C&E: Let’s start with where things are now. What are some of the capabilities you all are developing now?
Allen: Maybe it’s just because of where we are in the election cycle, but one of our biggest interests right now is a constant endeavor to push down the fundamental insides of microtargeting and modeling to lower-level races. It’s always fascinating to look at where everybody’s coming from. The traditional pollster–media consultant relationship, in our opinion, needs to be just as strong, with pollster–mail, pollster phones, pollster direct voter contact, and that works well in a big race where you can afford your own microtargeting.
Rivlin: I get asked a lot, “How do you work with pollsters?” One of the simplest ways I like to answer is, we look through the other end of the telescope. The pollsters are more involved in messaging and working out exactly what the language is and what the landscape is. We’re more nuts and bolts, the people you need to speak to. We find that the data is not refined enough to get to the nuance of messaging, but the data can get to divisions that really speak to peoples’ big ideological breaks.
One of the big repercussions of the 2008 election is that more people want microtargeting. There are a whole lot of people who worked on the Obama campaign—their first-ever campaign—who are used to seeing a model score on the voter file. They’re surprised to open a voter file and not see a model score.
C&E: Maybe you could expand on that a little for those who don’t know what a model score is.
Rivlin: Model scores come in a lot of different guises. There are models that do the clustering and divide the population into different categories. Another way of doing a model score is to give a predicted probability or a ranking of someone having a particular attribute. How likely are they to vote, how likely are they to vote democratic, how likely are they to have a particular issue? You know the way that these things have been increasingly used, is that people do a select on a list of people from age 35 to 47; you can say, I want a list of people that are 35-47 who are likely to vote. So that would be how a lot of people on the ground interact with those scores, and it’s interesting. In 2008, I spent a lot of time trying to persuade people that this dark magic was actually not so scary and was accessible and they should trust it. Microtargeters aren’t the wizard of Oz. [laughter]
Drechsler: As Joel alluded to, a lot of it is dispelling myths of what microtargeting is. Some out there see it as magic, and some see it as bullshit.
Allen: The truth lies somewhere in between.
Drechsler: The Obama campaign’s success with microtargeting has made everybody in some way want it. Some people have very high expectations [that they’ll have the same result Obama did], which isn’t necessarily true. The campaigns that embrace microtargeting and run with it will have success. Those campaigns where they say, “We don’t need it,” without having an understanding of what it is, can be frustrating.
C&E: When you approach a client, what are some of the things you impress upon them? What is the education process?
Drechsler: The approach is different depending on budgets. The reality is, we can do programs as small as $10,000 or up to a half million. It really depends on what your objectives are. I think one of the best selling points is that it pays for itself.
Meyers: We try to walk through with clients on this idea of, “Is my campaign big enough for market targeting, or is it too small?” And then we try to say that it doesn’t really have anything to do with how big a district you are in; it has to do with how big your direct contact budget is. If you have a $30 million budget, and you’re going to spend $29.9 million of that on TV and the rest on market targeting, that doesn’t make any sense. Our general rule is that if the project is going to cost you 10% or less of your direct contact budget, then this isn’t an easy decision for you. If you’re in the 30 to 35% range, you’re probably at the point that you should just be looking to put more mail in the mailboxes.
C&E: Does it make your job difficult, figuring out what the other side is doing?
Meyers: We all have some idea what each other does, but we’re probably missing half of it or misinterpreting half of it. There are certainly a few people that claim to do microtargeting, but you can just tell by the way they talk about it that it isn’t the same thing.
Rivlin: Part of the arms race isn’t so much working out what weapons each person has as assuming the other side has the arms and making sure you’re muscled up so that when you go to battle, you’re not at a disadvantage.
