . . 30% . . . . . 26% . . . . . . . 28%

Total DEM Contact . 14.7 M . . . . 2.7 M . . . . . . . 17.4 M . . . . 15.7 M . . . . . 33.2 M
As % of voters . . . . 142%  . . . . . 49% . . . . . . . . 110% . . . .  32% . . . . . . . 51%

Unvoted DEM . . . . 14.7 M . . . . 2.7 M . . . . . . . 17.4 M . . . . 15.7 M . . . . . 33.2 M
As % of voters . . . . 142%  . . . . . 49% . . . . . . . . 110% . . . .  32% . . . . . . . 51%

RV activated . . . . . 78% . . . . . . . 25% . . . . . . . . . 60% . . . . 16% . . . . . . . 26%  
Voters activated  . . 84% . . . . . . . 26% . . . . . . . . . 64% . . . . 19% . . . . . . . 30%
Sources: Pew, YouGov, Exit Polls North Carolina and Iowa, CBS/NYT, NBC/WSJ, ABC/WaPo, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

Obama totally outclassed the Romney effort. In short, Obama reached more than 2 times more voters and achieved per contact more than twice the activation result. In the Battleground states, 33.6 million contacts either directly or through Facebook friends and the app, or via SMS-to-call volunteers, which achieved 10.2 million activations, more than 5 times more activations than what Romney's effort achieved. The activation efficiency in Battleground states by Team Obama was 30% - in the four Critical states it was 33%. Chicago hit 110% of total Obama voters which means 10% spilled over to Romney's side (to try to capture some of those) and in the four Critical States, the reach went 142% ie 42% beyond those who actually voted for Obama. Of all registered Democrats (and leaning) voters 60% were activated by team Obama, and of his actual voters in the Battleground states, 64% received an activation contact on election day. Within the four Critical States, 78% of all Democratic-leaning Registered Voters and 84% of those who actually voted for Obama were reached on election day. A truly herculean achievement.

Nationally the numbers are also impressive. Chicago achieved 69.8 Million contacts that achieved 19.4 total voter activations on the day. The activation efficiency was 28% nationwide. Of unique contacts, 33.1 million Democratic and leaning voters were reached on election day which is 51% of all voters. Of all registered voters leaning Obama's way, the Obama election day campaign hit 26% and of his actual voters, they reached 30%. Bear in mind, Obama's team made no reminder contacts directly that hit voters who had already voted (although some of the Facebook friend contacts and activations would have, as these were beyond Chicago's control).

SOME NOTES ON THE DATA AND STATS

I should mention a few data points. I used any and all recent surveys and reported findings from the 2012 Exit Poll to various voter polls by Pew and the different news organizations as I mention in the above where relevant. For the voter activation percentage, I found this useful very recent survey by inc/WomanTrend which found that Facebook reminders will have a 57% activation effect on election day; SMS reminders 50%, eMail 33% and telephone calls 18%. It did not differentiate between Robocalls, real person-to-person calls, nor calls to landlines or mobile phones. I used the same activation percentage even though, we can be pretty sure a recording of a robocall will be least effective, while a timed, personal person-to-person call to an individual's mobile phone, will be most effective of phone types. I used 18% activation for all.

For the home visits, I made a very cautious assumption that any home visited by Team Obama would have on average 1.25 voters living in it. Every time, obviously, there would be at least one, but often it could be a couple with two voters, or could even have senior citizens living with their adult children, and/or have adult-aged children still living with their parents. A household - one door knock - might easily have six voters, mom and pop, grandma and grandpa, and say 2 voting age children say age 19 and 18.. I used the 1.25 voter rating just to be very conservative.

Also there was no survey data suggesting how big an impact a visit to a home would have on voting. It must by all logic me more than a phone call, email, Facebook or SMS reminder, but how much bigger. Again, I was very cautious, and used only a 60% voter activation level (remembering that a Facebook reminder would be at 57% and an SMS text message reminder 50%).

