How much is a customer worth? How much does it cost to acquire a customer?
If you can’t answer these basic questions, it’s hard to know if your business is sustainable or not. Your unit economics analysis will help you answer these questions. In this blog post, I will guide you through how I use unit economics to better understand a business. We’ll take a look at customer lifetime value (CLTV), retention data and customer acquisition cost (CAC), as well as discuss how to use them.
In this post, I will use a non-subscription e-commerce business as an example, but the general theory behind unit economics works just as well for other business models as well, including subscription e-commerce and SaaS.
If you want to skip the theory, just go to my (e-commerce) unit economics template in Google Docs.
Why Unit Economics?
Using unit economics analysis, you can estimate how much a customer is worth (CLTV) and how much they cost to acquire (CAC). If CLTV is significantly higher than CAC, the customer is worth much more than it costs to acquire and it probably makes sense to, for example, spend more money on marketing to acquire more customers.
CLTV – Customer Lifetime Value
CLTV (Customer Lifetime Value, also abbreviated LTV and CLV) is the net income that a customer generates to your business through purchases under the lifetime of that customer. Add all purchases by that customer, multiply by your gross margin and subtract any direct costs related to the purchases, et voilà! – you’ve calculated your CLTV.
Of course not all customers are the same, but aggregated customer behavior can help you predict the future. It makes sense to separate customers into different segments, as well as analyzing them all as one group. The most high-level analysis gives you a hint on how you perform in average, but you also need to understand that certain customer segments might behave very differently and the CLTV can vary greatly between customer segments.
Here’s an example of what a CLTV analysis overview might look like (explained below, click for higher resolution):
- Month: Month 1 is the month when the customer became a customer, i.e. made the first purchase.
- Percentage customers buying: In the first month, per definition, all customers make a purchase. The next month, on average 23.24% of customers makes at least one new purchase, i.e. are active. One year after becoming a customer, 10.00% of customers are active. These 10% are not necessarily a part of the 23% who ordered in the second month. Those groups could be totally separated from each other, but they both belong to the group of customers who first became customers in what is defined as month 1. Learn more about these numbers in the section Retention Cohorts below.
- Transactions per active customer: How many purchases does an active customer make in a month when the customer is active? The first month when a customer joins, it makes on average 1.05 purchases. The customers who make at least one order in any given following month, are on average making 1.25 orders that month.
- Transactions per customer: From all customers who joined in Mar-15, how many orders are made in a given month, per customer? You get this number by multiplying Percentage customers buying with Transactions per active customer.
- Average transaction value: This is simply the average order value (AOV).
- Total transaction value: The average monthly transaction value per customer from all customers who joined in Apr-15. Simply multiply Transactions per customer with Average transaction value.
- Gross margin: No explanation needed.
- Revenue: The average monthly revenue per customer from all customers who joined in Apr-15. This is the Total transaction value multiplied by the Gross margin. This is a key number, since it tells you how much monthly revenue you will make in the future from an average customer that you acquire today.
- Cost per transaction: Perhaps you have some cost per transaction, which is not already accounted for in the gross margin. This could for example be salary for the average support time needed or order transmission costs. Do not include fixed overhead costs unrelated to the order volume.
- Cost: The average transaction costs per customer, i.e. Transactions per customer times Cost per transaction (duh!).
- Profit: Revenue minus Cost (no surprise here!). This column tells you how much monthly profit you will make in the future from an average customer that you acquire today.
- Accumulated CLTV: Accumulated profit over time gives you the CLTV. In this example, after the first 12 months, an average customer will have generated an accumulated profit of €8.01. I call this number C1YV (Customer 1-Year Value). In this example, after 60 months (you can’t see it in the image), the accumulated profit is €20.54. The CLTV is obviously higher than the C1YV, but it’s also harder to predict how the customer will act 5 years ahead (or more!). For this reason, it can be convenient to distinguish between C1YV and CLTV (often approximated as C3YV or C5YV).
When I talk to entrepreneurs about unit economics, the most difficult part to grasp is the retention cohorts. A cohort is a group of customers. By tracking cohorts over time, you can learn about how new customers should act in the future. For this purpose, every group of customers who joins in the same month is defined as their own cohort. The cohorts’ average retention data is used to estimate the Percentage customers buying in the CLTV analysis overview above. Let’s take a closer look at how this works.
To estimate how new customers will behave in the future, you need to analyze how other customers have behaved in the past. In the case of retention, take a look at how many customers returned x months after they first became a customer. Segment your customers by the month they made their first purchase. Listing all this data can look something like this (click for higher resolution):
A few examples to clarify the table:
- Out of the customers who made their first purchase in Feb-12 (first row), 100% made a purchase in Feb-12 (first column), as per definition.
