Article
How to design a referral program - andrewchen
andrewchen.com
Quoted on this wiki
Every place a page here uses this source, in the order the words come in it.
Why a referral program? Referral programs — the “give $5, get $5” offers you see in many apps — have become popular in recent years. They have big advantages over paid marketing channels, in that you give your CAC to your users, who then spend it within your product, as opposed to handing it over to Google or Facebook. Because they are a form of viral marketing — utilizing your network of users to bring in more users — they tap in to your product’s network effects, as I describe in The Cold Start Problem. This is particularly useful for products that target high acquisition cost niches, whether that’s crypto users or on-demand drivers, whose CAC are often >$200, since the users often know each other. “A successful referral program can be 20-30% of your acquisition mix, as one of several acquisition loops.” The history of the referral program How did a structured form of customer referrals come into being? It’s said that the first documented referral program was created by Julius Caesar, who in 55 BC would paid his soldiers 300 sestertii (something like a third of their annual pay) to refer a friend to join the army. And thousands of years later, we still use, plus or minus, the same idea. It seems as though every consumer app has implemented some form of a referral program, though I argue it really kicked off in ~2008, which is when Dropbox’s innovative referral program was rolled out.
Why a referral program? Referral programs — the “give $5, get $5” offers you see in many apps — have become popular in recent years. They have big advantages over paid marketing channels, in that you give your CAC to your users, who then spend it within your product, as opposed to handing it over to Google or Facebook. Because they are a form of viral marketing — utilizing your network of users to bring in more users — they tap in to your product’s network effects, as I describe in The Cold Start Problem. This is particularly useful for products that target high acquisition cost niches, whether that’s crypto users or on-demand drivers, whose CAC are often >$200, since the users often know each other. “It’s not a silver bullet, but it’s worth adding to complement other marketing efforts.” The history of the referral program How did a structured form of customer referrals come into being? It’s said that the first documented referral program was created by Julius Caesar, who in 55 BC would paid his soldiers 300 sestertii (something like a third of their annual pay) to refer a friend to join the army. And thousands of years later, we still use, plus or minus, the same idea. It seems as though every consumer app has implemented some form of a referral program, though I argue it really kicked off in ~2008, which is when Dropbox’s innovative referral program was rolled out.
The Ask Product folks often start by agonizing over the ask. They wonder if it’s too trivial to create a “Get $5, Give $5” referral program, or if that’s too basic. But I think that’s the wrong place to focus — after all, you can always word smith and test many variations later once you have the program up and running. “What you find, after instrumenting all your referral UI, is that there’s just a certain conversion rate on this screen.” Thus, make the referral ask part of the main flows. After the user is buying something within your app, ask them if they want $X cash back now, by inviting someone. Or if they interact with a friend within the app — assuming the product allows invitations of some sort — follow up by asking if they want to invite others. And add it to the onboarding flow, and at the end of key transactions when the user is otherwise done, and you might as well capture engagement. And for god’s sake, don’t make it look like “an ad” with big splash text and graphics — make it plan, like something that’s part of the normal UI where the user can interact.