I made no attempt to guess how many individual contacts might be made to the same person using the same method, but some degree of logic can be used. A personal visit is so expensive and cumbersome, that any team wouldn't try more than once. A phone call might be tried more than once, as might a robocall. The Obama SMS-to-call volunteer action might be particularly good, if a direct call reminder from the Obama team wasn't enough to get the voter to the polling place, a call from another supporter (usually of similar background, ie a student calling a student, a military vet calling a military vet, a union member steel worker calling another union member steel worker etc) might be then more effective than a second call from the same Obama office some hours later.

In terms of SMS reminders, we know in 2008 Team Obama sent 3 reminders on election day. This time its likely to be the same amount but obviously to far more potential voters. An email system can be automated to send reminder emails at certain intervals, and as most emails are ignored, sending a couple on the day would probably not be a bad idea. Incidentially, I used the 80% email unopened rate and 50% voice calls unanswered rates as the current rules of thumb in the telecoms/IT industry as fairly standard global metrics. As to Robocalls, we had the Pew study that said that most robocalls are not listened to, and only 35% are.

For the SMS mobile phone number list, I didn't get that directly from anywhere but a Pew survey after the election found that 37% of 'Liberal voters' and 25% of 'Conservative voters' had used a mobile phone to engage with the campaign. I used those as proxies of how many voters from either side had given their phone number to the campaign (Liberals = Obama voters, Conservatives = Romney voters) and thus also received SMS text messages. The rough percentages are fairly consistent with cellphone ownership - youth more than the elderly, minorities more than whites, renters more than home owners etc; and SMS use bigger among the youth than elderly, etc. The data is very consistent with a McKinsey study of mobile phone use vs landline phone use by age and a Pew study of SMS use by age.

For example we learned from an interview of David Axelrod that when they analyzed their voter data, in the under 29 year bracket, of Obama supporters, more than half lived in households that had no fixed landline phone at all. Nationally the ratio is about 30% currently according to CDC data.

DID IT MATTER?

First, President Obama won by a margin of nearly 5 million votes by current count (still more than a dozen states have not finished counting ballots) and roughly 3.5% margin at about 51% to 47%. If Team Obama activated 10.1 million actual voters and Team Romney only 2.0 million - then yes, theoretically the difference of 8.1 million is bigger than Obama's winning margin of 5 million. But that is too simplistic a measure. Many of those who were activated ie reminded to vote, would have done so anyway. We have really no way of knowing, until someone does empirical field studies of this election asking such precise questions as 'would you have voted without the reminder' - and we may never actually get that kind of research done or published. One way to think about it, is that the voter activation effort is literally happening on election day. If we say that the average of opinion polls in the last days before the election is accurate to that point, and then measure the result on election day as the difference, then the Obama advantage could be said to be the difference between what RealClearPolitics final average of all latest polls was - an Obama national vote edge of 0.7% and the final count (as it stands today) at 3.7%. So the 'activation advantage' of 8.1 million for Narwhal over Orca, would translate into 3.0% margin or 3.9 million votes. Note that obviously President Obama would have won even without this, by this assumption, but it means that Chicago's magnificent turnout machine managed to boost Obama's win margin by nearly 4 million votes and turn a razor-thin squeaker election into a clear dominant win.

That is still only conjecture, we don't have better data to evaluate the effect more accurately, but its a fair point to say, that Team Obama outperformed essentially all last election polls and on average by 3%. Would that have changed the election outcome. Clearly Obama would still have won the popular vote by 0.7% and about 1 million votes over Romney, but how much is nationwide shift down of 3% across all states? Which states would have 'flipped' from Obama to Romney? Two. Florida obviously, and Ohio just so barely, that it would have gone into recounts. The next in line, Virginia, was too far gone for Obama, that even a 3 percent swing would not win it for Romney. Even without this gain from Narwhal outperforming Orca, President Obama would have been re-elected. But rather than his margin in Electoral Votes of 332 to 206, his margin would have been only 285 to 253 Electoral College votes.