Allen: Nothing that anybody at this table does is reinventing the wheel in terms of analytics or modeling. The tools already exist. Building your analytical tool chest is fairly easy, especially since the user community always lags behind the state of the art in terms of what we can do with the data or the behavior of the data. We have to push the envelope incrementally, because if we try to throw three new tools at [clients], they will probably use none of them. When you do a big presidential level race or a top-tier senate race, you’ve got people who are presumably focusing a lot of time and energy on adoption.
Rivlin: I find that this conversation has to happen on two levels—that you have to have buy-in from the people who are signing the checks and ordering the projects, but increasingly you also have to have buy-in from the people who are implementing it on the ground. In 2008, we ran projects where people would spend a lot of money and we would be in contact with the campaign folks, and some of the people on the ground didn’t know the scores existed or didn’t know how to use them and didn’t trust them.
Field organizers are my favorite people in campaigns, but they have to know how to pull the list. They have to know what things are, what they can trust. It’s very tempting to see a score on a screen like it’s on tablets of stone. You know the reason those people have the jobs they have is that they get the local politics, they get where they should use the models blindly, where they should use the models straightforwardly. You should empower people so that they don’t either distrust it or trust it so much that they turn off their political brains.
Drechsler: The worst thing that could happen is to have a model that sits on a shelf. You spend the time and the effort and the campaign spends the money, and nothing ever happens.
Rivlin: Or, they layer on things they shouldn’t be layering on. They say, “This is a model score, so I will use this within the universe that they would have already selected.” Sometimes they have good reasons to do that, but sometimes it defeats the whole purpose of the model.
Allen: That’s always the friction: How do you get what you can provide to that sweet spot of what can be used and what will be used? We could talk till we’re blue in the face about propensity groups, clusters, classes, all sorts of fun stuff. There are always three skill sets worth looking for in our firm: people who are smart and can do great analytics; people who can sell the stuff, because we’re a for-profit industry; and then people who can help communicate it back to the users.
Drechsler: [You’re] educating somebody who might be a regional field director on this cycle, but two cycles from now might be a campaign manager on a statewide level. That’s the investment.
C&E: We’ve talked a bit about going down ballot, where budgets get smaller and smaller. What’s the furthest down ballot you have seen microtargeting used effectively?
Allen: Well, if I start talking about the state senate or state assembly races in California, those are spending more money than half the congressional races in the country.
Rivlin: It’s not so much what level the race is, as much as what level it’s purchased at. So if you are working for a statewide entity, whether it’s a caucus or an interest group, the economies of scale are worthwhile for buying some modeling that isn’t going to be as precise for each individual race but will give them a lot of purchase across their races. If the cost is minimal or zero, then, unless you’re not doing any direct voter contact, you can still use it. The question is whether it’s worth investing in from the beginning if it’s a lot of money.
Allen: That’s the real challenge: The places where, intellectually, these kind of targeting and analytics tools would be most powerful and useful are the campaigns that are most reliant on door-knocking, phone-banking, and direct voter contact. For the New York City mayoral race, I’m sure they bought the whole world. So much of the movement of that race—and I didn’t poll it, so I have no idea what moved what—but presumably it was television. Everyone’s got a little bit of money to throw around to buy some TV. It’s when your only tool in your arsenal is direct voter contact, mail, phones, and doors that the quality and sophistication of the targeting is going to be the most impactful in terms of real wins/losses, because it’s the only way you’re communicating. The flip side of that is where budgets become smaller, because if you had a lot of money to buy TV you wouldn’t be relying so much on knocking on doors and phone banking. That’s the challenge we all face.
Rivlin: The other place where it’s most beneficial is where you don’t have obvious cues to target. So, the places that can’t afford it are often the places that need it the most. The classic example of that would be in a primary election. You can target on who is likely to vote pretty easily by vote history. Knowing who is likely to support your candidate rather than the other candidate is crucial. You don’t have partisan registration, partisan primary history, and this is where it is most impactful. There are not that many places where primary candidates have enough money to really invest in it, and then the question is whether they have enough direct voter contact to make it worthwhile.