And of course, the simplest thing to do is a “give $5, get $5” and give that offer to everyone, in an untargeted fashion. But a product leader soon realizes that this is inefficient — perhaps it’s best to give some users $15 and others $5, depending on their value. This is exactly what many marketplace companies have done, when it’s easy to segment their network into high-value cities like New York and SF versus, say, Memphis — you can set custom referral amounts in each place. But why stop at cities? Perhaps you do an analysis and figure out certain leading characteristics of high-value users as their account balance, or the types of other apps they use, or otherwise — once you think of this as personalizing an ephemeral offer to users, then you can run whatever promotions you want. “As a result, most referral programs have tended towards dollars over time, though I think the important idea is to prioritize new, outside users, and think about how to make the incentive as concrete as possible.” There’s the basic question of how to set the incentive amount. Typically this is based on a basic calculation of CAC/LTV, which has major weaknesses as it doesn’t take into account cannibalization (which we’ll discuss later). Instead, the focus is often to pick a simple number — if you know that the average user who signs up spends $20, then you can create a referral program that rewards a $5 give/get with some margin of safety. But the big lever on the incentive, of course, is to increase the amount — and the largest amount generally comes from tiered offers that have some form of breakage. An example of this is to say, “$100 when you sign up and buy 5 things” rather than “$5 when you sign up.” Given that the difference between a signup and a repeat conversion rate might be 100x, you might be able to safely raise the amount 20x. At Uber, this went so far as to combine two distinct numbers: A headline number that combined both the initial signup conversion as well as the first month’s earnings (again, as long as you drove X trips in the first few weeks). This resulted in a $3000+ number, a huge upgrade from the initial $200 numbers we started with. These larger headline numbers always tested much better on A/B tests, whether in email marketing or banner form, and while it might feel like the reward becomes unattainable, it’s possible to create a second or third or fourth tier to go along with the big headline number. You could say, earn $X when you fulfill all the requirements, but then a smaller number, $Y, when you only fulfill a few. That way you get the marketing impact of the big number but still have a fallback for users who don’t hit all the milestones.
The last aspect of the incentive structure I’ll discuss is a symmetric versus asymmetric offer — that is, should it be a “give $20, get $5” or “give $5, get $20.” Which one sounds better to you? This is anecdotal, but in testing I’ve seen, the inviter-centric amount generally works better — that is, catering to their self interest. However, I’ve also seen B2B contexts where in a professional setting, people tend towards inviting more if they are perceived as altruistic, giving out a large $ discount to others. In the end, probably just worth A/B testing to see what works best. “You’ll need some kind of ROI metric to drive the strategy of the referral program.” After all, if the lifetime value of these users exceeds the cost of acquiring them, shouldn’t you just go full steam ahead? Well, maybe. What if you can get much cheaper acquisition via another channel, like TikTok ads. Then any dollars that go to this might be better spent on ads. Or what if improving a referral program takes engineering team away from critical features? So yes, of course look at the CAC/LTV of your referral initiative, but think about how you might compare the tradeoffs against everything else.
Ask When do you ask the user to refer? Why do you refer? Is it tied to a holiday, or a particular promotion? What’s the message? “Target Which users do you target? All of them? How do you set referral amounts?” Incentive What’s the incentive, is it extrinsic ($) or intrinsic (points, storage, etc)? Do you give the inviter or recipient the same reward?
And of course, the simplest thing to do is a “give $5, get $5” and give that offer to everyone, in an untargeted fashion. But a product leader soon realizes that this is inefficient — perhaps it’s best to give some users $15 and others $5, depending on their value. This is exactly what many marketplace companies have done, when it’s easy to segment their network into high-value cities like New York and SF versus, say, Memphis — you can set custom referral amounts in each place. But why stop at cities? Perhaps you do an analysis and figure out certain leading characteristics of high-value users as their account balance, or the types of other apps they use, or otherwise — once you think of this as personalizing an ephemeral offer to users, then you can run whatever promotions you want. “As a result, most referral programs have tended towards dollars over time, though I think the important idea is to prioritize new, outside users, and think about how to make the incentive as concrete as possible.” There’s the basic question of how to set the incentive amount. Typically this is based on a basic calculation of CAC/LTV, which has major weaknesses as it doesn’t take into account cannibalization (which we’ll discuss later). Instead, the focus is often to pick a simple number — if you know that the average user who signs up spends $20, then you can create a referral program that rewards a $5 give/get with some margin of safety. But the big lever on the incentive, of course, is to increase the amount — and the largest amount generally comes from tiered offers that have some form of breakage. An example of this is to say, “$100 when you sign up and buy 5 things” rather than “$5 when you sign up.” Given that the difference between a signup and a repeat conversion rate might be 100x, you might be able to safely raise the amount 20x. At Uber, this went so far as to combine two distinct numbers: A headline number that combined both the initial signup conversion as well as the first month’s earnings (again, as long as you drove X trips in the first few weeks). This resulted in a $3000+ number, a huge upgrade from the initial $200 numbers we started with. These larger headline numbers always tested much better on A/B tests, whether in email marketing or banner form, and while it might feel like the reward becomes unattainable, it’s possible to create a second or third or fourth tier to go along with the big headline number. You could say, earn $X when you fulfill all the requirements, but then a smaller number, $Y, when you only fulfill a few. That way you get the marketing impact of the big number but still have a fallback for users who don’t hit all the milestones.