MEASURING PERFORMANCE

But we can actually look at a few other ways of this. First, lets see how the two systems performed within this year's election against actual voters turned out against Registered Voter counts. So in the Battleground states where 24% of the voters live, what was the actual turnout as percent of Registered Voters, compared to the rest of the country. For Romney, the Rest of USA achieved 78.9% turnout of Registered Voters, but in Battleground States, were Orca was deployed, the turnout was 79.8%. A gain of 0.9% in Republican voter turnout increase. How did Obama's Narwhal do? The turnout in Rest of USA was 69.1% from Democratic Registered Voters, but in the Battleground States it was up to 74.0%. So Narwhal was able to generate 4.9% more voter turnout by this measure.

Then another way. We can examine how Orca and Romney full effort voter activation performed against 'the normal' for the Republicans. By looking it how much better Romney performed in the Battleground states vs the other US states. By performance, lets compare how Romney did vs 2008 when McCain ran. In total votes, Romney achieved 1.3% more votes than McCain in 2008. When we compare Romney in Battleground states and Rest of USA, we get an interesting comparison. In the Battleground states Romney achieved 2.0% more votes but in the Rest of USA only 0.2% more votes. So Romney's Orca and Get-out-the-vote in the targeted important states did achieve 1.8% better results. One could say, that Romney was even without the new Orca voter activation methods, 0.2% more desirable to Republicans in 2012, than McCain was in 2008. And yes, the Orca machine could be said to boost your own voter turnout by 1.8%

Now lets do the same for Obama. Obama's turnout nationally was down 5.9% from 2008. But in the Battleground states, his support was only down by 0.6%. In the non-Battleground states, ie the Rest of USA, Obama's support was down 7.5%. It was the same man, with the same plan, but in the states where Narwhal was deployed in full, Obama's voter turnout was 6.9% better than elsewhere.  

We have two measures, they both show a gain in the same direction (both show a gain in both measures, it would be contradictory if in one case the 'benefit' was found to be a decline in turnout). In both cases Orca's benefit is far smaller than Narwhal's. Lets take an average of these two measures as our handy number for both efforts. For Team Romney, their average gain of voters when Orca is deployed is thus 1.4% of own votes (or half that, 0.7% in a national or state-wide vote 'margin' compared to the other candidate). For Team Obama, the average gain of votes when deploying Narwhal is thus 5.9% (or 2.95% in a national vote margin difference).

We can also now compare their relative effectiveness, by this average we get Narwhal outperforming Orca by a factor of 5.9 to 1.4 ie a factor of 4.2 times more effective. Taking the difference between the two, 5.9% minus 1.4% gives a differential gain of 4.4% for one side (Obama) which means in a national or state-wide vote, when both systems are deployed against each other, Narwhal gains 2.2% vote margin for Team Obama over Team Romney. 2.2% margin gain in the Battleground states is 700,000 votes. Yes, Obama would still have won, if neither Narwhal or Orca would have been deployed, but then, the margin would have been truly razor-thin 200,000 votes across the nine Battleground states, rather than the 900,000 it was now. Of President Obama's critical 900,000 vote margin in the must-win Battleground states of the election, the superiority of Narwhal over Orca, gave Team Obama 7 out of every 9 votes it won, or 78% of the winning margin was due not to Narwhal by itself, but rather, by how much Narwhal was superior to Orca. Had Orca not been in the game, President Obama's victory would have been by over 1 million votes in these states.

Now, was Orca a failure, as has been suggsted in the media - absolutely NOT. It clearly outperformed those states where it was not fully deployed. It definitely boosted Republican voter turnout by a significant number, between 0.9% and 1.9% in the nine Battleground states where it was fully deployed. 

For as what Narwhal and Team Obama could do, their system boosted Obama's vote turnout by a massive anything between 4.9% and 6.9%. And obviously Narwhal was between 3.6 and 5.4 times more powerful in generating more votes, than Orca. Lets call it 4.2 times more effective.

What is the number of votes delivered? For team Orca, using the 1.4% boost in the Battleground states, the actual votes achieved was 200,000 votes. Of Romney's total gain above McCain in the Battleground states, 700,000 votes, 29% were thanks to the supercharged Get-out-the-vote effort that McCain did not have, and half a million votes were gained by Romney the man, his plan and campaign, over McCain and Palin's 2008 attempt, in these nine states.