That’s where you can do more interesting stuff, in states where it is easier to identify the supporters and lack of supporters. That’s where the added value is if you start modeling issues or subsets within. We tell people not to do projects if they can get 90% of the way there using what they’ve already got in the file. It’s a waste of their time.
Meyers: Ballot issues, too, are a huge area where this can be of benefit. Some of that isn’t even so much that they don’t have the money. They’re just this group of people that is used to ballot stuff. They are just completely not used to doing mail and phones because there was never really any value to it.
C&E: Then you run into the California problem again, where ballot issues are king.
Allen: Especially ballot issues that have non-standard constituents.
Meyer: Or where it is just sort of muddled. We did some joint stuff in Ohio, where one party might lean one way or the other, but it wasn’t so strong that you just wanted to get all the Democrats and leave all the Republicans and vice versa, that you were going to have to get some votes out of these guys. You know, we did this Obama initiative stuff in Ohio on primary day, and it was a little scary in this Tea Party age. Are we going to be able to thread this needle?
Drechsler: And even in party-rigged states, we sometimes find ourselves asking, “Why do we need this?” You might have the advantage of 400,000 votes, but you’re down in the polls. You’re going to need that crossover; you’re going to need to figure out who you can reach out to, who your soft Dems are. A Democrat in one part of the state is very different in another part of the state, just like a Republican is different.
C&E: Something you would never have known had you not gone through this process.
Meyers: That can be a struggle too, when you find something that is counterintuitive. How do I validate this enough? I come out of the purely political world, even trying to convince myself that I’ve really found something. And then, do I have confidence enough to take this up the ladder?
Rivlin: That’s where we make our money; it’s why we have a mix of political backgrounds, rather than just stats classes from MIT. Our job is to interpret the politics and see when something comes up that’s counterintuitive: Is it counterintuitive because there’s a lot of counterintuitive things in the world that become perceived wisdom? Or because there’s something wrong with the data and the stats? Our job is to make that decision, and that’s where the art comes in. That is what’s fun about the gig.
Allen: If it were just plugging things into a computer and spitting out lists…ugh.
C&E: Let’s talk about looking ahead. What gets you excited about what you are going to be able to do? Some of the advancements you think are coming?
Meyers: I think in 2012 we’ll see the full integration of the digital side of campaigns. We’re seeing that more now, but it’s not fully there yet. By 2012, you’ll have the full integration of it—the Holy Grail we’ve been seeking since we started. Every time someone has asked me this since 2004, “Where I can deliver my commercial to somebody,” is beginning to exist. We could do it if they would let us, there’s no technological impairment.
They don’t know how to put a billing system together for an individual ad that goes out. They know how to bill the cable system, and they don’t know how to bill a household. The hard part of that is, how do you make people who aren’t partisan watch your stuff? Getting someone in the middle to watch something is so hard. Outside of the presidential candidate, why would anybody opt to do that? It’s like, well, you have a website that people can go to get information, but what would you need this for?
Drechsler: You can take the approach of shooting fish in a barrel: You’ll hit somebody. Or you can do the scalpel or laser. I think that’s the greatest benefit of microtargeting for campaigns—it hones in on how to deliver the best, most precise message to your voters. it complicated and hard is that you can’t simultaneously run seventeen different messages in a campaign. When we talk about this stuff, you’ve got to take the micro out of microtargeting. No campaign, not even presidential, can run 17 niche message streams to 17 different groups of people. You can’t afford to do it, it’s impractical to do mail runs or phone strips, and the logistics are ridiculously important. At the end of the day, each of us has to get 50 1. You’ve got to be able to deliver narrow messages, but also not get so focused that you’re looking at your own navel. You have to find really nice clean universes that have 5% coverage of the population so that you end up winning big in the places you target and ignoring the rest of the population.
