Wednesday, July 6, 2011

Interview with Political Data, Inc.

I was in LA the other day, so I stopped by the offices of Political Data, Inc. (PDI) to say hello. Gary Brown, their Vice President, and Lindsay Hopkins, their Director of Software, were kind enough to let me interview them about PDI. Here's what we talked about.

Michael: PDI is the biggest vendor of political data in California. Can you tell me about the history of the company?

Gary: Jim Hayes started PDI in 1987 and revolutionized how political data vendors work. Before PDI, you would go to a data company and say "I'm running in the 26th Congressional district", and they'd go to the county, get the voter file and the vote history, and compile a database for you. Jim said, I'm going to build the whole state file, so that when somebody calls, it will be ready.

Michael: It's amazing - we take that for granted now.

Lindsay: Today, somebody can place an order, and it will be ready in 10 seconds. The thing that takes the most time is dealing with the client's paperwork. It takes more time to create a password for our Online Campaign Center than it does to get the data. That's the thing that Jim revolutionized.

Michael: Gary, you've been with PDI quite a while. How have things changed since when you started?

Gary: I joined PDI in 1995. Pretty much everything has changed since then! Those days, data was on paper - we'd print millions of paper labels, mail people their voter files, and all communication was by fax. Information was stored on reel to reel tapes. In the 80s, movie studios weren't doing well, and went to local businesses in Burbank offering time-sharing on their mainframes. We were on Warner Brother's lot, in two old trailers that used to be dressing rooms for movie stars. We'd crunch data on their mainframes, and then ship boxes of paper lists and labels from the airport nearby. We still have a $50,000 printer that we can't even give away now.

Lindsay: These days, we're moving everything online with our Online Campaign Center. People are asking more and more for tools that enable them to help themselves.

Michael: I love that philosophy - with Impact Dialing, we really focus on self-service.

Gary: We're getting ready to release the next version of the Online Campaign Center. A lot of other companies built voter file software in the mid-2000s, and we're excited that we're the first company to be building the 2nd-generation version of our product. That's the nice thing about being a smaller company - we can turn on a dime and use new technologies that weren't available before. The key thing about our software is that we're first and foremost a data vendor: we know exactly how our data is supposed to be used, and our software is basically just a delivery mechanism for that data. And the great thing is, when we update our files, everybody using the Online Campaign Center instantly gets the new data. You don't even have to think about file updates - it's just part of the product.

Michael: Nice. Tell me a little more about your data.

Lindsay: A big challenge is to get people to think about the value and importance of good data. It's so important to know that you're getting good addresses, the correct precincts, accurate polling place data, and so on.

Gary: In the last 45 days of an election, we're doing about 200 different county file updates, 500 absentee data updates, and putting in polling places. There are 58 counties, and once absentee ballots are out, most update their files to report who has voted 3 days a week, and some as much as 5 days a week. People ask when they'll be able to get absentee data, and the answer is always that we have it available 10 minutes after we get it. Getting all of this data requires so much coordination, I still can't believe that we're able to pull it off.

Lindsay: Some campaigns go to the county and buy the voter file directly, but then they realize there are so many things they don't know about how to update the addresses and phone numbers, where to get polling place data, how to find out who has already voted absentee, or how to manage a database for all of the records. We have 14 people who are California voter file experts - that's all they do. We want to make things as easy as possible for our clients so that they can focus on the hard work of talking to voters.

Michael: Can you share a story about a campaign that used your data effectively?

Gary: A few years ago, a campaign in a heavily Vietnamese area felt like they weren't reaching everybody, and asked us to get cell phone numbers. At first, I was skeptical, but we got the numbers, and I was really surprised how effective it was. A lot of Vietnamese families in this area ran small businesses, and if you called their landlines, you'd never reach them - they were always at work the times that we typically call people. Now, we offer wireless numbers standard, because you can reach so many younger voters and non-white voters who you can't reach any other way.

Michael: Thanks for sharing the story of PDI. It's exciting to see where you folks are going.

Monday, June 27, 2011

Your phone in a browser

Last week, Google announced that they had begun working "to enable Chrome with Real-Time Communications (RTC) capabilities" - in other words, to allow your browser to send and receive streaming audio and video. Instead of having to download to download software to make calls, such as with Skype, or install a special plugin, such as with Gmail's voice and video chat, these applications could be built directly into the browser. Whoa!