For team Obama, using the 5.9% boost level, the achieved votes gained was 900,000 votes, essentially the total of the his winning margin in the Battleground states. Yes, if Romney had run his Orca, but Obama had not run his Narwhal at all, the votes in the Battleground states would have been tied. Obviously, if neither side ran their projects, Obama would have won by these nine states by 200,000 votes, not 900,000. But yes, Narwhal can be said to have delivered the Battleground state vote winning margin to Team Obama and re-elected the President. In a Billion-dollar campaign, that $100 million dollar investment in some data nerds and statistics geeks was an investment very well worth it.

But yes then, if you think Narwhal delivered 900,000 extra votes to Team Obama in total (in Battleground states) and we take into account it cost $100 million dollars to build, then the cost per vote delivered is $111.11 - that is quite expensive in terms of vote activation, where a Princeton study in 2008 said that a vote costs $67 when created by direct mail (lots of direct mails needed to generate one extra vote), or as little as $1.56 when generated by SMS text messaging reminders. Of course Narwhal did far more than than just deliver votes on election day, so for example where it helped identify TV advertising audiences more accurately. Time magazine reported that Team Obama's TV advertising buying was 14% more efficient than Team Romney's mostly due to Narwhal. Or yet, another way to spin this issue is, that Narwhal more than paid for itself, by helping target voter contributoins, which grew from $500 million in 2008 to $700 million in 2012. So in reality, the Narwhal team were $100 million dollars in the profit to start with, and could easily pay for all those 900,000 election day activities and still show a net zero cost.

TOTAL CONTACTS BY AMOUNT, CAMPAIGN

As I researched this topic, I also was able to produce a rough draft model of the total contacts across the full campaign, in terms of multiple contacts made, by type and by accuracy. This is not as robust in its data, has far more of my assumptions, often mere best guesses, but I believe this is close to the truth. Remember, first, that we are still comparing only the actual Obama campaign and Romney campaign, not their other supporters, like labor unions and women's groups on the side of the Democrats, and religous groups, the gun-lobby etc on the side of the Republicans. But lets look at the main types of contacts, just nationally, and by number of contacts by each method.

CAMPAIGN CONTACTS BY AMOUNT IN 2012
Contact Type . . . Obama Voters . . Times . . Tot Contacts . . Romney Voters . . Times . . Tot Contacts
Twitter/voters . . . 12.8 M . . . . . . . . 404 . . . . 5,171 M . . . . . . 1.4 M . . . . . . . . . 16 . . . . . . . 22 M
eMail/forwarded . 11.2 M . . . . . . . . . 77  . . . . . 860 M . . . . . 21.5 M . . . . . . . . . 77 . . . . . 1,652 M
SMS texts  . . . . 24.2 M . . . . . . . . . 60 . . . . 1,452 M . . . . . 15.2 M . . . . . . . . . 12  . . . . . . 182 M
Facebook/voter . 18.6 M . . . . . . . . . 27  . . . . . 503 M . . . . . 15.4 M . . . . . . . . . 34 . . . . . . . 525 M
Mail . . . . . . . . . 30.5 M  . . . . . . . . . 6 . . . . . . 183 M . . . . . 23.3 M . . . . . . . . . 12 . . . . . . . 279 M
Leaflet . . . . . . . 34.4 M . . . . . . . . . . 3 . . . . . . 103 M . . . . . 46.3 M  . . . . . . . . . 6 . . . . . . . 278 M
Robocall/listen . 10.7 M . . . . . . . . . . 6 . . . . . . . 64 M . . . . . . 8.1 M  . . . . . . . . 18 . . . . . . . 146 M
Call . . . . . . . . . 26.5 M . . . . . . . . . . 5 . . . . . . 132 M . . . . . 21.1 M  . . . . . . . . . 2 . . . . . . . . 42 M
YouTube . . . . . . . n/a . . . . . . . . . . n/a . . . . . . . 66 M . . . . . . n/a . . . . . . . . . . n/a  . . . . . . . 15 M
Home visit  . . . . 11.1 M . . . . . . . . . 1 . . . . . . . . 11 M  . . . . . 5.4 M  . . . . . . . . . 1 . . . . . . . . . 5 M
TOTAL CONTACTS . . . . . . . . . . . . . . . . . . . 8,546 M . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,148 M
One-directional media contacts . . . . . . . . . . . . 417 M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 719 M
Digital interactive contacts  . . . . . . . . . . . . . . 7,986 M . . . . . . . . . . . . . . . . . . . . . . . . . . . 2,382 M
Interpersonal live (home visit, phone call) . . . . . 144 M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 M
Social media contacts . . . . . . . . . . .  . . . . . 5,674 M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548 M
Contacts usable on election day activation . . 8,194 M . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2,576 M
Sources: Pew, YouGov, CBS/NYT, NBC/WSJ, ABC/WaPo, Inc, NY Times, DailyKos, Huffington Post, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

