Wednesday, December 22, 2010

Should campaigns hide LGBT people?

I worked for No On 8, and one common criticism of the campaign was that it didn't show ads with same-sex couples. Many people said that Prop 8's passage showed the mistake in this strategy. The campaign strategists who made this decision, however, pointed out that No On 8 might have lost even more badly if they had shown LGBT people in ads. They argued that the ads targeted a narrow "movable middle" of people who hadn't made their minds up, and that these people were uncomfortable with images of same-sex couples. Showing supportive straight people, the argument went, would build support; showing same-sex couples would actually have been harmful.

Who was right? We'll never know. But there is a better way to answer this question than focus groups and "expert" consultants: randomized controlled testing.

I'm helping design a randomized controlled test for Equality Maine, Maine's LGBT advocacy group, that will look something like this. They'll randomly select 30,000 Maine voters, and randomly divided them into three groups of 10,000. (10,000 happens to be the number necessary for statistical significance.) Group A will receive a piece of mail with a photo of a same-sex couple and text trying to convince voters to support same-sex marriage. Group B will receive a piece of mail with a photo of a straight couple and similar text to Group A. Group C will receive nothing.

A couple days after voters receive the mail pieces, a survey company will call all three groups and ask them if they support same-sex marriage. (Shameless plug: If you run a survey company or a call center, check out my company's predictive dialer, Impact Dialing.) Based on the survey results, we'll be able to see which mail piece is the most effective. We'll also be able to see if some groups of people respond better to one message, and some groups better to the other message. For example, maybe older voters have so much ingrained homophobia that images of same-sex couples will make them less supportive, but images of straight parents of LGBT people will appeal to their sympathies. And perhaps younger unmarried voters who see images of a committed same-sex couple that can't marry will be able to relate more to that mail piece.

We'll never know if No On 8 made the right strategic choice by not showing same-sex couples. But going forward, we have better tools to figure out the answers to this and other questions. Leave a comment if you're working on a voter contact or fundraising campaign and want to test a message, a medium, or anything else.

Wednesday, December 8, 2010

You are being watched (or, Free burritos)

I usually hate junk mail, but the other day, I got a coupon for a free burrito in the mail. Thanks, Chipotle! Every once in a while, I'm pleasantly surprised to get something I wanted that I didn't ask for. What if all of your unsolicited mail was for things that you wanted? You'd be pretty excited - you'd get less in your mailbox, and you'd actually look forward to the things you did get. And marketers would be pretty excited, too - they'd spend less money on their mailings, and the ones they did send would produce the results they wanted.

To realize this vision, though, marketers would need to know more about you. They'd need to know what kinds of products you purchase (mexican cookbooks?), what magazines you subscribe to (food magazines?), your income (do you have enough disposable income to eat out much?), your family size (do you have a spouse and kids to feed?),... and the list goes on. How would you feel about them having that kind of information about you?

Well, if you're not comfortable with that, I've got bad news for you. Data companies already compile all this information and more. I'm browsing through a catalogue from one of the biggest data companies, where I can buy lists of people who have allergies, who buy organic food, who use credit cards, and who vacation in Canada. There are over 1,000 different data points on almost every US consumer.

A couple posts ago, I wrote about how non-profits can use modeling to better target their donor acquisition strategies. Such modeling is only possible with the use of these vast sources of individual-level data. This data can also be used to target voter outreach; for example, my company is helping run a project to figure out what people can be persuaded to support marriage for same-sex couples. (More about that later.)

But is it worth it? Are you comfortable with data companies tracking you in exchange for free burritos? Does it seem harmless - there are over 220 million American consumers, so who could possible care about you except as part of a large cohort? Or does it seem dangerous - what if somebody is out to get you, and uses this data as blackmail? Leave your comments below!