This is a report on the mass email I sent during this campaign. Summary: 23,000 recipients, 2000 bounces, 30 unsubscribes, 15 annoyed recipients, 25 friendly respondents, 4800 email viewings, 370 click-throughs to my site. Cost: $15 ($1 voluntary penalty per annoyed recipient), 20 hours.
About ten days before the election I found out that the most recent voter registration data available from the county included email addresses for 5 to 10 percent of the voters. The form by which voters give their email address was no longer online, as the online registration deadline had passed. However, a county elections official opined that the fraction of voters who voluntarily gave their email address would have a reasonable expectation of getting campaign emails. Local party activists were divided over whether I should do the mailing, out of concern for the party's image or even that my ISP might cancel my service.
Spam?. The former concern was addressed by adding a disclaimer about the LP to the email. The email I sent is at http://marketliberal.org/Email.html. To address concerns over spam, it included this postscript:
The voluntarily-disclosed email addresses of registered voters are available to any candidate by state law [EC2194(2)]. This will be my only such email for this election. To never receive mass email from me again, click here. This email is from me, Brian Holtz, and not from any unit of the Libertarian Party. I personally bond this email as not spam. If this email was so unwelcome that you want to sanction me, then click here. For each recipient who does so, I will donate $1 to SpamCon's anti-spam legal defense fund.
The opt-out and penalty links are basic mailto URLS, and are a convenient way to collect responses. Political emails are in fact exempt from the federal CAN-SPAM act, but my email was in compliance except in one respect. The law requires commercial mass emails to include a physical address of origin, but I didn't have a campaign P.O. box and chose not to include my home address.
Beacons are a technique to track the occasions on which a piece of content (web page, email) is downloaded for display to a user. To the image in my email I added "?visitor=NNNNN" at the end of the image's URL. The image still loads as usual, but the extra information is logged in the web server's access log whenever the image is retrieved by the user's browser or email program. If you make the information unique for each recipient, you can track viewings on a per-recipient basis. If you make the information unique merely to the email campaign, then the beacon will just give you aggregate numbers, and can double-count re-viewings by the same user.
Link-tracking adds a similar "?visitor=NNNNN" tag to the URLs in the email. In a way similar to beacons, you can track user clicks on the links in your email.
Rate Limiting. I had hoped to script the sending of my emails, so as to generate a unique visitor ID for each email recipient. Unfortunately, tests showed that my ISP (Comcast) rate-limited outgoing emails to just a couple per minute, and so to get the emails out in time I had to send to as many recipients at a time as possible. At first I sent to 100 recipients per message using Outlook, but at about 1000 recipients Comcast shut off my SMTP access for two days. I switched to sending via Comcast's web mail interface, which actually sped things up as it allows more recipients (150) than does Outlook. I sent to the remaining 22,000 recipients this way.
Recipients. I extracted the email addresses from the county text files by using standard Unix commands (grep, cut, sed, join) available on Windows via Cygwin. I was concerned about spamtrap addresses registered by anti-spam activists, so I culled out about 70 of them (that matched the regular expression 'election|vreg|spam|vote') and sent them a custom email. (No negative responses, three positive.)
Results. Of 23,000 recipient addresses, I got about 2000 bounces. 30 users invoked the unsubscribe option, and another 10 invoked the voluntary $1 anti-spam penalty. 5 more sent hostile emails, although one responded to my follow-up by apologizing. There were about 25 friendly personal replies, and about 4800 downloads of the image in my email. This tends to overstate unique viewers due to re-viewings of the email, but also tends to understate unique viewers due to text-only mail readers and users who turn off images. (For comparison, I received 90 image downloads for an email to the 50 voters who had sent me an electronic form letter about "Big Tobacco".)
The links in the email generated about 370 click-throughs to my campaign site. (About a dozen of the clicks were my own tests, so the numbers below add up to more than 370.) The clicks were distributed as follows:
218 Bet.html55 Platform.html45 front page, linked via candidate image25 SpecialInterests.html20 Mortgage.html19 Lesson.html
Two of the three attempts to meet my $5 voter challenge were received soon after, so I attribute those to the email. The last-minute email did not solicit contributions, and none were received. If the email had been earlier in the election cycle, I would have included a mailto URL to allow people to subscribe to receive campaign updates.
Instant Messenger. I didn't get any instant messenger replies that I can attribute to the active instant messenger icon included in the email. My campaign web site has included an active Yahoo instant messenger control throughout the campaign, and I'm signed in to Messenger for a very high percentage of my waking hours. Despite this, I only received about five or six contacts this way. However, the contacts have tended to be very high-quality, resulting in non-trivial conversations leading to ongoing communication.
Brian Holtz
2004 Libertarian candidate for Congress, CA14 (Silicon Valley)
http://marketliberal.org