Then lets look at the two teams and how their top means of communication compare, by ranking them by the percentage used:

CAMPAIGN 2012 OBAMA VS ROMNEY TOTAL CONTACTS BY PREFERENCE USED
Obama Contact . . percent  . . . . . . Romney Contact . . percent
Twitter (voters)  . . . 61% . . . . . . . . eMail (forwarded) . . 52%
SMS texts . . . . . . 17% . . . . . . . . Facebook (voters)  . 17%
eMail (forwarded)  . 10% . . . . . . . . Mail . . . . . . . . . . . . 9%
Facebook (voters) . . 6% . . . . . . . . Leaflet . . . . . . . . . . 9%
Mail . . . . . . . . . . . . 2% . . . . . . . . SMS texts . . . . . . . 6%
Calls . . . . . . . . . . . 2% . . . . . . . . Robocalls (listened) . 5%
Leaflets . . . . . . . . . 1% . . . . . . . . Calls . . . . . . . . . . . . 1%
YouTube videos . . .  1% . . . . . . . . Twitter (voters) . . . . . 1%
Robocalls (listened) . 1% . . . . . . . . YouTube videos . . . . 0%
Home visits . . .. . . . 0% . . . . . . . . Home visits . . . . . . . 0%
Sources: Pew, YouGov, CBS/NYT, NBC/WSJ, ABC/WaPo, Inc, NY Times, DailyKos, Huffington Post, analysis by TomiAhonen Consulting Dec 2012
This data may be freely shared

A dramatic difference clearly, Romney's team did the majority of its contacts using eMail and Facebook. Team Obama did its most using Twitter and SMS. Romney also used Mail, Leaflets, SMS and Robocalls. While Obama only really used eMail and Facebook to significant degree beyond the top two. Very drastic distinctions between the two campaigns. Clearly, Romney's is more suited for older voters and Obama's more for young voters, but also all four of the primary means used by Team Obama were suitable for forwarding and digital distribution; only half of the favored six means of contact by Team Romney were suitable for digital redistribution to friends and contacts.

AND WHAT OF THE DELUSIONS?

We know now, that during election day, up till at least 5PM of Tuesday, the Romney Orca team was convinced that Romney was winning in the election - and they thought they were winning quite easily in many of the Battleground states above and beyond the minimum they needed. As we learned, as Romney prepared to meet his supporters, they had a big victory celebration planned with massive fireworks and Romney didn't even bother to draft a speech in case he'd have lost. The shock of losing was so total, apparently Ann Romney was crying for several days afterwards. They were so convinced they would win. Why was that? And bear in mind - so did the Obama team.

Jim Messina admitted after the election that on Friday leading into the weekend, he had looked at his data (from Narwhal) and was convinced that the election would be won. On election day, when President Obama asked his three top campaign bosses how would the election go, all three - David Axelrod, David Plouffe and Jim Messina said Obama was gonna win, and they all picked every single one of the states correctly except one - Axelrod thought Obama would also win North Carolina (which he lost) getting all other 49 states correct. Messina thought Obama would lose Florida (the narrowest win for Obama) while getting the other 49 states right. And Plouffe picked all 50 states correctly, including Florida which was so close, it wasn't decided until four days later. So both sides were totally confident they would win. Clearly, we see with perfect hindsight, that Team Obama's confidence was 100% well-placed and Team Romney's optimism was overly optimistic and one could say 'delusional'. Some of their consumer survey data was recently leaked with quite hilariously absurd voter projections for some of the swing states.

