I joined Airbnb because I wanted to join a high-growth, tech company; I had no idea what affiliate marketing was. When I was asked in the interview if I could and how I would theoretically find 100 partners in Korea, I basically said 'why not' and that I’d figure it out. That was about 5 years ago.
My first few weeks, I felt like I was thrown into the deep end of a pool. Was there a difference between a partner and publisher? What were CPA networks? Why shouldn’t I reach out to big brands to partner with them? (Hint: brand partnerships need to be calculated, strategic efforts because they take a lot of bandwidth and often fail.). One thing I struggled with early on was understanding the marketing funnel and all the related concepts like LTV, CAC, Efficiency and Incrementality. I think I would have benefitted from someone giving me a Performance Marketing 101 lesson.
If you’ve been in the space for any small amount of time, you might think this is super simple and intuitive, but I think in the early stages of understanding performance marketing, it’s not very clear. There’s likely a moment in time where for every strong Account Manager that they get it and they feel comfortable doing gymnastics with funnels.
To succeed at HG, comfortability with understanding how pieces of performance marketing connect is a baseline requirement. So, I’m going to write that for our team and clients, the resource that I wish I had.
Start with The Funnel
A funnel is just the customer journey with conversion rates attached to it. If I know a user has to sign up for a free account, then a free trial, then they can become a paid subscriber, that’s my funnel. If I know for every 100 users who make an account, 10 sign up for a free trial and 2 become paid subscribers, I can calculate the conversion rates. In this case, the conversion rates (can be abbreviated as CV or CVR) across these funnel events is:
If I know these high level CVRs I can pretty quickly estimate certain things. For instance, if the above are averages for a program–which should always beg the question: ‘over what time frame?’ –over the last 12 months, if I notice that all the users that came from one channel have a 1% sign up to paid subscriber rate, I know that channel is driving users that are less likely to convert. (This could be okay if the value of those users is higher, but we’ll get into that later.)
Streaming Brand Example
Let’s look at a slightly more complex example of a sample streaming brand. Let’s say there’s a brand called HFlix that allows users to stream unlimited videos of hamsters for a $10 monthly subscription. HFlix’s marketing team has been running ads on Facebook and Google and is now looking to diversify into and build out their affiliate program. They come to Hamster Garage and we ask them about their funnel. Here’s what they tell us.
They also share the following conversion rates they’ve seen from their current marketing efforts.
Here are some things to think about when you observe a brand’s funnel and conversion rates.
Brand Specific
Every brand has a slightly different funnel, but most funnels usually share commonalities. In this example, HFlix has no sign up per se, but ‘New Account’ might be the user inputting their email and credit card information before they can start their trial. Many subscription brands will have a free trial before a user converts to a new subscriber, but some might not. Meanwhile, retail brands won’t have a free trial.
Understanding Events
When a user makes a new account or starts their trial, each of these can be considered an event. Every brand will define what events they are measuring based on how they think of their funnel. For instance, perhaps the HFlix team realizes that they can separate the New Account event into 2 events: ‘sign up’ and ‘credit card verified’ and now they track these as separate events. Maybe initially these were on 1-page with 2 input fields and now they are 2 pages. They might feel they are losing users at the point where users are entering their credit card information, but they would rather get a users email to retarget them later, so they want to split the events. Now, you could measure the conversion rate between events.
Sensible Conversion Rates:
Conversion rates share a story. In this case, it makes sense that the conversion rate from click to ‘new account’ is high since users are signing up for a free trial and not paying up front. Similarly, the users who make a new account are very likely to start the free trial since they aren’t necessarily paying right then. Finally, the drop off at the ‘trial start’ to ‘new sub’ conversion rate also makes sense since this is the moment users have to decide if they want to pay $10 to start their subscription. In some world, if HFlix decided that at the ‘new account’ event, they don’t require the user to put their credit card info, we might see the click to new account CVR go up and the trial start to new sub go down as the point of friction has moved.
From the above CVRs, we can also calculate the following conversion rates:
These are useful because we can always start with 100 clicks and know where those 100 users will end up. In this case, if a partner sends 100 users to the site, we can expect 8 of them to start a trial and 2 of them to become subscribers.
LTV and CAC
Conversion rates help us understand how many users we can move through the funnel from start to end. However, in reality, this is only part of the equation. The other half is asking how much are these users or customers worth? More people understand conversion rates than how to incorporate lifetime value (LTV) into their analysis. Yet, if you don’t understand both and don’t understand how they connect, you might miss some insights.