- The next month (Mar-12), 20.1% of the same customers made at least one order. In Apr-12, 16.3% of the Feb-12 acquisitions made at least one purchase. A common mistake is to assume that these 16.3% are necessarily related to the 20.1% ordering the month before. This is not the case! These percentages simply relate to the whole group of customers who joined in Feb-12. It could be true that the 16.3% are all a part of the 20.1%, but it could also be true that no single customer from Feb-12 made a purchase in both March and April.
- Out of the customers who joined in Nov-13, 21.7% made at least one order in Jan-14.
In the bottom of the table, you see an “Average retention” row. These numbers are simply the average retention numbers from the table above. For example, in average 23.8% of customers who joined on month made an order the next month (the average of 20.1%, 19.0%, 19.6%, 19.4% etc). As you can see from the blue raw data, the first-month retention was lower in 2012 than in 2013, probably because the company improved product, service, marketing etc. Don’t use older data than necessary, since if you are constantly improving the product (and retention), old data will be misleading.
Plotting the average retention gives you the black dots in this graph (excluding month 1 for clarity):
Given this data, you would like to estimate what the retention will be in month 25+. You can do this by calculating the best fitting trend line (the thin grey line). This consists of two parts: assume what kind of equation generates a line similar to the black dots, and calculate the right parameter values for fitting the trend line.
It turns out that in all e-commerce business I’ve looked into, the dots follow something very similar to a power function with a negative exponent, i.e. y = ax-b where a and b are constants between 0 and 1, x is the number months since the user joined minus one (e.g. for first month user can return, x=0, giving y=a), and y is the retention. I’m not sure why this kind of aggregated human behavior turns into a power function. If someone has a good guess, please add a comment below! What’s great about it though, is that you only need a few months of data to be able to somewhat reliably predict the future behavior!
Finding the best trend line when you have 23 data points and 2 unknown parameters isn’t too difficult. I won’t go into details, but you can read about it here if you’re interested. In my spreadsheet template I’ve created formulas calculating the constants for you, so you don’t have to worry about it.
In the graph above, the best power function trend line is y = 0.232x-0.339, which corresponds to the grey line in the graph. Now you can estimate your customers’ future retention using this formula. For example, in month 36, probably around 0.232×(36-1)-0.339 ≈ 6.95% will make at least one purchase. This data now goes into the CLTV overview analysis.
CAC - Customer Acquisition Cost
Understanding how much a new customer is worth in 1, 3 or 5 years is not enough to understand if your business model is sustainable. You need to compare the CLTV with the customer acquisition cost (CAC). This is quite straightforward: Divide the costs related to customer acquisitions by the number of new customers in the same time period, and you have you overall CAC.
Just like the CLTV, the CAC can be very different depending on the channel. For example, word of mouth is free, SEO is usually cheap (depending on how much efforts/salary you put into it), TV is often expensive, while SEM is somewhere in between. Unfortunately, most cheap channels are not easily scalable while the more expensive ones are.
Usually you separate paid and organic channels, or even paid/owned/earned channels. No matter what your marketing strategy looks like, make sure to understand what channels have low and high CAC.
In search channels (SEO and SEM), I also think it makes sense to distinguish branded from non-branded search queries. People searching for something including your company name in the query will behave very differently from people using generic words (perhaps not knowing about your company at all). You could also argue that quite likely, people googling your company name will find you even if you wouldn’t have a Google ad for that keyword.
As an example, in a given month, the CAC analysis might look like this:
I think the table is rather self-explanatory, but for example, this month €10,800 was spent on paid marketing channels, resulting in 100,986 visits and 1,934 orders from new customers and 1,445 orders from returning customers, giving an average CAC in paid channels of €3.41 and a CPR (cost per reorder) of €2.91. In total, the overall CAC is €2.86, but you may notice that it differs quite a lot between different channels.
The reason I’m separating the channels into scalable (paid) and non-scalable (organic) is because the organic channels are usually difficult to scale, while the paid channels typically (but not always) are scalable. No matter how much money you spend on Google, more people will not search for your brand name, so SEM (brand query) is a non-scalable channel. On the contrary, you can always buy more traffic from non-branded search queries by extending your campaigns by adding more ads and keywords, hence SEM (no-brand query) is a scalable channel.
Unkown Traffic Sources
You’re probably getting lots of traffic for which you can’t identify the source. Some of them saw your billboard ad and entered your URL in the browser. Some of them heard a friend talk about your awesome product. I have grouped all this non-attributable traffic as “Direct traffic”. In reality you want to try to break this group down into smaller units. For example, you can use an exit poll to estimate the effects of your billboard ads and try to break them out from the direct traffic.