But yes, what of Romney team's expectations. Lets again ignore the hype and hysteria and be driven by the math. Boston believed that they had a turnout advantage and that their voter enthusiasm was greater and that Obama's turnout was down. The Romney team believed - their data told them - that their turnout was up and outpeforming John McCain's. This is true, it happened. Romney got nearly 900,000 more votes than McCain did in 2008. The Romney team understanding, of the Republican voter turnout and enthusiasm was accurately picked up.

What of the Democrats. Well, in 2008 Obama's vote total was 69.5 million votes, now four years later, he got 4 million votes less. The Obama turnout was severely down from 2008 - by 6%. Definitely yes, there was a marked downturn in the enthusiasm level of the Democrats. When we add to this, the actual bonus performance of Orca - adding 1.4% more bonus votes in reality to Team Romney's performance - this while Orca stumbled and crashed during election day - before the vote, certainly Team Romney had every reason to feel confident that they had a 'secret weapon' - which did even in its injured state, help boost the Republican turnout even more. Nobody could guess before voting started, that Orca would crash so badly all day, and no doubt, some of the troubles weren't even discovered until late in the day on Tuesday evening - like problems of volunteers in North Carolina and Colorado being totally blocked from the system, and that even their newly assigned replacement login names and passwords did not work.. Orca was expected to deliver more Romney turnout - and it did. But not many outside of Boston's data team would have known how badly Orca was struggling on the day.

Team Romney had witnessed a heightened enthusiasm for their candidate after Mitt Romney's first debate performance. That enthusiasm lasted up to the election and definitely delivered an increase in Republican voter turnout. That assumption and expectation and analysis of what was happening in the electorate was accurate. The campaign also had observed and measured a decline in President Obama's support, compared to 2008 and this was quite significant and it truly happened also. Finally, Boston also felt they had further gains from their Orca secret weapon for election day, a boost that had not been shown to the public and nobody could anticipate exactly its ability, outside of Team Romney - so they felt it was a bonus secrete weapon to add to their winning margin. And also yes, the facts really were, that Orca, even in damaged mode, did boost Romney's performance.

The belief by Team Romney, so strongly, that they would win - was based on solid science and correct reading of existing trends. What of course, any forecaster would say, is that the exact points where those trends would overlap, would be the critical point. When does Obama's downward trend in support dip below the surging Romney support. So that could not be known perfectly. But we have seen, that the numbers suggest, that were it not for Narwhal - a surprise nobody could anticipate that it would be so huge in its effect - boosting Obama's voter turnout by 5.9% - the election could easily have been tied, or then hanging by a tiny thread, one way or the other. If we remove Narwhal from the equation - and add 'no faults' to Orca - if Orca had performed flawlessly on election day as everyone on Team Romney expected - then yes, a Romney win was very reasonable to expect. What - as RealClearPolitics polling on election even said = should have been a 0.7% election, was derailed by two datamining machines. The Romney machine stumbled, and the Obama machine stunned everyone by its massive performance. That tipped it. But again, to accuse the Romney team of being delusional, no that is not fair. When they examined all available data, it all pointed clearly to a decline in Obama votes, an increase in Romney votes, and they knew that Orca would further boost those. Their belief, that Mitt Romney would win the election, was not misplaced hope. It was well founded in the reality which Team Romney could observe at the time.

