Date
Mar 2023
Client
Internal Project
Earth Breeze
Role
Senior Product Designer
UX
Earth Referral Program
Earth Breeze is an E-commerce giant that has acquired more than 2 million subscribers purchasing eco-friendly laundry detergent without plastics, while donating a portion of each sale to charity.
Through Meta ads, they have built an incredibly large and devoted fanbase. I was tasked with creating a referral program for them.
The company had found its previous success by taking third-party Shopify apps and piecing them together. I was hired to design and manage the development of their first custom software tool from the ground up.
Rather than designing an entirely new concept, as I had in many previous roles, this was an optimization project. Could we make a tool that performed better than existing tools on the market? Would this tool prove to be a financially viable stream of income for the company?
Listed below is the design approach I took towards answering these questions.
There were two objectives that determined the success of the program.
Given these objectives we settled on a single north star metric for the entire project:
New Subscribers for every 1000 people contacted.
We liked this metric for the following reasons:
A broad understanding of project success does not matter if you can’t drill down into the smaller parts of the project to understand what specifically needs to be improved.
We decided on comprising metrics visualizing each stage of the journey for both the referral sender, and the referral recipient.
Before this project, our company tried to implement a referral program using a third-party shopify app called Talkable. As the final step, I used the best performing campaign of talkable, to set benchmarks for each of our new referral program’s metrics.
If we could not beat what the company had previously tried to implement with a pre-built app, the project would be a failure. Let the games begin!
What is the optimal way to incentivize our customers to share Earth Breeze in a way that people would buy? To figure this out:
The first thing I did was brainstorm the different ways we could build our referral program. I like to start by creating a visual model of the relationships between each of the parties involved. Then, I define the questions that need to be answered in order for each party to be satisfied with our solution.
I decided a referral program model should answer the following questions:
What the company was ultimately looking for, is a way to acquire customers for a lower cost than the average acquisition on facebook. I took some reasonable models and compared estimated completion rates and cost of goods needed to fulfill each reward.
I presented this sheet to our finance and executive team.
This model worked best for our team:
One disagreement within our team was that some folks did not want to provide too large of a discount to recipients, as it would attract a customer that could be priced out of the standard product cost. What would be the point of offering a free pack to someone that wouldn’t eventually pay for it?
On the other hand, completely selfish programs tend to be unsuccessful.
We decided to set the standard 40% subscription discount as the baseline, but frame it to each party as a discount specific to the referral program. Then, we would test other offers against that baseline.
I needed two pages: one for existing customers and another page for freshly referred customers.
I started by iterating possible layouts of each page. Below is an example of some low-fidelity wireframes for the sender page which needed the following elements:
The recipient page was more simple, as it was just a wrapper around the existing product.
Below was my final low fidelity sketch for the layout of each page.
I created medium fidelity wireframes in a flow diagram using Miro (one of my favorite tools) to illustrate how the program would work and to collect feedback from the rest of the team.
The page for existing customers had 3 sections:
The first section was a summary panel for monitoring existing referrals and sending new referrals. It included a quick breakdown of the program incentives, a progress indicator displaying sent and remaining referrals, and finally, a call to action.
The call to action opened a modal that asked users for all the information necessary.
Below the panel, we showed detailed statuses on each of the friends that our customer referred.
Finally, we offered a more detailed, illustrated description of how the program worked, with one more call to action.
The recipient page wrapped our standard product page with the personal touch sending customers entered in the modal.
After we launched the site, we began gathering data on customer behavior. Below are some examples optimizations we tested, good and bad.
One of the first rates that caught our attention was the referral rate given that the customer opened the referral modal.
Opening the modal signaled an intent to send a referral, yet only 40% of customers actually sent one after the modal was opened. What was wrong?
Our page required users to log in before they could send a referral, and our hypothesis was that people were not remembering their login information, causing them to bounce.
The email sent to the customer already identified them, so logging in was technically an extraneous step. We implemented a token system that allowed users to log in through a special link in their email, which would bypass logging in only on referral program pages.
This was a striking success! The conversion rate given that they opened the modal shot from 40% to about 97%.
Our original email to the recipient only said “You have a gift.” I liked the curiosity this created, but I wondered if there was a lack of trust when they received an email from a totally new company.
Much of our traffic came from “advertorials.” These were blog posts on our site to which we drove facebook traffic. They had a specific and proven selling formula. I decided to try including this same formula in the recipient emails.
This test was a striking failure! It reduced our recipient open rate by half.
We removed all the curiosity behind receiving a gift, and sent an annoying salesy email to people who had never heard of us. The formula worked, but only for customers who have had their interest previously peaked using a facebook ad.
After trying some other emails, we went back to our original message: “you have a gift”
Simple copy often works best.
If we successfully convinced an existing customer to send a referral, that customer would be more likely to send another. We wanted to see if reducing friction in the repeat send process would increase the number of referrals sent per person.
To make this happen, we changed the form to immediately prompt customers to send another referral after they finished the first one, while highlighting the limit of customers (e.g. “9 friends left”) to push customers towards the goal of maxing out their referrals.
This test was not striking, but it was significant. We averaged about 15% more referrals per sender.
In our analysis of this test, we created an interesting breakdown of our referral program users, dividing them by the total number of referrals they’ve sent. You can see this data in the graph below.
Most people sent 1 referral, followed by 2 referrals and 3 referrals, as expected, but the following count was the most interesting. The 4th most popular number of referrals to send was 10 referrals.
These customers would account for a disproportionate amount of our referrals, and backed my point that a max goal pushed users to send more referrals.
The idea to give customers a space to write a personal note proved to be one of the most delightful aspects of the program. Our customer service team did an excellent job of delighting each customer, and this provided an avenue for new customers to express their happiness.
The project scaled to become a reasonable portion of the company's revenue. It never surpassed the facebook campaigns, but the cost of acquisition was the cost of a single pack, which was an order of magnitude less than the $30-$50 average cost to acquire a customer on facebook.
For no additional cost, we were able to multiply the amount of product sold from a single facebook conversion.