The beautiful thing about WebRTC, as the project is called, is that it will give developers a standard way to write web applications that use real-time audio and video, using the same Javascript programming language that they already use to make web pages more interactive. This should open the door to more of these types of applications, as it will be much easier to develop them. And since WebRTC is sponsored by Mozilla, Google, and Opera, users of their  won't have to install any special software to use web applications that take advantage of WebRTC.

One obvious application of WebRTC is to build a phone in your browser. Perhaps Impact Dialing will use it so that you don't have to use a separate phone to connect to our hosted predictive dialer - instead, you'd connect directly through our web application, using WebRTC.

What else could WebRTC be used for?

Tuesday, June 14, 2011

The future is self-service

At the airport in New York the other week, I went to the counter of a Chinese restaurant to order some food for the long flight back to California. Nobody was at the counter, and nobody seemed to be paying any attention to me. And then I realized that the computers on the counter were facing the customers, not the staff. I was shocked. What a brilliant idea! How could nobody have thought of this before? Everywhere, businesses are finding ways to turn the computer to face the customer: all banks now have ATMs, most airlines have self-checkin kiosks, some grocery stores have self-checkout stands, and now restaurants are experimenting with lettering customers place their own orders.

The key change that's allowing self-service to expand, and the sticking point that causes it to fail and frustrate when done poorly, is user interface design. Remember old checkout interfaces at grocery stores? The green text on white screens, the cashiers furiously keying in codes? The amount of time needed to learn a system like that made self-service impossible. Self-service checkstands now have big colorful buttons with pictures, and, most importantly, fewer options. In the best-designed self-service grocery stores, you just scan the barcodes of all of your items. It doesn't get much simpler than that. Where self-checkout doesn't work as well is where this simplicity breaks down - most notably, when you have to tell the system which fruits or vegetables you are weighing. Some grocery stores fix this by packing produce and putting barcodes on them.

Impact Dialing is doing something similar with auto dialing. It used to take an expert to set up an auto dialer: there were a massive number of settings and options to configure, you'd have to install special hardware and software in your call center, and you'd need consulting and training from your auto dialer company to set things up and learn how to use the system. We've taken away all of that pain: made everything web-based, so there's nothing to install; simplified complicated settings and used plain English in our interface; and made the whole thing available to anybody who wants to try it out.

Are you creating beautifully-designed self-service tools? Have you used any that made your day? Or that made you angry because they didn't work as well as having a real person work with you? Leave a comment!

Wednesday, June 8, 2011

Auto dialer terminology (or, Wow, this is confusing)

Ever wonder what the difference is between an auto dialer and a predictive dialer? Can't keep your robo calls and your outbound IVRs straight? It's all very confusing.

In designing Impact Dialing's website, I spent a lot of time researching what search terms people actually use in the world of making lots of phone calls. I chose the the terms below based on this research and my own experience.

The term "auto dialer" is the most confusing phrase of them all, and a good starting point because of its generality. I did a Google search for auto dialer and came up with a few excerpts from its snippets:

  • "Software for broadcasting voice message by phone"
  • " automated phone dialer for voice message broadcasting..."
  • "An autodialer will call one or more programmed phone numbers..."
  • "... allows for hands-free calling..."
There are a lot of terms and ideas thrown around here. About the only thing they all have in common is that they involve automatically dialing a list of phone numbers. At Impact Dialing, we refer to all of our services as "auto dialer products", to reflect the different meanings that different people attach to the phrase "auto dialer". 

So if auto dialer is a vague, nebulous phrase relating to anything that automatically dials a phone number, then what are concrete examples of auto dialers? One term that showed up a few times in the above Google search is "voice broadcasting". Voice broadcasting is a kind of auto dialing in which a pre-recorded message is played when somebody answers the phone (or is left on an answering machine). Voice broadcasting messages are also sometimes called "robo calls". 

Pollsters often use robo calls to survey people, by asking them to press a keypad on their phone corresponding to different choices: for example, "Press 1 if you play to vote for the Democrat, press 2 if you plan to vote for the Republican." Sometimes people call these robo surveys, but they're more commonly referred to by a more technical name. Gathering information from pressing phone keypad digits is called Interactive Voice Response, or IVR. When it's used on outbound calls (as opposed to inbound systems - "Press 1 for Sales, 2 for Support"), it's sometimes called an "outbound IVR system". 

Our Google snippets showed us that people also use the term "auto dialer" for a system that connects answered dials to live people. The most common term used for live phone calls is "predictive dialer"; when a predictive dialer is offered as an online service, rather than as a piece of hardware, it's called a hosted predictive dialer. Some people also use the term "power dialer" to refer to an outbound dialing system. Although sometimes "power dialer" and "predictive dialer" are used interchangeably, the more technical distinction is that a predictive dialer uses real-time analysis to determine the optimal time to make dials (see this blog post for details), whereas a power dialer simply dials a pre-set number of lines when a caller finishes the previous call. 

