Here's How I Used Old-School Direct Mail to Power a High-Growth Online Business
Andrei Utkin is the Chief Marketing Officer at Insureon, the Chicago-based online insurance agency for small businesses.
Read on for a transcript of his presentation from our January 16, 2018 Here’s How Startup Marketing Conference, where he describes how he used direct mail to power an online business.
Hi everyone, for those of you who don't know Insureon, I'll give a quick introduction to the company. We are a small business insurance agency. We help small businesses find the right type of insurance and make the process of buying it a lot easier. We were started about seven years ago as an idea and have grown into a 250-plus strong firm. I hate this analogy but you can think of us as the esurance for small business insurance. You come to us as a small business owner, fill out one application and we give you multiple quotes from multiple carriers, which you can pick from and become a policy-holder. But unlike any aggregator or another intermediary, we are an agency that will stay with you for life. We will assign an account manager to you and if you have any questions or have a claim to file, we will help you with that.
We are very close to our customers. When we started as a small company, we didn't have a brand or much money. So how do you attract small-business owners as a startup? Well, you go for hand-raisers. You look for information when the customers volunteer themselves about their intent. Small-business owners typically don't buy insurance on a whim. They buy insurance either when the company is just established, when their current policy is up for renewal, or when they’ve gone through a major growth event such as hiring a bunch of people and needing more insurance to be adequately covered. In the first four or five years of our history, we have relied very heavily on low-funnel or hand-raiser channels to grow and to attract customers. Of course, those are SEO, SEM and to some extent, retargeting. I'm sure you all understand what low-funnel channels are and why they're so easy and profitable. It’s because you have the luxury of knowing exactly who you're going after, in fact, you're only there when the customer says, "Hey, I'm interested in your product."
So we started building up our presence through Google and Bing, bidding on several hundred thousand keyword combinations and have since become a formidable player on Google. You know you’re an important company to Google when they assign a representative to you who is willing to help you continue to grow your business, which is what happened after three or four years. Brenna, our former Head of Content, did an amazing job with her team setting up an SEO machine and writing content by the hundreds and thousands of pages. I think we have over 6,000 pages of content today. So if you Google any sort of question about small business insurance, chances are you'll see Insureon or TechInsurance, our other brand, on the first or second page of results. That worked really well for the first three or four years, but then we started hitting diminishing returns, especially in the SEM space where we were already at number one, two, or three on most small-business-insurance-related questions. Don’t get me wrong, the channel works great and it’s still our cash cow, but if you want to continue to grow you need to look elsewhere.
So we started doing that and said, "Okay, now that we’re a little bit bigger and have a little bit more money, let's experiment with brand building." We looked at higher funnel channels. From there, we hired an agency, shot a commercial, and ran it on national TV. It cost over a million dollars to do all of this. It definitely generated measurable impact in terms of site visitors and applications, but overall the ROI just wasn't there. We couldn't justify running the long term, especially with our limited cash flow. We just couldn't see the appropriate amount of response measurable to the cost of the actual production. The same is true for online displays. We put millions of impressions out to different ad exchanges. We even had an agency to help us with buying the right type of audience and targeting, but again, the results weren’t enough to justify it. We saw there that you really need to understand measurement and proper attribution if you want high-funnel channels to work. If you’re only relying on cookie-based methodologies where you say, "Okay, I know this person will serve the display ad for me so I'm going to see if they show up as either a click-through or a view-through." You'll be severely short-handed in your understanding of the effectiveness of the channel because a lot of people use anonymous browsing and multiple devices. So if I see my banner on my desktop but I apply on my mobile, traditional measurement mechanisms are broken. Of course, there are all the media mix modeling kind of companies, Visual IQ, MarketShare, etc., who help you solve for this, but you honestly need a much bigger scale to properly build an MMM model, plus a lot of money, which we didn't have. So we said, "Okay, nice experiment. We’re going to learn a lot of valuable lessons from this, but if we want to continue growing and show profitable growth, we need to look elsewhere."
This is when we had the idea of using direct mail. In the digital space, a lot of people are skeptical about direct mail. It still has a space, especially in the subprime space of profitability, and you can use this channel successfully. So we decided to see if direct mail could help us solve the measurement question. We were also hoping that it could solve the timing problem because we didn't know how long it would take for someone to become a customer after seeing our ad. Again, because people don't buy insurance immediately after seeing an ad – it's not an impulse buy like electronics or a vacation – we wanted to make sure we understood what the response curve was.
We hypothesized that with direct mail you know exactly who you're targeting and who's coming in as a customer, so you can put one and one together and it provides a much more robust measurement mechanism that you can use to inform your future strategies. So we did that and started investing in direct mail. When we started playing with the channel, in true analytical fashion, we said, "Okay, let's look at all the levers that can help you improve your performance in direct mail." The levers are, at least in the case of this channel: pricing, creative targeting, data, and measurement.
