Data Wars: DNC vs RNC

Micah L. Sifry | October 23, 2007 - 10:35pm | 3 comments | Email This!

The Republican National Committee just launched a new online game called Scariest Democrat, a Halloween-themed contest attacking the Democratic presidential field that invites visitors to "click on the Scariest Democrat." Complete with creepy sound-effects, the site drew 65,000 visitors by 4pm on its first day, and RNC e-campaign manager Cyrus Krohn told Marc Ambinder that "nearly 9,000 people had spent the time to give the RNC their e-mail address in order to vote." (The winner, by an overwhelming margin, was Hillary Clinton.)

But Josh McConaha, the Democratic National Committee's internet director, isn't impressed by these numbers. He compared the RNC's latest push to a recent online letter-writing campaign the DNC did around the S-Chip bill, and his numbers suggest the Democrats are still way ahead of the Republicans in the online engagement game. He told me:

On our SCHIP work, the most recent we've done, the DNC had roughly 68,000 letter writers on 120,000 page views (114,000 were on the launch day if we want to compare it to the RNC). We required users not only to know their zip+4, but to write their own letter and fill out a long form of information. Even with all the required information, our bounce rate was only 43% (which is high, but not for needing that level of information).

[Bounce rate refers to the number of visitors who come to a webpage, but fail to do anything.]

The RNC has 9,000 voters on 65,000 launch-day page views. They require only first, last, email and zip -- the most basic information you can ask for. They had an 86% bounce rate.

Our list stands at about 3.5 million people; the RNC has, if I remember correctly, claimed around 7 million (the difference is largely due to buying email appends to voter files in 2004).

So if you email 7 million people and 9,000 people respond, you've only got a .12% response rate.

If you email 3.5 million and 68,000 people respond, you've got a 1.94% response rate -- a little over 16 times better.

It's not totally fair to compare same-day stats with overall stats, but 95% of all action happens on launch day (which is clear given our page view numbers on SCHIP). So, let's just say that the RNC will see another 6,000 votes, bringing them to 15,000 overall (60% day of, 40% after -- even though the real numbers are 95 and 5). Their push, which they're actively bragging about to the media, was only 22% of what our last push was -- and that's spotting them an unrealistic increase in post-launch traffic and a list that's double the size. Divide for their list size, and their thing was 11% as effective as our last thing, and probably far less so if you took into account the amount of effort required by the user.

This isn't the first time, by the way, that a DNC email campaign has generated impressive numbers. Back in early September, McConaha told me that they emailed their entire list asking people to use their PartyBuilder tool to start county-level chapters, and produced a 25% increase in local groups--more than 800 ones--in 48 hours.

I've emailed Cyrus Krohn asking him for his comment on these numbers (and coincidentally enough I'm going to be on a panel with him and McConaha tomorrow!). Stay tuned.

Our readers also like:

The Silent Online Majority
90 recommended posts
Categories: 

Comments

Apples and Oranges...

by Patrick Ruffini on October 24, 2007 - 1:47am
The difference here is more easily explained by the different sign on processes. The RNC application is a two page process in which you: 1) Vote 2) Give your email address Is the higher "bounce" rate anything more than natural attrition you'd experience on a secondary page? Especially since when you land on the page, it says, "You have selected Hillary Clinton" or whomever in big type, followed by a request for an email address in smaller type? People might have missed the request, or given that the results are posted before you vote, figure it's not worth the hassle to give your address. When I was at the RNC, a similar survey action generated more than 80,000 votes with a very low bounce rate. (Actually, this was a more intensive, multi-question survey.) However, a key difference was that we did not require someone we knew to be on our list to give their email address to vote. It was a one-step process for most people. So our action rate was higher as a result. The key though is that the RNC has built something viral and worth talking about in the press. Because it wasn't a standard, cookie-cutter, issue-based call to action, it attracted attention beyond the email list. I'd be willing to bet that most of the people who signed up were new to the RNC. How many people who signed the DNC call to action were new to the DNC? Probably relatively few. It was probably the same people signing with the email address already on file, or the same people signing with a different email address.