So, in summary:
  • auto dialer: anything that automatically dials phone numbers
    • voice broadcasting: when somebody answers, play a message
      • robo calls: another name for voice broadcasting
      • robo surveys: press a keypad button to provide feedback
        • outbound IVR system: another name for robo surveys
    • predictive dialer: when somebody answers, connect them to a live person, and use realtime analytics to determine when to dial again
      • power dialer: use a simple ratio when somebody finishes the previous call
This is not an exhaustive list, but it touches on some of the most common terms and hopefully provides some clarification. 

Monday, June 6, 2011

Smaller is better: designing for usability

Yesterday somebody called to learn more about Impact Dialing and raved to me about how well-designed it is. He had looked at a half dozen other dialers and said that ours was far and away the best-designed and easiest to figure out. I don't claim to be a design expert, but I think we're on to something in terms of our usability, and I want to share our design philosophy.

My experience suffering through my now-competitors' poor design was one of the reasons I started Impact Dialing. For example, on logging in, one system presents the user with a blank screen and a series of drop-down menus. The only way to figure out how to use it is to explore the menus and sub-menus, opening up new windows and navigating deeper and deeper into each window's sub-application. Another dialer tried to be more user-friendly with a step-by-step wizard, but in order to change anything, you have to go through the entire wizard again. Both systems used specialized jargon that non-experts would never understand. From my frustration, I came up with a series of design principles:

  • Users shouldn't need to read a manual
  • Make everything accessible with minimal clicks
  • Use good defaults, so even if you have no idea what you're doing, you can still get set up
  • Use plain English and avoid jargon

I started out by breaking down the task of setting up a predictive dialer: you need logins for your callers, a script for them to read and record answers to, and a list of phone numbers for them to call. I used Mockingbird to wireframe a layout in which these three concepts - Callers, Scripts, and Campaigns (which contains lists and a few other pieces of information) - were tabs along the top. This would present users with everything they needed to access intuitively and with minimal navigation, so that they could work from left to right to set things up the first time.

Even with this simple design, I knew some people still might not think about needing to set up caller logins, so we made a default caller login for every new account. Similarly, new users might not know what their scripts should look like; since many of our clients make political calls, we made an example script and corresponding call results for political campaigns. Finally, I knew that some people still might need a helping hand along the way, so we created a "Welcome" tab that opens when users log in and displays a quick run-through of how to use the system, along with a video guide.

At this point, we released Impact Dialing. We just had the bare minimum features needed for a predictive dialer, which in my mind, was actually an advantage: we had a clear, simple, easy-to-use design, something that none of our competitors can say.

As our product has developed, we've been very careful with how we add new features. We eye anything that adds very much complexity with great suspicion, and try to think of simpler approaches whenever possible. All else equal, we always build simpler features before more complicated ones. And if somebody requests a feature that's tied in with other features we're considering, we'll try to wait to build it until we really understand the whole set of features.

For example, we originally only allowed one set of results per script, so if you had two questions in your script, you'd have to list out every combination of answers in your results list. A few people had asked for multiple results, but we knew that we'd eventually support free text entry and patch-through calls, and so we waited until we were sure we understood all of the use cases we wanted to support. Patch-through calls were going to take a little more work, so we just built out the multiple results and free text, and came up with a fairly elegant interface. We've got some more ideas about how to make the system even more flexible, but even now, it's still as straightforward and simple as our original design. That's how we like to improve things.

The final piece of our design approach is to remove and streamline when we can. For example, we originally required new users to enter their first name, last name, organization, email, password, password confirmation, and agree to the Terms of Service. When we re-designed our site, we stripped down to just email, password, and Terms agreement. If somebody entered their password incorrectly, they could always recover it later. Everything we decided was nice to know but not essential, and we'd rather err on the side of simplicity.

Do you have or know of a well-designed web site or web application? Please share in the comments below!

Tuesday, May 31, 2011

Why our new site is amazing

I recently decided it was time for Impact Dialing to get a well-designed website. There were two principles that were important to me in this new design. First, we needed a consistent look and feel between our web site and our web application so that users wouldn't have to learn a new layout when they jumped from our site to our app. Also, we could share a common codebase between the two, reducing development and maintenance, and making our app load more quickly. Second, we needed to use responsive design, which adapts pages to different-sized screens without using a completely different design. That way, our site would work well on mobile, tablet, and desktop sites, and even work well if some new device or format pops up that uses a completely new screen size. 

