Stealing Dropbox’s Referral System & 4 Other Growth Hacking Lessons

Every startup entrepreneur will go through the experience of figuring out growth hacking methods. Having been through it multiple times ourselves, we thought it might be helpful to show others how we put the concept into practice.

For those of you who aren’t familiar with what we do, RocketClub is a platform for early-adopters and evangelists to earn shares in cool startups (yep, thats right!). For the rest of this post, we will walk you through how we growth-hacked our member share page and demonstrate principles we adhere to throughout the process. Hopefully you will find it useful!

1) Keep improving your feature set

All RocketClub members visit our campaign share page following sign up to help companies on our platform. In fact, an astounding 58% of our members share via social network on this page. Here it is:

Now here’s the important thing to note; we did not get 58% members sharing on this page from the get go. The number of members sharing was in the low 40s (%) when we first started. This conversion grew through multiple iterations of our share page. This is what we were working with initially:

One of the key things we focused our attention on soon after launch was the social networks themselves. Our intuition was that social networks are not all created equal, and hence result in varying degrees of virality. Some of these differences between the social networks include Facebook disallowing pre-defined messages for users to share, Twitter tweets being much more fleeting, and the nature of LinkedIn being primarily for work.

Using the analytics platform Mixpanel, we found that between July 6th and 12th, the number of Facebook and Twitter shares on this page were similar (484 vs. 448). Further investigation into our analytics, however, showed that 2,088 users visited from Facebook compared to the 644 visitors from Twitter (LinkedIn was not even on the radar).

We decided to make small tweaks to prod our members towards sharing on our desired social network — Facebook. First, with a ‘Click here’ image, and then later with suggestive text. We started seeing Facebook shares increase significantly.

In August, our Facebook shares totaled 1,050 compared to 282 shares on Twitter.

Facebook messages now dominated our referrals and incoming traffic, but we didn’t stop there. Remember there is always room for improvement; the question is finding out where conversions are hit the hardest.

Since Facebook does not allow pre-defined messages, we weren’t able to see what messages accompanied our member’s Facebook shares. The small sample of messages from our own network, however, led us to believe many Facebook shares didn’t have a message at all. That’s when we decided to prod our members some more.

We weren’t able to identify how many Facebook shares tagged friends and resulted in more conversions, but referral visits increased by 6.2% per each Facebook share. We were able to even notice a difference on simple tests with our own Facebook network. After all, Facebook users receive a notification when they are tagged and this typically gets an user’s attention.

You probably get the idea now. Always be improving (ABI). Some improvements are incremental, but sometimes they can be drastic as well, as we’ll see below. When optimizing for virality in particular, the key is mapping out your viral loop to understand all the conversions in between. Andrew Chen has a great post about this.

2) Steal from the best in business

Comparing your product’s user experience, user flow, and features with other successful products is also a great way to hone in on conversion rates. It’s an opportunity to learn and steal from the best.

RocketClub members shared a lot but we knew we weren’t converting new members at a high rate. Referral visits were low per Facebook share and we weren’t closing the viral loop. Simply put, not enough users were clicking on the Facebook shares by RocketClub users.

We believe the most challenging part to going viral, the intent for a user to share, had already been solved. Perhaps the Facebook ad didn’t look enticing enough? We started running experiments to A/B test variations of our open graph data to optimize for referral visits until we had another idea.

“Good artists copy, great artists steal.”
– Pablo Picasso

For those who work in the industry, we all know Dropbox grew virally through their ‘Refer a friend and both receive 500MB of disk space’ marketing campaign. We asked ourselves –how did they completed the viral loop? What did Dropbox do differently? We took another look after signing up for their service years ago.

Nothing about sharing on social networks! It wasn’t an accident; Dropbox probably has an entire team focused on optimizing for virality. The primary way to share on Dropbox is email. The alternate secondary method is via a custom link. This got us asking other questions: Is it because emails are more targeted? Is it because emails allow them to run drip email invitations? Are users more receptive of new services via email?

At this point, there was only one way to move forward. Copy the professionals at Dropbox and test to see if there is any validity to their referral system. Dropbox has been so successful growing through their referral system, we thought the smart thing to do was to copy exactly the way they did things. After all, their resources and experience in conceiving this page were far superior to ours.

At first glance, Dropbox’s referral page seemed simple. But it was anything but. We spent a weekend hacking together a version for RocketClub that mimics their page feature by feature. Below is the final result:

Those who are inclined to find out how we teared down Dropbox’s share page feature by feature, and implemented it for RocketClub, check out this adjacent article we published:

Breaking down and Implementing Dropbox’s Referral System (warning: developers only)

3) Only change features upon evidence

While we believe Dropbox’s way of doing referrals must be right and should result in increased virality for us, we are still running A/B tests to ensure new changes are supported by data. This means half of our users visiting our share page still see our old version, while the other half see the new Dropbox-inspired version. It’s really important to obtain data on which one converts better (and why). We have a rule here at RocketClub –changes are not tolerated unless supported by evidence. It keeps us from “spinning our wheels” and making product changes for the sake of changing them.

With the help of Optimizely, we’re running an A/B test on the share page for every RocketClub campaign. We also tag all our campaign referrals appropriately to understand who and thru which channels they come from. Since we are making a major change to our share page, we perform one simple test where either version is displayed 50% of the time.

Want to check out the A/B test we’re running? Sign up and get shares in RocketClub here:

You can then visit these two pages and compare the two versions yourself:


4) All experiments must have a hypothesis

A/B tests are useful and results-oriented, but they are even more helpful when backed by a hypothesis. Identifying one behind an experiment puts it into perspective.

“In research, the purpose of a hypothesis is to make a statement of the expected results of the study being conducted and makes predictions on how to or more constructs will be related.”

Here are hypotheses from our earlier experiments:

  • The difference between social networks result in varied and disproportionate referral results
  • Social proof increases user sharing
  • Multiple sharing options increases user sharing

Once proven, hypotheses allow for smarter future iterations of your product. In the case of our recent A/B test between the social sharing and email-only share page, claiming a particular page converts better than another only identifies a specific causal relationship. By including a hypothesis such as ‘email sharing brings in more referrals than social network sharing because emails are by default directed at recipients,’ our experiment becomes much more meaningful. Validating these hypotheses are how startups make progress.

“The key to making progress as a startup is validated learning about customers.”
– Eric Ries

Last Words

Growth hacking is empowering. If anything, we’ve learned that the key is to stay disciplined, open minded, curious, and creative. As developer marketers, we sometimes find ourselves chasing after the next shiny cool feature and forget the big picture. To recap, here are the guidelines that work for us:

  1. Keep improving your feature set
  2. Steal from the best in business
  3. Only change features upon evidence
  4. All experiments must have a hypothesis

And one more thing, if you found this article useful, help entrepreneurs you know by sharing it with them!

Originally published at on October 21, 2015.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.