The first thing to understand about LTV is that it’s essentially a projection of how much a customer or user is worth over the lifetime of the brand’s relationship with the customer. For instance, for HFlix, since it's a subscription brand, we can calculate LTV by looking at the retention rate over some period of time. If on average, users that sign up for their first month of HFlix stick around for 8 months and they’re paying $10 a month, then the LTV is $80.
LTV matters because we have to factor how much a customer is worth when determining how much we can pay to acquire the customer. If a customer is worth more, we may be open to paying more to acquire the user. How much we are willing to pay to acquire a customer is called the Customer Acquisition Cost (CAC).
Let’s say HFlix says to us that they are willing to pay up to $20 to acquire a new subscriber. That’s their CAC for a New Sub. Because we know the CAC for one specific event, we can actually calculate the value across each point in the funnel. Let’s remind ourselves of the conversion rates for HFlix.
If we reverse engineer the above, we realize that:
All we’ve done is actually apply some basic math to figure out the proportional value of a different event in the funnel because we know the likelihood of that event converting to a New Sub.
To double check this, all we need to do is run it through an example - my favorite is always 100 clicks. If we paid $0.40 per click for 100 clicks, we’ll pay $40. Meanwhile, 100 clicks will lead to: 5 New Accounts, 4 Trial Starts, and 2 New Subs. HFlix is willing to pay $20 for 1 New Sub, so $40 for 2 New Subs. Our math checks out.
Summed up: if you know the conversion rates across the funnel and the CAC, you can determine the value of each event. This all assumes that the conversion rates are accurate, or stable, which is not always true, but we’ll delve into that later.
Paid Placements
Paid Placements or flat fee opportunities are one of the first areas that a deep understanding of conversion rates and LTV becomes important. However, before we jump into an example, we also want to understand that publishers also have a funnel.
So far, when we’ve been talking about funnels, we’re talking about the user journey when the user lands on a brand’s site (e.g. HFlix’s site). However, in reality, publishers are also driving users to their site, users are navigating their site, and then some portion of those users will ultimately click an affiliate link before landing on the brand’s site.
The click is the doorway between those entities. At the most abstract level, if I know how many clicks a publisher can send me and how much that’ll cost me, I could theoretically gauge if the partnership could work. The conversion rate of those clicks is probably the most important conversion rate because it most accurately reflects how strong the audience match is with some important nuances.
Let’s jump into an example. HuffFeed is a leading publisher with 10M monthly visitors. They’re proposing to write an article about HFlix for $10,000. Should we consider working with HuffFeed?
First, we ask the partner to tell us exactly what this will get us. They share with us that:
This is pretty typical of what partners will share with us for a media placement, but it’s not enough for us to gauge if this is worthwhile. For the $10,000 cost, we are getting 2 different promotional opportunities and both have different funnels before the user lands on our site. That is, since it’s not 500K readers and all 100,000 pageviews that will come to our site, we need to estimate the funnel down to clicks to our site. To do so, we ask the partner for the following:
- Newsletter: assuming that the link will link to our landing page, we want to know what the open rate and the typical click through rate for that specific placement.
- Article: what’s the typical click through rate for a 1,000 word article?
Let’s say they share the following:
Great, so now we can calculate a few things.
From our earlier funnel analysis, 14,500 clicks has a 2% conversion rate from clicks to New Subs, so we can expect 290 subscribers from these clicks. This would be a cost of about $35 per subscriber. This is double our accepted CAC for a New Sub of $20. At this point, we can make a decision on if we think it’s worth it because we think there’s value to being published in HuffFeed beyond the New Subs we immediately acquire.
Efficiency
The final equation here is efficiency. It’s effectively the same as Return on Ad Spend (ROAS). It’s understanding how much you’re getting for what you’re paying. Efficiency is a function of LTV and CAC.
In the above scenarios, we said that HFlix is willing to pay $20 for a New Sub and that the LTV was $80, which we calculated was from the average New Sub staying for 8 months. With these numbers, we get the following:
In the example of the paid placement we shared, the efficiency would be:
While a 200%+ efficiency sounds great, in reality, we’re only looking at LTV relative to marketing costs. In reality, for most businesses, there are other costs. For instance, if, instead of HFlix, we were looking at a candy subscription box, if the cost of producing each box is $59, then a $20 CAC means the company is at least making $1 per new sub. Yet, in this paid placement example, the company is losing $14 per New Sub since their cost ($59) plus the CAC ($35) is $14 above the expected LTV.