Some don't like responsive designs because they usually create a layout for a desktop, and then re-arrange, re-size, and hide page elements for progressively smaller screens. The problem is that smaller devices, with slower connections, still have to download all of the larger code. I found a "mobile-first" template called 320 and up (mobile screens are often 320 pixels wide) with which you create your mobile site first, and then progressively load more and bigger elements as your screen size increases. This way, your site loads quickly on mobile devices, while still keeping the advantages of responsive design.

Best of all, I managed to hire the designer of 320 and up, Andy Clarke of Stuff and Nonsense. He built our new site in less than a week, using 320 and up as a the starting point. Check it out here. If you don't have a mobile phone or tablet to try it with, open it in a desktop window, and then slowly resize the window to make it smaller. You'll see how the site responds to smaller and smaller screen sizes. 

One exciting part of having a responsive site is that our web application now works on phones. Many of the political consultants we've been talking with have been excited about the idea that volunteers can use our hosted predictive dialer with just their smart phone. 

Let us know what you think of the new site!

Monday, May 23, 2011

Impact Dialing wins Lean Startup Challenge!

[Update: we're on VentureBeat!]

A couple months ago, I wrote a blog post about how Impact Dialing uses a "lean" approach to its software development. The post was an entry in a contest to get investment funding that would help grow our business. To my surprise, we won! A few weeks after writing the post, I got an email from Pivotal Labs, a company that helps startups build better software, asking if we were still interested in the funding. Pivotal really liked the way we were building our company and our software, and told me they wanted to help incubate Impact Dialing by providing a desk in their office (we currently work from home) and help hiring another developer (by the way, we're hiring). Next, I talked to the Band of Angels, Silicon Valley's oldest seed funding organization. They too were excited about Impact Dialing's approach and product, and offered to invest in the company. I'm excited to announce that we've taken up Pivotal and the Band on their offers. The most immediate results of their help is that we've just launched a beautiful new website. Make sure to follow this blog (or our Twitter or Facebook) to get updates on what we do next!

Friday, April 29, 2011

How to start a company

I recently met someone with a great business idea who has done a lot of research and wants to start a company. When I was in a similar situation last year, my two big questions were: 1) how do I find a programmer, and 2) how do I find investors? I don't know if I have the perfect answer to these questions now, but I can at least share my recommendations from my own experience. (If you've started a successful company before, have access to angels or VCs, or are a programmer, this post doesn't apply to you.)

Last year, I thought my best choice was to find a developer who would work for free in exchange for 50% ownership. If you have a friend who is interested, that's probably your best option. But if you don't personally know someone who's interested, it's very hard to find a stranger who will believe in you and your idea enough to work for free on it. There are a lot of people with business ideas, and most of them fail. Good developers have better options than risking their time on a venture that, statistically speaking, will likely go under without paying them anything.

So if bringing on a 50% partner isn't an option, then you need to raise money, right? I'd suggest that, if your idea is so great, spend your own money on it. (If you don't have money to spend, then you should probably work on your own financial management first - that's pretty essential to running a business!) You need to be confident enough in your idea and plan that you're willing to put a few thousand of your own dollars behind it. Figure out your minimum viable product, so that you're minimizing your risk, and come up with a marketing plan that you really believe in.

Now you've got a plan and you're ready to shell out some cash to make it real. Unfortunately, you're still not in a place to hire a good programmer. Chances are, anybody good will either want a stable job, or to work for a startup run by someone with a history of success. Instead, you can hire a contractor or outsourced development firm to build your product. Find someone who is used to working with startups - many companies and contractors these days are.

Once you've got your minimum viable product built, congratulations! Sell it, and prove yourself - make back the money you've invested so far. If you can do that, you'll be able to convince your friends and family to invest in your business. With their money, you can hire a good developer, since you've got more cash to pay up-front, and a proven model that will convince them to take equity for lower pay. When you're able to pay back your friends and family, you can decide if you can grow out of cash flow, or if you need to raise larger amounts of money from angels or VCs. If you want to go that route, you'll be able to get investment from good sources and at a good rate, since you've got a good track record.

I hope this is helpful for folks thinking about starting a business!

Wednesday, April 13, 2011

Acquiring new donors, part 2

Before, I walked through an example donor acquisition campaign, and mentioned that non-profits typically rent a list of names and addresses from other non-profits or direct marketing companies. These list sources have a pretty obvious problem: multiple organizations are soliciting the same people. To the extent that people have a limited budget for charitable contributions, organizations end up poaching donors from each other, or the lists' performance degrades.

