Thirty percent in 2020. Six when I picked it up.
This was a complete overhaul of the program, not a small fix, and I was responsible for the strategy end to end. Refer a Friend was on HelloFresh’s meta backlog of top priorities, and the team already on it was focused on a visual refresh. I took it upon myself to jump in, pitched senior leadership my hypotheses for why it was failing and was rewarded with the project. We launched in the US and Canada first, then Germany, where HelloFresh is headquartered.
It wasn’t failing at the page, it was failing to be reached. Everyone who had touched it seemed to think more places meant more traffic, and my site-wide audit found it plastered everywhere, even in the first modal a new customer sees, before they’ve cooked anything.
Visited the referral page
17.83% 10.45% −41%
Share of active customers · North America highlighted
We don’t give them any information.
The sender was asked to send a free box without being told what a free box is: how many meals, for how many people, and whether shipping is free. Six of the seven people we studied asked what their friend would get.
The friend got a templated email that didn’t say who it was from and honestly felt like spam, and “Claim your gift” threw them into our 21-step signup, big screens meant for a warm lead who’s already invested in signing up. Once I started treating the friend as a proper new acquisition funnel instead of just a referral, it was clear this is an awful lot to ask somebody. It’s misleading too: my mom had to pay for the free box I sent her, because it only covered a set number of people and days.
“I don’t want to make a promise that doesn’t wind up being true. I don’t want to look bad to my friend, or have the guilt or shame of misleading them.”
Your friend lands here. 21 steps later. The card is required at step 21, and the menu comes after payment.
Friend lands here
Card required. Menu afterThe friend comes first.
In our research, 86% of the people who accepted the offer did it because their friend sent it. It’s not “I got a free box,” it’s “my friend Sam sent me this,” and I kept that power of the sender in mind throughout.
Even at four steps, you’re asking me to sign up before I know what it is, so I designed a page that shows the friend what HelloFresh is, personalized with who sent it and one of the sender’s favorite meals. A 20-second clip of the exact box and how the ingredients are packaged shows what you’ll get better than any text could. Then we explain in simple terms that there’s no catch: try it for a week, cancel any time, and know that it’s a subscription.
Moments of joy.
We scaled down the touch points in the app and made the rest more intentional, and weekly invites after meals are picked went from under 100 to over 600. I was calling these moments of joy: the sigh of relief once you’ve picked the week’s meals and dinner is solved, and the box arriving safely on your doorstep. Both are the perfect time to think, you know who would love this?
The two moments: after the week’s meals are picked (web and mobile web), and when a box arrives.
Leadership brought the reward back.
The old positioning was money-heavy and wasn’t working, and the research said people weren’t sharing for the $40, so the team and I built the whole program around altruism, helping out a friend who’d had a baby or just moved. After about three weeks of building, we pitched it to senior leadership, who said, “I don’t like this, people only care about the money.” I did everything in my power to relay the evidence, and they said, “I get it, I don’t like it, please go back to the other direction.”
Unexpected pivot, for sure, but that’s the name of the game, and instead of dropping the altruism, I looked at how to massage the money so it wasn’t the number-one thing. Our best customers get the most credit, so I wanted something just as enticing that could also help the business.
I came up with three ideas: a reward like an Oura ring, the standard credit, and a raffle, which we ran for FIFA tickets. I built all three into one reusable component, so the offer can change each month as partnerships come and go.
FIFA raffle · US
+10.7%
Invites, but $11,700 lost after prize cost
Fissler cookware · Germany
+15.7%
Conversions, with €5,646 returned after costs
I offered to shoot it myself.
I wanted to build out a proper campaign. I still have contacts with studios in the city and with modeling agencies, so I said, “Give me four days, and I’ll shoot a whole campaign at no cost to the company.” I pitched it to leadership with every reason I had: I’ve been a photographer for 20 years, the website is someone’s only interaction with us until they receive the product, people very quickly scroll past AI imagery, and I just wanted to show real people in the moment.
They said no because they wanted AI for the speed, so I came back with mood boards and more ammo on why real people made more sense. If they were worried about time or money, I was offering my services, so there was no risk, only a plus. They said no again, so it was another pivot and another constraint. With that constraint in mind, I put together a plan for how to get the best AI imagery possible with the look and direction I wanted. Most of the website’s images are very studio-esque and don’t feel real, so I went for the real side: a messy kitchen, all from what feels like the same day.
More people shared. Growth took longer to follow.
What we can prove
>4.5%
Weekly US invites vs control, statistically significant
What the channel did
+20%
Weekly US invites, Q2 to early Q3
What it cost
9.7–12.6%
Invites that became customers, down from 15.4%
Invites are up more than 4.5% versus control, which, mind you, comes after six years of constant decline, and weekly invites were up 20% from Q2 to early Q3.
Invite conversion fell from 15.4% to 9.7–12.6%, but that’s because a much larger number of invites are going out. Before, a very small base was sharing and it shrank every year, so the conversion was a little higher on a smaller amount. The Q2 average was 1,318 new customers a week, and by late August we were looking at 1,658.
US figures. The lift over control is statistically significant. Everything was tested as one bundle, because platform migrations left no way to isolate changes, so no single part gets the credit. The 20% and the new-customer figures are time comparisons outside the test; late August was the first time weekly conversions passed the Q2 average.
Start with someone you have in mind.
When you reach this page outside a moment of joy, “send a free box to someone” makes it hard to imagine who would actually benefit. Tapping it suggests occasions like a new baby, a birthday, a move or a busy week, which brings people to mind and lets us hyperpersonalize the friend’s page, so Sam’s becomes baby themed. I’m well aware that more screens mean more friction, which is exactly why I want to A/B test it.
What does the next month look like?
In any subscription service the most important metric is churn, and I was highly focused on stopping people from canceling right after their free box. We see it most at the lowest loyalty levels, where the sender is purely doing it for the money and the friend takes the free meals and cancels.
After launch, the analytics team found that 52% of referred friends cost more than they brought in, and a referral from a long-standing customer was worth about three times one from someone new. That’s a finding about the channel, not a measured effect of the redesign.
Invites sent and new members joined can look really great in isolation, but the churn rate is where you get the real indication of how it’s doing.