This is a guestpost from AdRoll.

The first thing most heads of marketing do when they join a new company is to change strategy; it's no coincidence then that the median tenure of a top marketing position is less than two years. It’s a story as old as time — a marketing manager joins a new company and decides to rebrand and update the go-to-market messaging. The customers don’t respond well and, shortly thereafter, the marketer loses his or her job.

Testing any new messaging before you put it out in the world can go a long way toward making sure things go well. But testing comes with its own set of issues. Asking around the office isn’t very scientific and focus group results can be questionable. Running a controlled digital experiment is a good plan, but, unless you run an incredibly popular brand that gets hundreds of thousands of visitors a day and have a short buying cycle, it can take a long time to get results. What is the scientific marketer who wants to keep their job to do?

The answer could be referral marketing. Using referrals to test your go-to-market messaging has strong advantages over more traditional testing approaches. But before we get to those, let’s clear up what we mean by using referral programs for go-to-market messaging testing. Referral marketing is a formalized version of word-of-mouth marketing. You can use it to rapidly increase visits and conversions, as well as use the actual design, language, and offer of the referral to test the messaging itself.

Hitting the Numbers

One of the fundamental problems in any digital experiment is getting a large and diverse sample size. Getting a large enough sample size often means either waiting or paying for traffic. Getting a diverse enough sample requires careful planning of how you manage segmented ad buys (you do segment your ad buys, right?) A referral program is an attractive alternative to traditional paid digital. For one, referral programs are often cheaper than an ad buy, and come with no up-front costs and easy budgeting - you only pay when someone takes advantage of a referral, and you know what your CPA is well in advance.

Even more importantly, referrals, in general, tend to convert at much higher rates and much faster than most other methods of generating a lot of traffic quickly. They also tend to convert with significantly fewer touchpoints. This is important — if your product is new, you don’t have time to engage and nurture an audience. And if your product has been around for a bit, you are unlikely to be running your updated messaging across all channels, and you want to minimize “cross-contamination” in your tests. Faster time to convert with fewer touchpoints means you get real action-based data rather than having to wait or rely on indirect or vanity metrics.

Getting It Right

Referrals can be a good way to test your go-to-market strategy, but there are some special considerations to keep in mind.

1. One fairly unique aspect of referral programs is Self-Selection Bias. People tend to associate with others who are similar to themselves, so any referral links they send out are likely to go to a similar audience profile. Getting around this requires you to spread your test messaging across diverse groups.Alternately, you can use the same referral messaging, but split test the landing page for visitors from each unique referral code. This can be a good way to add additional controls and add to the validity of the results.
2. It’s important to remember what you’re testing. The first thing people like to do when setting up experiments with referral programs is to play around with the referral incentives. That’s fantastic if your goal is to identify what the best offer is for your referral program. It’s not so good when your goal is to isolate and test the messaging.Adding too many variables into your test can very quickly turn good data into a lot of statistical noise. This is true for testing offers in tandem with testing messaging, but also for testing too many variations on your messaging at the same time - multivariate tests are doable, but take a lot of discipline and some tricky math to get right. Instead, focus on a rapid iteration model. Test two options, pick the winner, add a variant and test it against the winner, etc. Because of the speed with which referral programs can be rolled out and generate results, you won’t be missing much speed-wise by going iterative rather than multivariate.
3. On the note of keeping your testing simple, you should test one thing at a time. Unless you have the organization and statistics to test multiple variables accurately, you should refrain from testing the referral language and the landing page at the same time.Testing the language of the referral creative you supply to your customers is great for testing short, top of funnel messaging — what will grab people’s attention and make them sit up in their seats a little. Landing pages, on the other hand, are better for more nuanced experiments — which benefits, and in which order, hook the most customers, for example.

Product launches and rebrandings should be fun, but all too often we marketers let them turn into nail-biting exercises in waiting and seeing and hoping. That may have worked for Don Draper in the 60s, but these days we have the technology to build better, stronger, faster launches based on better data. Referral programs should be an integral part of your testing toolbox for their ability to generate data quickly and at a lower cost than buying a massive ad blitz or paying off a gaggle of influencers. It’s a secret weapon that can help any marketer break the cycle of hire, launch/rebrand, fire; and that’s something we should all work toward.