A different source of prospects is a modeled list. Data companies assemble information on consumers, including demographic data like age and income, as well as buying habits, such as magazine subscriptions or tendency to buy environmentally-friendly products. Non-profits can pay data companies to find out what characteristics their existing donors share. A statistician can then take these characteristics and build a model to show which people who are not existing donors share characteristics with the existing donors. Using this "look-alike model", the non-profit can buy a list of prospects from the data company, and attempt to turn them into new donors.

Next time, I'll go into the process of actually soliciting donations from these prospects.

Wednesday, March 23, 2011

Phone calls are dead

At least, according to this New York times article:
"NOBODY calls me anymore — and that’s just fine. With the exception of immediate family members, who mostly phone to discuss medical symptoms and arrange child care, and the Roundabout Theater fund-raising team, which takes a diabolical delight in phoning me every few weeks at precisely the moment I am tucking in my children, people just don’t call....
Whereas people once received and made calls with friends on a regular basis, we now coordinate such events via e-mail or text. When college roommates used to call (at least two reunions ago), I would welcome their vaguely familiar voices. Now, were one of them to call on a Tuesday evening, my first reaction would be alarm. Phone calls from anyone other than immediate family tend to signal bad news."
In some ways more interesting than the article are the comments, which include some agreement:
"I haven't used a phone for casual conversation since the late 90's. Anyone I really want to talk to for more than 5 minutes, I'll spend time with. Texting is efficient and cheap. What's not to love?"
 and some outrage:
"... there is nothing fashionable or stylish about what is a great indication of a society of persons who isolate from one another. This behavior has led to increasing prevalence and incidence of mental health disorders, which cost the country more than we can afford in costs related to treatment and loss of productivity."
The irony of this article for me was that I was monitoring Impact Dialing while reading it and watching as almost half of the dials we made were answered! Apparently more than just a few Luddites still answer their phones. Next post, I'm going to talk about multi-channel fundraising, so I'll leave with this final comment from the New York Times article:
"I think we make a mistake when we presume that our attachment to a particular method of communication is "right" and that others are "wrong" or somehow a dire reflection of societal ill. I work in an industry where I listen to people in person all day and use the phone to make confidential appointments. I text and email family, colleagues and friends. However my cousin works at home exclusively at the computer and in her spare time wants a phone or in person conversation because she has had enough of screens, keyboards and isolation. We have different lives and different needs to balance. That seems okay to me. And great to have the options."

Thursday, March 17, 2011

Impact Dialing's Lean Startup Model

[This post is part of the Lean Startup Challenge. Post a comment wishing us luck!]

I didn't want to start Impact Dialing at first. I had been looking for a predictive dialer to help Equality California call voters about same-sex marriage, and I couldn't find anything I liked. All the dialers were expensive, difficult-to-use, and not scalable. I once even suggested to the CEO of one of my now-competitors that he look into technologies like cloud computing, which provides computing power on demand, and voice over IP, which can be used to provide telephone lines on demand. He brushed me off. Finally, an acquaintance said to me, "There's a market out there. Why don't you just build it yourself?" 

I didn't have a lot of savings to invest, so I knew that Impact Dialing needed to make money as quickly as possible. I had heard of a company called Twilio that makes it simple to build web applications that interact with phone lines. This would help us build our dialer quickly. Twilio's pay-as-you-go pricing was appealing, too: we wouldn't have minimums or up-front costs, so Impact Dialing could cover its phone bills by requiring customers to pay in advance, and then using that money to pay Twilio. 

Likewise, we chose to run our software on Amazon Web Services. Amazon lets you "rent" virtual servers on-demand, so we could seamlessly grow our capacity as our volume increased. Services like Twilio and Amazon have made it possible for people like me to start companies with minimal funding. 

My business partner, Brian, who does our programming, decided that to program with Ruby on Rails, an application framework that makes it incredibly fast to build web applications. We also used a pre-built CSS template, which defines how a web page looks, so that we wouldn't have to spend time building our own look and feel (which can take a lot of time!). 

With the technologies picked out that would allow us to start quickly and at minimal cost, now I had to define our product. Once again, I was focused on getting to market as fast as possible. I made a list of all the features I would like to have, and then crossed out everything that wasn't absolutely necessary. We would build the minimum viable product, start selling it, and then build new features as we went and as our clients and leads requested them. 