We hired a national agency.
In terms of pricing, we initially started with a local vendor and then we learned that there is a huge difference between printing and mailing costs, depending on who you work with. We started talking to other people and ended up hiring a national agency that helped us save over 10 cents on production cost and seven to eight cents on mailing costs, which doesn't sound like much but if you look at it, given that the cost of sending a piece of direct mail is about 50 to 55 cents, it's actually a huge percentage of your overall cost base. That saved us a lot of money and because this is one of the largest vendors in the nation, basically what they do is commingle your mail with all the other players they have, so you pay basically wholesale prices even if you're a relatively small player like we are.
We bought data from specialized vendors.
We looked at the data side of things and started working with one of the biggest vendors for small businesses, Data Innovation. They have about 10 million records, which they continuously update and maintain fresh. We began buying data from them and realized that data is nice if you want to cover the entire industry, but if you have a focus on specific verticals, different types of small businesses are worth very different amounts of money. Our most lucrative vertical is the tech industry, so IT consultants, web designers, system developers, etc. You just don't get enough high-quality data from the national vendors, so we started looking at more local and specialized players who only focus on those verticals, and we’ve had a lot of success and we’re seeing less return mail and higher response rates.
We added urgency to our creative.
When we first started playing with direct mail, we just created a nicely branded piece of mail that talked about the benefits of working with Insureon. It used a very peaceful tone and was good looking, but it didn't really work. The same agency we hired said, "You guys are doing it all wrong. You have to create urgency. You have to make sure that, within five seconds, the person understands why they should be paying attention to your product, otherwise your mailings will go into the wastebasket." So we redesigned our creatives and made it a bit more “salesy” if you will, but it worked. We created a type of messaging around, "Your business might be at risk.” It used a little bit of fear mongering, which I don't necessarily like, but it works.
We included punch-out cards in our mail pieces.
We put punch-outs in the mail so people who were not ready to buy from us immediately could take those cards and put them on their fridge or desk and remember our name when it was time for them to actually think about purchasing small business insurance. This was very important to solve the timing issue, which I’ll address in a little bit.
We built a PII-based attribution logic.
Then we looked at the attribution logic and said, "Okay, so given that direct mail is so adjustable, we know that John Smith at 123 Main Street received mail from us. Let's make sure we have a proper measurement system to see if John Smith shows up as a customer.” This required building a custom address-matching logic because if you mail someone at “123 Main Street” but they put in their address as “123 Main St.,” you want to catch that. The other thing we did was make data more controllable. So when we pull data, we say, “Here are a million names we want to mail to, let's make sure we put aside 100,000 of those so we understand the rate they show up as our customers anyway, because it will happen even as we maintain a pretty strong presence at other marketing channels. Then we will only take credit for the incrementality of the response rate among the 900,000 that we do mail.”
That was very important because we did see that a lot of the control group people actually showed up, which is great because it means we have a fairly strong presence in the marketplace. But looking at the response curves, between the control group and the mail group, it allowed us to take accurate credit for the effectiveness of the channel.
We built a response model.
When you work with a big data vendor like we do, the benefit of pulling data from them is not just that it's accurate and up-to-date, it’s also the fact that you get a lot of variables along with just the name of the company. You can get things like the size of the company, when were they formed, the title of their main person we're mailing, etc. We put all of these variables, literally more than 100, into a big regression model and let the data speak for itself. Once we'd accumulated enough response to tell us what the biggest predictors of response were, we built a response model that separated our entire universe of 10 million potential prospects into 10 deciles. The difference in response rate between a top decile and the bottom decile was over 10x. We were able to split the response well and ended up focusing on the top three deciles, where the response was way above average. By focusing on those three alone, you suddenly have an ROI-positive channel, where the randomized targeting approach did not break even.
Putting all these things together allowed us to get the most out of the channel and build a case for a rollout, which is what we did. When we put the best creative team, best data vendor and appropriate response model together, our ROI went from minus 60 percent during our first test, to plus 20 percent. Our first official rollout of this channel just went out a couple of days ago. We are planning to spend a seven-figure number on direct mail this year, it will account for over 10 percent of our acquisitions. So we are not yet an SEM or SEO killer, but we are a very formidable, additional component for a marketplace.
And maybe, most interestingly, we learned about the response curve. It was actually shocking, even to our vendor, because we learned that even three, four, five months after you send a piece of direct mail, you can count an incremental response. It proved our hypothesis that what we are selling is a product that requires a very low-sale cycle. And if you only have a 30-day window to measure your response, or you have a mechanism that only counts immediate reactions to your ads, you will be severely under counting your measurement and will never be able to grow out of those other channels. So this will now influence our long-term strategy. We are going to be able to be smarter about our measurement and our methodology for calculating returns.