Target Which users do you target? All of them? How do you set referral amounts? “Incentive What’s the incentive, is it extrinsic ($) or intrinsic (points, storage, etc)? Do you give the inviter or recipient the same reward?” Payback What is the success criteria for the program? How do you think about cannibalization?
There’s the basic question of how to set the incentive amount. Typically this is based on a basic calculation of CAC/LTV, which has major weaknesses as it doesn’t take into account cannibalization (which we’ll discuss later). Instead, the focus is often to pick a simple number — if you know that the average user who signs up spends $20, then you can create a referral program that rewards a $5 give/get with some margin of safety. But the big lever on the incentive, of course, is to increase the amount — and the largest amount generally comes from tiered offers that have some form of breakage. An example of this is to say, “$100 when you sign up and buy 5 things” rather than “$5 when you sign up.” Given that the difference between a signup and a repeat conversion rate might be 100x, you might be able to safely raise the amount 20x. At Uber, this went so far as to combine two distinct numbers: A headline number that combined both the initial signup conversion as well as the first month’s earnings (again, as long as you drove X trips in the first few weeks). This resulted in a $3000+ number, a huge upgrade from the initial $200 numbers we started with. These larger headline numbers always tested much better on A/B tests, whether in email marketing or banner form, and while it might feel like the reward becomes unattainable, it’s possible to create a second or third or fourth tier to go along with the big headline number. You could say, earn $X when you fulfill all the requirements, but then a smaller number, $Y, when you only fulfill a few. That way you get the marketing impact of the big number but still have a fallback for users who don’t hit all the milestones. “The last aspect of the incentive structure I’ll discuss is a symmetric versus asymmetric offer — that is, should it be a “give $20, get $5” or “give $5, get $20.”” The Payback You’ll need some kind of ROI metric to drive the strategy of the referral program. Are you spending the right amounts, or should you increase the numbers? How much product effort should be put into implementing new surface areas? Etc. Is it working? These fundamental questions are often answered with a classic CAC/LTV analysis, and there’s a reason to doing that.
The Incentive You’ll note in the original Dropbox offer, the incentive itself was storage space not dollars — this is the dilemma of intrinsic versus extrinsic rewards for users that participate in your program. Many referral programs for mobile games tend towards intrinsic rewards as well, earning you points if you invite friends. The advantage of intrinsic rewards is that it’s particularly cost effective when the incentive is something you can control, like points. The problem with intrinsic rewards, of course, is that external users — people who have never heard of your product — are the least responsive to points or otherwise. Dropbox’s storage offer is maybe somewhere in the middle, since it’s at least a concrete form of value. As a result, most referral programs have tended towards dollars over time, though I think the important idea is to prioritize new, outside users, and think about how to make the incentive as concrete as possible. “You could say, earn $X when you fulfill all the requirements, but then a smaller number, $Y, when you only fulfill a few.” The last aspect of the incentive structure I’ll discuss is a symmetric versus asymmetric offer — that is, should it be a “give $20, get $5” or “give $5, get $20.” Which one sounds better to you? This is anecdotal, but in testing I’ve seen, the inviter-centric amount generally works better — that is, catering to their self interest. However, I’ve also seen B2B contexts where in a professional setting, people tend towards inviting more if they are perceived as altruistic, giving out a large $ discount to others. In the end, probably just worth A/B testing to see what works best.