Totally agree...

by mcconahaj on October 24, 2007 - 10:04am
It is apples and oranges. The RNC process is much simpler than the DNC process for what we're comparing. There was 1) vote and 2) give your email address. The DNC had... 1) Give a form of your address or zip+4 (because of traffic, the geocoding process gives way to making people use the USPS lookup, so that's 1.5) 2) The same fields as the RNC, plus full address 3) User-supplied subject and fully body (no defaults) 4) Preview and send But if we're looking at past campaigns and 80,000 is the best number, then the multiple 100,000+ campaigns or the 250,000+ campaign we did on Path to 9/11 and ABC should come into the mix. "Viral" is a matter of context, I suppose. It probably went viral in page views, which is why the bounce rate is so high, but even as by "going viral" it still underperformed compared to an average send for the DNC. And if it did go viral, then the RNC list performed even worse than the math shows -- how much worse depends on just how viral. Let's suppose the bounce rates for the two lists were equal and that all action came from the email list -- 43% bounced for the DNC just from email, 43% bounced from the RNC just from email. That means that there were only 21,000 page views from RNC list, versus 120,000 for the DNC. With those numbers, it's a 0.3% click-through rate for the RNC and a 3.4% click-through rate for the DNC -- over 11 times better. But we know those page view numbers aren't accurate on either side, simply because there's always some amount of "viral" to everything...it's just what we have to work with. So let's look at the value of "viral" -- new names. My experience has been that new names range from 20% to 40% of all action takers (but it can range to either side of that). Let's give the DNC the low end -- 20%. Let's give the RNC the high end -- 40%. That means that the DNC had 13,600 new names and the RNC had 3,600 new names. In other words, given a 20% difference in new names, the DNC still had just under 4 times as many new names as the RNC's "viral" campaign did. Without more data it makes it educated speculation, but if the argument is that the RNC's was easier and it also went viral, then it would mean that the list performed even more poorly than the original calculations.

Response from RNC

by cyrusk on October 26, 2007 - 9:05am

I’m pleased to see the DNC has taken such interest in the RNC’s web efforts. The RNC web team is going to continue experimenting and assessing what works and what doesn’t, as all web practitioners should. We’re still in the infancy of this medium and nobody has mastered the web yet – not even Google.

I understand the risks associated with sharing web statistics publicly because they invite the type of analysis Josh offered. Since we stated our figures at 4 p.m. day of launch that was all Josh had to work with--but keep a few things in mind. The Halloween contest launched at 9 a.m. ET so he was only working off of 7 hours of stats. I believe the DNC’s stats for their SCHIP work were based on a full day, so let me revisit our Halloween program now that we have 24 hours of data. After exactly 24 hours, scaredem.com generated more page views than the DNC’s SCHIP project and acquired over twice as many registrants as was cited in our initial media outreach. Josh’s projections weren’t far off, but the raw numbers he based his calculations on were not accurate. I won’t divulge the size of our list but it’s larger than the DNC’s (thanks Josh!). Maybe that should be our next web game – guess the size of our list.

I’d caution anyone against using this one event to be the determinant in assessing our online engagement. There are several other metrics associated with this particular program worth taking into account to gauge the value of our effort. For instance, how did our media and blog outreach impact additional traffic to the site? How many new visitors did the aforementioned attract to gop.com and what was the engagement rate of that new audience?

I will tell you that over the 36 hours following the release of our Halloween project, we attracted over one hundred thousand new visitors to the site and 68% of that audience viewed 3 or more pages. We introduced new readers to our web site and they consumed a respectable amount of content. More importantly in the context of this thread, those users chose to sign up for our e-mail alerts, generating over 50,000 new e-mail addresses above and beyond the 16,000 addresses we acquired from the contest, bringing our daily total of new addresses to over 66,000. According to Josh, the DNC generated roughly 68,000 letters from a pre-existing audience which didn’t really speak to or attract new “customers”. As Micah titled his piece “Data Wars”, I’d say this thread illustrates we have many battles ahead of us. But I’m claiming victory in the Halloween vs. SCHIP exchange and I look forward to seeing who wins the “Data War” on the condition we continue to conduct civil discourse over analytics.

Onward & Upward.

Cyrus Krohn
eCampaign Director
Republican National Committee