Why aren’t we using HFlix as an example and thinking about a candy subscription box? Because as a content streaming platform, HFlix may have effectively zero marginal cost to service a New Sub. I’m bringing this up because there are always insights about a business when you consider the story the LTV, CAC and Efficiency are sharing. In fact, increasing efficiency is a benchmark of a maturing company. Often, startups allow for lower efficiency because they are venture-backed and trying to grow in a market. Larger companies may even do this in new, emerging markets, whereas, in their mature markets, they may have a higher efficiency requirement.
So, efficiency requirements can change. At bare minimum, a 100% efficiency means that you are paying $1 for every $1 of value you are getting. Less than 100% efficiency means you’re losing money and higher than 100% means you’re losing money.
Going Deeper into LTV
All of these numbers we’re discussing are usually determined by the brand. As an agency, we’re shared what the average LTV is, what an acceptable CAC is, and what the efficiency targets are. However, it’s imperative that we understand how some of these numbers are calculated so we can seamlessly calculate things as needed. Likewise, with time, you start internalizing certain insights and realize how everything is connected based on the relationships between these key numbers.
Typically brands provide us with an acceptable CAC. For instance, HFlix is willing to pay up to $20 for a New Sub. That’s enough for us to work with partners to understand the funnel. We don’t need to know what the LTV is or what the company’s efficiency benchmarks are if we have the CAC.
This is because CAC can be calculated using LTV and Efficiency. If a brand knows how much a customer is worth and how much of that value they need to keep, they can provide us with a CAC number and as long as we are acquiring users at that CAC, it’ll work for them… theoretically.
In reality, the above assumes that LTV is the same per user, when in reality, this isn’t true. Let’s say that users that are avid readers of HuffFeed love hamsters and when they sign up for HFlix, they stick around for 24 months. Subsequently, the company is actually making $240 over the course of the customer’s lifetime.
If HFlix needs 400% efficiency, then we can safely pay up to $60 for these specific users. Namely:
However, most companies don’t have custom LTV for cohorts or per user. Instead, falling back on an average LTV with a set efficiency standard to arrive at an acceptable CAC is what makes sense for most brands. It may make sense for brands to reevaluate the average LTV on a given frequency, for instance, every year, but for the most part, using a snapshot LTV and desired efficiency to determine CAC works for many brands.
Incrementality
The final piece of the equation that you might encounter is brands trying to incorporate incrementality into these equations. While understanding how exactly brands determine incrementality across channels is beyond the scope of this piece, we can understand how we can leverage that number once we have it.
For instance, if we are told from HFlix’s data science team that the affiliate channel is actually only 80% incremental. Imagine on a monthly basis HFlix is driving 25,000 New Subs across all of their marketing activities, and affiliate drives 5,000 of these New Subs. 80% incremental means that if the company turned the affiliate channel off, instead of losing 5,000 New Subs, they’d theoretically only lose 4,000 New Subs.
So how do we take these insights into account in our calculations?
If right now, we’re driving 5,000 New Subs per month and the company believes $20 to be an acceptable CAC for New Subs, then we’re spending $100,000 in the affiliate program. That’s how much we’re paying partners. Likewise, if we assume that the LTV of a customer is $80, then 5,000 New subs are worth $400,000 to the company over the lifetime of the customers.
However, we know that only 80% or 4,000 of these New Subs are incremental, so the Incremental LTV we’re driving is $320,000 and likewise we can only pay $80,000 for these New Subs, not $100,000. However, in order to get 4,000 Incremental New Subs, we need to drive 5,000 New Subs.
So, we adjust our payout from $20 per New Sub to $16 per New Sub. If we’re paying $16 per New Sub for 5,000 New Subs, we’ll end up spending $320,000 in commissions to drive 5,000 New Subs, of which 4,000 are Incremental New Subs.
What the marketing team is effectively saying is that they believe that 1,000 of those New Subs would have signed up anyway because maybe they saw a billboard or television commercial. The company may have spent $80,000 on that campaign and so when you combine that spend with the $320,000 of commissions, you have actually paid $20 to acquire each new user across multiple touch points, but the affiliate channel was only 80% incremental.