In less than five weeks of writing our first line of code we launched Impact Dialing and started selling it to political campaigns. We released new features and bug fixes on an almost daily basis. A couple times, someone was interested in using our product but really needed a feature we didn't have, so we built the feature and released it within a day or two. And by the time the November elections were finished, we were already profitable!

We've continued to work hard improving Impact Dialing, building new features and making the system more robust. We've enjoyed keeping the company small - it's still just me and Brian - but we've started to think about bringing on investors to help us grow more quickly than we can on our own. I would love to go to more trade shows and events to promote Impact Dialing, but we've got a tight budget that forces us make hard choices (which can be a good thing, too!). And both Brian and I have other jobs that help us pay the bills, though we'd love to work on Impact Dialing full-time. 

Recently, I heard about a contest called the Lean Startup Challenge that offers funding to lean startups, and given how lean Impact Dialing is, I figured we'd be a pretty good fit. Wish us luck!

Monday, March 14, 2011

"Data Mining: How Companies Now Know Everything About You"

Turning this post over to Time magazine for a great article on data mining:
"Part of the problem people have with data mining is that it seems so creepy. Right after I e-mailed a friend in Texas that I might be coming to town, a suggestion for a restaurant in Houston popped up as a one-line all-text ad above my Gmail inbox. But it's not a barbecue-pit master stalking me, which would indeed be creepy; it's an algorithm designed to give me more useful, specific ads. And while that doesn't sound like all that good a deal in exchange for my private data, if it means that I get to learn when the next Paul Thomas Anderson movie is coming out, when Wilco is playing near my house and when Tom Colicchio is opening a restaurant close by, maybe that's not such a bad return."
 Check out the full piece at,8599,2058114,00.html.

Friday, March 4, 2011

Acquiring new donors

Acquiring new donors is difficult because acquisition campaigns usually cost more than they raise. The donors acquired, however, give enough in subsequent donations to make up for the price paid to acquire them, but acquisition presents a cash flow problem because of its upfront expense. Let's start with what a typical acquisition campaign looks like.

Most acquisition campaigns are through direct mail. A non-profit rents a list of names and addresses of potential donors from a direct marketing company or another non-profit, mails them, and a small percentage donate. For example, an organization might mail a list of 10,000 people. Each letter costs 60 cents to mail, for a total cost of $6,000. 50 people donate, or 0.5% of the list. The average gift is $40, raising a total $2,000. The net cost of the campaign is $4,000, and each donor costs $40 to acquire. To re-coup the cost of acquisition, the organization needs to raise one more gift of $40 from each donor.

There are three main elements I'm going to explore that influence the cost of acquisition: list source, channel (mail, phone, email, etc.), and amount asked for. Of course, there are many other important considerations, such as the framing of the ask (does getting a celebrity to sign your letter help?), timing (if your issue is in the news, people are probably more likely to donate), the wording of your letter or script, and so on. I'm going to focus on the first three elements, because there is more research on them and they are, in some cases, easier to test.

But I'm going to leave this post at here for now so that it doesn't get too long. Next time, I'll talk a little bit about some testing that's been done to lower the cost of acquisition. If you've ever done a fundraising experiment or test, I'd love to hear about it, either in the comments or directly.

Wednesday, February 16, 2011

Microtargeting done right

If you've worked on political campaigns, you've probably heard the phrase "microtargeting" and probably associated with some sort of sophisticated way to pinpoint voters to persuade. Recently, some people have been questioning the usefulness of microtargeting, saying that there's no proof that it's effective. Someone I work with mentioned to me today that her campaign is considering using microtargeting, so I thought I'd take a moment to write a blog post about the subject.

Traditional microtargeting works by calling a selection of voters and interviewing them about a variety of topics related to the election. Within these surveys, the voter is asked several times their opinion about the issue or candidate. The questions are often framed as a difficult value choice: most people would answer yes to "Do you support more funding for schools?", but if you ask "Would you be willing to put more money towards schools if it meant less money for policing?", respondents are forced to make a value choice that, according to microtargeters, reveals more about their underlying beliefs.

Microtargeters look at respondents who answer the questions differently throughout the survey process, believing that these people with "unstable opinions" are the most likely to be persuadable on an issue. A statistician uses these people to train a look-alike model that, based on demographic and consumer information, rates every voter on how similar they look the respondents that the statistician is modeling. (For more information on how look-alike models work, see this blog post.) The voters who are rated highest on this model are considered the best targets for persuasion.