Also, direct traffic can originate from other trackable channels. Someone can find you through SEM on their smartphone on their way home from work, and complete a purchase from their computer when they get home. Your tracking algorithms will probably count this as one non-converted SEM visitor and one converted direct traffic visitor. It would be more accurate to count this as only one visitor, converted through SEM, but it’s not yet an easy task to track cross-device behavior.
When doing marketing, both old customers and potentially new customers see your ads. How do you know how much of the marketing cost was used to acquire new customers and how much was used to retain old customers? My example is simplified. I’ve assumed that you can track how big part of the acquired traffic comes from old customers and I assume that the cost for a visitor is the same no matter if it’s a new or old customer. This might not be true, but simplifies the example a lot. You might want to modify the formula to better split costs between acquisitions and reorders. Also, for some channels, like TV, you might want to calculate CAC in some totally different way.
Brand Awareness and CAC
Over time, when your brand awareness increases, people will be more likely to come up with your brand name when thinking of brands in your category (“unaided” brand awareness). Even if they don’t come up with it themselves, when they see your brand they will recognize it more easily (“aided” brand awareness). This increases the likelihood that they will choose your product and you will see a higher customer inflow, both through more direct traffic and by higher click-through rates on ads and higher conversion rates on your landing pages. As an effect, your CAC will decrease.
When spending money on brand-building activities, it will be difficult to calculate the positive impact on your CAC. The activity itself might not bring any new customers who can be traced to that activity, but the brand awareness might increase, lowering your CAC on other channels. Unfortunately, there is no magic formula calculating the value of brand awareness activities, but that doesn’t mean they are worthless. Thanks @robahldin for suggesting adding this section about brand awareness!
Using Unit Economics
By now, you should know how to approximate your C1YV, CLTV and CAC. Now what? It’s time to draw some conclusions. You often hear about the CLTV/CAC quota. It should definitely be over 1 and the higher, the better – since the customer then is worth much more than it costs.
- CLTV/CAC<1: You’re in deep shit. It costs more to acquire a customer than they’re worth. This could be temporarily defensible under some circumstances, e.g. if you have a good reason why you need to build a certain user base before CAC can decrease, or CLTV increase.
- CLTV/CAC>3: Now we’re talking. A general rule of thumb is that the CLTV/CAC ratio should be 3 or higher for a healthy business. Many startups don’t see these numbers in the early days though. Note that if the ratio is very high (above 6 or so) it might be time to ramp up marketing or sales to attract more new customers.
- C1YV/CAC>1: Good! It means that it takes less than a year to cover the CAC. This is a great argument when requesting a higher budget for increased marketing spend.
Be careful – there is no guarantee that your CLTV and CAC will remain the same when you scale. Spending more money on marketing can increase the CAC because you might have to target new segments. The CLTV may also decrease when targeting a larger audience, e.g. when moving from early adopters to early majority (see Diffusion of innovations).
There is much more written about analyzing CLTV and CAC. I recommend you to look up information and discussions relating to your specific business model, target groups etc.
Comparing Customer Segments
While it can be very useful to keep your eyes on the CLTV/CAC quota on a high level, be careful when comparing CLTV and CAC for different customer segments. Even if CLTV/CAC=100, perhaps CLTV/CAC for customers coming from a group buying campaign can be less than 1, especially if they never return without you giving another discount. Don’t fool yourself to believe that everything that you do is smart just because your high-level metrics look great.
It gets really interesting when you start looking at different marketing channels, or even specific ads. A banner on yahoo.com may have a high CAC but low CLTV while the same ad on a small niche site may cheaply generate high-value customers. Just make sure that you’re not comparing apples with oranges (or any other fruit for that matter).
Highest CLTV Wins
Is a high CLTV only desirable because it makes your unit economics look good? No, there’s more to it!
The higher the CLTV, the more you are able and willing to pay for your customers and still keep a good CLTV/CAC quota. Imagine a competitive market, e.g. AdWords, where you have to bid for the customers. The more you are willing to pay, the higher your ad will rank, and the more clicks and customers you will get. Hence, a higher CLTV will generate more customers as well!
Hopefully, you now understand why it’s important to keep track of your unit economics and how it can be done. I’ve published my own (e-commerce) template as a Google Spreadsheet here.
The template is optimized for non-subscription e-commerce businesses. If you have another business model, you can probably adapt my template to work for you. For a SaaS business, I can recommend David Skrok’s article and Christoph Janz’ dashboard.
Good luck and please give me some feedback on how this guide and the template can be improved!