But we get ahead of ourselves. Let’s start first by looking at how a referral program is usually defined. “There are variations of course, as some focus on giving and getting dollars. Some ask you to share a code, or a link, or connect your addressbook to invite friends.” One way to organize all these variations is to divide them into the following — and you need to answer a series of questions for how you structure the program:
One way to organize all these variations is to divide them into the following — and you need to answer a series of questions for how you structure the program: “Ask When do you ask the user to refer? Why do you refer? Is it tied to a holiday, or a particular promotion? What’s the message?” Target Which users do you target? All of them? How do you set referral amounts?
The real question is, WHERE do you make the ask? And my answer is simple: Ask many times, in many places, with different messages, and in-context with whatever action you’re asking the user to take. What you find, after instrumenting all your referral UI, is that there’s just a certain conversion rate on this screen. And that most users, if you put the referral functionality on a banner somewhere random in the product, simply don’t interact with the referral features. Rather than trying to raise conversions, instead, show the screen more often — get more impressions! “After the user is buying something within your app, ask them if they want $X cash back now, by inviting someone. Or if they interact with a friend within the app — assuming the product allows invitations of some sort — follow up by asking if they want to invite others. And add it to the onboarding flow, and at the end of key transactions when the user is otherwise done, and you might as well capture engagement.” One of my favorite ideas from Uber is the concept of “holidizing” a referral campaign. For drivers, as the holidays approached, you might tell them to earn extra money towards gifts and festivities, by participating in a referral program. Or for the run up to a major concert in town, you might run a special tiered campaign where referring 1 friend gets you X, but 5 gets you 5*X and a huge bonus on top. There’s something great about freshening up the messaging each month to align to major holidays, with new amounts, new imagery, and otherwise.
Incentive What’s the incentive, is it extrinsic ($) or intrinsic (points, storage, etc)? Do you give the inviter or recipient the same reward? “Payback What is the success criteria for the program? How do you think about cannibalization?” Let’s use an example to describe this.
Target Which users do you target? All of them? How do you set referral amounts? “Incentive What’s the incentive, is it extrinsic ($) or intrinsic (points, storage, etc)? Do you give the inviter or recipient the same reward?” Payback What is the success criteria for the program? How do you think about cannibalization?
There’s the basic question of how to set the incentive amount. Typically this is based on a basic calculation of CAC/LTV, which has major weaknesses as it doesn’t take into account cannibalization (which we’ll discuss later). Instead, the focus is often to pick a simple number — if you know that the average user who signs up spends $20, then you can create a referral program that rewards a $5 give/get with some margin of safety. But the big lever on the incentive, of course, is to increase the amount — and the largest amount generally comes from tiered offers that have some form of breakage. An example of this is to say, “$100 when you sign up and buy 5 things” rather than “$5 when you sign up.” Given that the difference between a signup and a repeat conversion rate might be 100x, you might be able to safely raise the amount 20x. At Uber, this went so far as to combine two distinct numbers: A headline number that combined both the initial signup conversion as well as the first month’s earnings (again, as long as you drove X trips in the first few weeks). This resulted in a $3000+ number, a huge upgrade from the initial $200 numbers we started with. These larger headline numbers always tested much better on A/B tests, whether in email marketing or banner form, and while it might feel like the reward becomes unattainable, it’s possible to create a second or third or fourth tier to go along with the big headline number. You could say, earn $X when you fulfill all the requirements, but then a smaller number, $Y, when you only fulfill a few. That way you get the marketing impact of the big number but still have a fallback for users who don’t hit all the milestones. “The last aspect of the incentive structure I’ll discuss is a symmetric versus asymmetric offer — that is, should it be a “give $20, get $5” or “give $5, get $20.”” The Payback You’ll need some kind of ROI metric to drive the strategy of the referral program. Are you spending the right amounts, or should you increase the numbers? How much product effort should be put into implementing new surface areas? Etc. Is it working? These fundamental questions are often answered with a classic CAC/LTV analysis, and there’s a reason to doing that.