Meanwhile on Team Obama's side? The difference is again a complete generation and a whole order of magnitude if not several. Narwhal was not test-run on election day. Narwhal had been running in full mode for months, been in development for more than a year and its users had been trained on it long before election day. Narwhal had the world's largest consumer insights panel - voter opinion survey panel - to drive its data and had unprecedented accuracy. Best of all, Chicago had been able to test the accuracy of Narwhal for months, they knew exactly how accurate it was. That is why Axelrod was on Morning Joe TV show promising to shave off his trademark moustache of something like 40 years, if Obama lost some specific Battleground states that were mentioned on the TV show. Axelrod really knew what he knew. This was not a fair fight. Its like playing Trivial Pursuit with anyone else, however genius and great memorizer, if you have access to Wikipedia and the other person does not. You win every time. Team Obama had the world's most accurate information about this election, and when Narwhal's simulations and projections said, Obama will win Colorado, Florida, Iowa, Nevada, New Hamsphire, Ohio, Virginia and Wisconsin - eight out of the nine - and lose North Carolina. That is exactly what happened. Not approximately what happened, exactly what happened. I wish I had such a machine in my pocket haha, even I got one of those 9 states off, and I'm pretty good at this forecasting game.

PS - If you didn't get enough of my analysis of the Election of 2012, and would like a fair, unbiased, non-US view based on the numbers and math - of how the 2012 election was won and lost, and where, and by what issues, based on deep analysis of the Exit Polls in each of the states, without any of the partisan spin, you may enjoy this Final Analysis of How 2012 Election Was Won and Lost

All info, stats and tables in this article may freely be shared. And about me - I have no dog in this fight, I am a Finn, I can't even vote in US elections, I am in this purely out of a fascination with the subject matter on a personal level - US politics - and professionally the datamining and customer insights, targeting etc part - while usually I do that kind of work on telecoms customer databases, segmentation, and subscriber billing systems. 

Comments

[Stoli89]

Say what you will, but Mitt Romney the "businessman" was soundly defeated by the "community organizer". The leadership and management skills claimed by the leveraged buyout specialist were clearly not all they were cracked up to be.

[Winter]

Tomi
I believe that in "Enter the colossus" you start mixing up Boston and Chicago from around:
"Boston was not afraid to squeeze to get more juice. "

[Winter]

About the delusional part of the Romney campaign.

Nate Silver did a simple count from published polls, just like you did. And he got the same results, a win for Obama.

Boston named Orca as the Narwhal killer. So they knew about Narwhal. They should have known who worked on Narwhal and as it was tested before the elections, they should have known that is was powered way beyond the 2008 campaign. They could have guessed that Chicago with a 4 year head start would go far beyond 2008 in every direction. Any reconnaissance (eg, just asking voters) would have made it clear what tactics Chicago was using to get out the votes. And I cannot imagine that Boston did not have some moles inside Chicago.

All these were known facts: Polls showing a consistent advantage for Obama, a massive targeted get-out-the-vote campaign reaching 1 out of every 4 Democrats, a truly massive ground campaign, and the size of Narwhal.

If you know all this, and still are completely convinced that you will win it easily, then you must be delusional.

The only alternative I see is that they were banking on some dirty tricks that did not play out (eg, the secretive voting machine updates in Ohio and the 6 hour lines in Florida).

But whether Romney was delusional or a criminal, both would have made him unfit to be a president.

[Decade]

Of course, you know about the technology behind Narwhal:
http://arstechnica.com/information-technology/2012/11/built-to-win-deep-inside-obamas-campaign-tech/

[v-pills yorumları]

Absoultly of course!

[http://truthaboutabstrial.net/download/]

Frenchie puppies for ownership are satisfied animals and you will delight in their firm. It can be an on-line book, about 142 PDF websites. The foundations ended up being peaceful for imports of vehicle products that prompted Western manufacturers to use making use of advanced technology and robotics for vehicle producing.

[Fitzpatrick]

Each individual contact I delivered him with queries was solved between 1-a couple of days. choice onion-'sliced. increasing.

[good apps]

On the other hand, this desire is just not practical, in the end you should continue doing the difficult do the job day by day. You can actually discover Spanish if you have a very good teacher or pick the right offer that allows you to educate yourself on the language all by yourself. Fruit-apricot, blackberry, clementines', damsons, grape fruit, guava, lemon, mandarin fruit, melons cantaloupe, plums, plums, raspberry.