How do we know if this works? Microtargeters offer a couple arguments. First, when they build their model, they leave out a segment of the respondents. Then when they build their model, they see how well the model predicts the responses of this group. But what if microtargeters are wrong, and people with unstable opinions actually aren't more persuadable than other groups? What if they just don't care? Or they flip back and forth all the time, whether or not someone tries to persuade them? Microtargeters point out that, in some campaigns that have used microtargeting models, the high-scoring voters on the microtargeting model have, throughout the course of the campaign, become increasingly more supportive, and the low-scoring voters have not. This is better evidence, but by no means conclusive. In these campaigns, the microtargeting may have identified people who were likely to become more supportive for reasons completely different from the campaign's activities (TV ads, phone calls, direct mail).

So is microtargeting useless? I don't know. The alternative to microtargeting, and the approach that I advocate, is randomized controlled experiments, as I outline in this post. But randomized controlled experiments can be more expensive and time-consuming than building a microtargeting model, so it would be great if it turns out that microtargeting really does identify persuadable voters. What I suggested to the person I talked to this morning was to do a randomized controlled experiment with the microtargeting model. They could build the model and then mail a random selection of high-scoring voters, as well as a random selection drawn from all voters. After surveying the mail recipients, they could compare the persuasion rates from the microtargeted voters and the random voters and see how well the microtargeting model performed. And they could build their own persuadability model from the random voters and see how similar it was to the microtargeting model.

Tuesday, February 8, 2011

Testing online donations

M+R Strategic Services has done some great work testing how to raise more money online. Check out a great blog post that gives some important insight into non-profit websites and email campaigns: It's great to see more folks running well thought out experiments like these!

Thursday, February 3, 2011

How predictive dialing works

I gave a training this morning to a group of political organizers across the country on how to use Impact Dialing, as part of which I explained how predictive dialers work. That made me think that maybe it's time that I explained it as part of this blog, too.

If you're ever had a long list of phone numbers to call, you know that at best, only maybe one out of two people will pick up the phone. In the middle of the day on weekdays, it can be much lower, often less than one in ten. And even when somebody does pick up the phone, it takes at least a few seconds for them to answer.

Predictive dialing uses statistics to reduce both of these sources of delay. By observing what proportion of people are answering the phone, a predictive dialer dials multiple numbers every time someone becomes available. So if one 1 out of 4 people is picking up, the dialer will dial somewhere around 4 lines when a caller is available. Additionally, it will start dialing even before a caller is done with a phone call. If a typical call lasts 100 seconds and it takes people about 10 seconds to answer the phone, the dialer will start dialing at around 90 seconds.

All of this math is a gross simplification, but it gives you an idea of the advantages of predictive dialing. It should also give you a hint about the shortcomings: if the dialer guesses wrong and somebody picks up the phone unexpectedly, there will be no caller available to talk with them. These are called "abandoned" or "nuisance" calls. That's why, when you're calling a very valuable list (such as donors), it's often better to use a preview dialer that only dials a single line after you've finished your last call.

This math also hints at when predictive dialers are most effective. The law of averages states that any chance error gets smaller as the sample size increases, so the more people using a dialer, the more efficient it is. Similarly, the less variation there is in the lengths of the phone calls, the more accurately the dialer can predict when to make the next dials.

Impact Dialing is doing a lot of research into pushing the limits of predictive dialer efficiency, as well as even making preview dialing faster. This research is under wraps until we get it hammered out and patented, but I can't wait to share what we're working on.

Tuesday, January 25, 2011

Some good software

Here are a few pieces of software I use for political campaigns and non-profits; I'd suggest checking  them out if you're in this field.

I just finished building a website ( in Squarespace, a content management system (CMS) that makes it really easy to create good-looking, maintainable websites. More and more non-profits and political campaigns are using Squarespace for their websites, because you can hire a web developer to build the site in Squarespace, and then fairly easily maintain it without technical expertise. It's my favorite tool for building websites.

Another important piece of technology for non-profits and political campaigns is their donor database. There are a huge number of databases out there, and I haven't seen any that I feel does everything quite right. However, I got a demo of ActionKit the other week and was very impressed. It's more for longer-term non-profits than political campaigns, and it's only available to progressive causes. A few people have asked me to help them pick a new donor database, and so I may have more to say in this front in the next few months.

Donor databases often include email tools, but if your database doesn't, the best standalone email software I've found is Mailchimp. It's dead-simple to use, makes gorgeous (and CAN-SPAM compliant) emails easy, and is even free for lists under 1,000 subscribers.