Meyers: We get frustrated, especially on big campaigns, with some of the laziness. There isn’t a real reason you can’t put together a dozen telephone scripts. There are some practical reasons why you can’t put together a dozen mail pieces every time you send mail out, but mail scripts aren’t hard.
Allen: No campaign is going to say, “Okay, I’ll be the experiment, we’ll mess around with something that might or might not work and see you on Election Day!” You get this nice field experiment thing because not everyone is going to implement what you are doing in the same way. Not everyone is going to adopt it to the same level. Not everyone is going to make the same choice based on the advice they’re given. Across multiple campaigns, you’re going to get this nice field experiment mode. Then you say, “All right, let’s see if we can learn anything, and hopefully we’ll learn something important and be able to adapt what we do.”
Meyers: And convince anybody that you’ve actually learned anything.
Rivlin: I think that the testing question, this is one of the advantages of being in a firm that does other stuff. We have a research and development budget that is mixed between the mail side and the analytics side, and we do a fair bit of it. Even earlier this cycle, we did tests and some of it was subsidized through legalities that allowed us to learn what works and what doesn’t work. I think modeling has gotten to a place where people are comfortable. The thing that’s dragging us back a little bit is the quality of the data. We can improve the modeling. I think the next step is working out the effectiveness of programs.
Allen: Persuasion and GOTV is tough to test.
Rivlin: Yes.
Meyers: I don’t know how you guys feel, but I’m frustrated trying to extend these tests that get done to all campaigns for all times. Somebody had touched on this earlier. “I’m going to run this campaign like a business.” There’s no business in the world that would spend $20 million to go from 47% market share in California guaranteed to 51% market share! There’s sort of this silliness to it.
C&E: Is there a concern that there isn’t enough policing of the industry? That’s why there is always such a thing with polling—when something goes wrong, everyone goes nuts because everything’s accepted as legitimate.
Rivlin: I think your reputation matters; it matters in everything in politics. It’s the same reason we down-sell a huge amount of time. The clients will come to us and ask us for x, y, and z. We’ll sell them x, and tell them to spend their y and z for voter contact, because they need to come back for more and for your reputation. They’ll see pretty quickly that certain things are harder to test, but if they are running any kind of ID program, they’re running things again. There is policing almost after the fact, and, as with everything in this business, reputation and the people giving you recommendations are key.
Drechsler: Training that regional field director, who might be way down on the totem pole today but could rise above down the line—that’s where you’re building your reputation. Its standard practice to be nice to people, to take their calls and walk them through it. There are times where some people get it better than others, and it’s frustrating and you want to beat your head against the wall.
On the flip side, occasionally you have people who come in and question everything, and it’s just one of the things you have to deal with. In some ways we police ourselves. There are some who claim they do microtargeting, and they don’t do microtargeting! I think that’s the most frustrating thing. You sit there and you pull your hair out, trying to go around and dispel that, but is it always the best use of your time? You think, “They won’t be around next cycle; we will be, because we’ve been doing this and we have a track record.”
Allen: There is enough money chasing things in politics that even people who aren’t very good at what they do manage to make a decent living sometimes. But there are hierarchies, and if you are successful and win things, people sort it out pretty quickly.
Meyers: Oddly, the lousy competitor is the one who scares us most—going out and selling real cut-rate stuff.
Drechsler: It’s tricky because you don’t want to outright trash them, because then you are trashing the people who hired them for making a bad decision. It’s a delicate walk through that minefield.
Meyers: Right, when we lost Bloomberg to you it was a little difficult; it was the only time we ever lost to another democratic firm. Had to cancel the summer home!
Andrew Drechsler works for Strategic Telemetry, one of the premier Democratic microtargeting firms.
Joel Rivlin is director of analytics at MSHC Partners.
Michael Meyers is a partner and president of Target Point Consulting, one of the first Republican microtargeting shops.
Bryon Allen is the chief operating officer at Wilson Research Strategy, a Republican polling and targeting firm.