If you're running a voter contact program, you'll need a list of voters to contact. I often recommend folks to TargetSmart (only available to progressive campaigns); they make a good case that their data is the best because they use multiple sources and verify them across each other, so that you're getting the most and best-quality phone numbers, addresses, and email addresses. Another great source for California lists is Political Data, which works across the aisle.

For campaigns that need a database to store their voter lists in, the Voter Activation Network (VAN) is the standard tool for progressive campaigns (sorry non-progressive folks, I just don't know as much about the tools available to you). If you know of any other software like VAN, I'd love to have something to compare it to.

And, of course, Impact Dialing is the best software for contacting voters by phone - it triples the number of voters you can contact compared to dialing by hand!

Thursday, January 13, 2011

How to win elections, Part 2

In Part 1, I talked about how to get your supporters to vote with a Get Out The Vote campaign. Now, I'll focus on how to persuade people to become supporters in the first place. You can persuade people in a several ways, like mail, phone, door-to-door, and TV. But unless you've got enough money to talk to everybody in your district, how do you choose which voters to target? And how do you know which messages are most effective?

I've actually already begun to cover this topic in "Should campaigns hide LGBT people?". Here's a re-hash. Let's say you're planning on sending out direct mail, and you have two designs that you're trying to decide between. Randomly choose 10,000 voters to get one mail piece, 10,000 to get the other, and 10,000 to get nothing. (10,000 is a magic number for statistical significance.) After you send the mail, call all the groups and survey them on their support for your candidate or issue. Then you can see which mail piece is the most effective, and, by comparing the level of support of people who got the mail to the people who received nothing, see how effective your pieces are.

Using statistical modeling, you can also see which demographics were supportive in the group that got the mail piece but weren't supportive in the group that didn't get the mail, and then use that model to predict which voters in the entire population will be most persuadable. You might even find that one mail piece works best with one group, and the other piece works best with other groups. Now that you've picked the best mail piece and know which demographics to target, you can send the right mail piece to the right voters.

What if you're making phone calls in addition to sending mail? Add another 10,000 random voters onto your test and give them a call at the same time you're sending your mail. Then, when you do your modeling, you'll want to see if some demographics are more cheaply persuaded by phone and others by mail. A mail piece is much cheaper than a phone call, but many more people will listen to a phone call than will look at a mail piece. It turns out that different demographics have different response rates to phones versus mail; for example, older people are more likely to read their mail than younger people. [Shameless plug: if you're making phone calls, check out Impact Dialing's awesome predictive telephone dialer!]

Finally, you might be running a door-to-door canvass or TV ads. With these tactics, you have to target geographic groups of voters - either media markets for TV or "walkable clusters" for canvasses. The principles of testing these tactics is similar to mail and phones, but instead of randomly choosing individual voters for a test, you randomly choose walkable clusters or media markets. Then, after you model which voters are most likely to be persuaded, you'll look for the media markets and walkable clusters with the highest density of these voters.

Unfortunately, campaigns rarely do much testing and targeting. Instead, they rely on "expert" opinions, focus groups, long-form polls, and other approaches that often don't hold up when actually tested in the real world, and they deliver their message to either to broad an audience or to a poorly chosen targeted group. But if you want to do any of this testing, I'd love to help you out, so don't hesitate to get in touch: michaelrkn[at]gmail[dot]com.

Wednesday, January 5, 2011

How to win elections, Part 1

A client recently asked me about how best to identify new supporters of his organization's cause. They're looking at a ballot measure a couple years from now and want to begin the campaign work now. I admire this kind of forward thinking, but I told him he's better off procrastinating. Here's why.

Most campaigns run a Get Out The Vote (GOTV) operation in the last week or so before an election, and in order to get your supporters to the polls, you first have to know who they are. It used to be that the only way to know who was on your side was by calling them up and identifying them, one by one. Each of these IDed voters would then be contacted by mail, phone, or door-to-door to remind them to vote. GOTV campaigns often weren't run in bigger districts, because it took too much work to identify enough supporters to make a difference.

Statistical modeling has changed all this, at least for larger races. It only takes a couple thousand surveys to project to the rest of the population how likely each voter is to support an candidate or issue (see this post for more on how this modeling is done using consumer and demographic data). It turns out that these models can perform just as well as IDs in correctly identifying supporters. They cost a fraction of IDs, and the surveys and models are best done just a couple days before the GOTV campaign starts.

What about persuading voters to support your issue or candidate? Aren't ID calls good for that? Well, sort of. But you'll have to wait until Part 2 to learn about that.

If you're running an election and want to chat about your strategy, drop me a line at michaelrkn[at]gmail[dot]com.