Copy Notion Unit-economics Glossary Page for your team. The link is in Telegram. Glossary is a tool to revisit metrics when you forget what they mean.
We gain new knowledge only by testing a hypothesis. All knowledge in the universe, including scientific knowledge, is created through testing hypotheses. There are no other ways. In the world of products, only one out of ten hypotheses succeed at best. So, most likely, you will fail with most of your products. Therefore, the main task is to learn how to quickly and cheaply test hypotheses.
It is believed that an idea or hypothesis is worth nothing. This is a popular misconception. A fast team can test 2–4 hypotheses in a month. The cost of such a team ranges from 25–100 thousand USD per month, depending on the country. Therefore, the cost of a single hypothesis idea ranges from 12,000 to 50,000 USD. If you are starting from scratch with just an idea, you are not starting from zero; you are starting from a negative position.
Data-driven decision-making is crucial for both early-stage startups and large ventures. It provides the necessary information to assess a company’s overall health and viability as it operates in the market. This approach helps steer your company in the right direction and quickly and transparently evaluate the impact of one change or another in the product.
Unit economics answers whether you earn on your user (unit) or not. To understand this, you need to find out:
For most online businesses, the unit will be the user. To avoid complexities in the future, always stick with the user, even though you can take anything as a unit.
Important to understand. Unit economics and Finances are different tools for different goals. These are two different views on finance. Finances are needed to calculate PNL accounting, and investments — this is a projection onto the time axis (by months). And unit economics looks to the future because its main idea is scalability.
If you spend more money on user acquisition than you earn on it, then scaling such a business does not make sense. Sooner or later, the money will run out, and your business will cease to exist.
The unit economics is considered to understand whether it is possible to develop this business. Can it be scaled through this growth channel or not? How much should we raise LTV (revenue per user) or lower CPA (cost of user acquisition) for the economics to converge.
Let’s introduce the axiom: The cohorts approach should be used in unit economics to obtain adequate metrics and draw accurate conclusions.
It is necessary to calculate unit economics separately for different cohorts. This enables you to observe how the economics varies across different segments. It is possible that your unit economics may converge in some segments but not in others. Otherwise, you will end up with a jumbled picture that provides fewer useful insights.
Segmenting cohorts allows you to compare groups of users from different time periods and analyze users with different characteristics, such as acquisition channels, devices, platforms, regions, and more.
A cohort refers to a group of users identified based on specific characteristics. For example:
The example below (table) demonstrates that if you don’t separate the data from mobile and desktop channels, you might wrongly perceive that the overall performance of the Facebook channel is much worse compared to other channels. However, this assumption is incorrect. It is possible that the issue lies specifically with the mobile version of advertising or your application, affecting the display resolution and leading to poor results.
Considering that Facebook offers the lowest cost per user (in this example), without analyzing the data separately for mobile and desktop users, you might make the mistake of discontinuing Facebook advertising and incur losses as a result.
This article is more for beginners, so we will skip the cohorts analysis when considering case studies.
The formula for calculating the unit economics comes down to comparing two indicators: AMPU (LTV) vs CPUser (CPLead).
If you don’t know the AMPU and CPUser metrics, you will need to perform the following steps for a specific user cohort:
2. Find out how much money you spent on attracting a cohort of users;
3. Divide the cost by the number of users in the cohort and get the cost of attracting one user in the cohort;
4. Calculate the gross profit from the cohort of users;
5. Predict what the gross profit from this cohort will be in the future;
6. Calculate the LTV of a user from a cohort;
7. Compare LTV and CPA to understand the unit economics for the analyzed cohort.
Okay, but why do we need buzzwords like UA, C1, C2, COGS, COGS1_1st Sale, APC, ARPC, AMPU, AMPPU, AC?
The framework described below aims to decompose high-level metrics such as CPA and LTV into components, enabling teams to identify specific areas for improving the unit economics of the product. The framework consists of four main sections, each with its own set of metrics:
Now you can open Excel and actively engage with the figures and concepts presented. I have created a simplified Google Sheet Framework Template with actual business figures. To maintain anonymity, I have included a factor that allows you to adjust the ratios. I have used Iliya Krasinski’s framework within the unit economics realm as it is user-friendly and straightforward.
When you come across incomprehensible words, look for them in Glossary. The Glossary includes all the essential metrics you can readily apply to your company. Here is the link: [TG]
The best way to learn is to start doing something with examples.
The main thing is not to memorize the formula, but to understand the essence, then you yourself can easily derive these formulas. So let’s take a look at the cases. Now we will calculate the unit economics, find the shoulders of the metrics, evaluate the existing product hypotheses, and put forward new ones. To assist you, I have provided simple templates below. These templates offer initial steps that you can follow.
Sheet 1 in Excel. Let’s play. You play as a Founder of a music service like Apple Music, and you want to make the world even better and create another music service that will be better than existing solutions.
You have done good market research, looked at all music apps, found out the metrics on the internet, and asked domain experts at thematic conferences, and already now you have some target numbers.
Given:
Step 1: choose the right AMPPU model. All business models are differentiated, just with AMPPU calculation. The picture below shows the formulas for different business models. For our case general AMPPU formula is
AMPPU = AvPrice × Margin × AvPaymentCount,
Let’s calculate AMPPU for one-month and one-year cohorts.
Step 2: Find C1.
We need to understand the conversion rate to a customer (C1). In this case, we know that the conversion rate to a customer is 1%. For your business, calculate the conversion rate to purchase for each marketing channel.
Note. Everyone counts conversion in different ways, but How to calculate it right? C1 — or conversion to the buyer should be calculated from user acquisition, not just user registration. Often, teams calculate conversion only for the steps they are responsible for.
For example, the marketing team is responsible for leads and registrations, while the product team calculates conversion to the customer from registrations, then:
conversion from user → lead = 10% conversion from lead *→ customers = 11%
So the conversion from user to customer would be = 10%* × 11*% = 1,1% — not such a pretty figure for reports, but it clearly shows the problem that only 1,1% of users are converting to customers. This indicates the need to build funnels and identify blockers in the user journey.
Step 3: Counting AMPU
The difference between AMPU and AMPPU is that AMPPU shows marginality per paying customer, while AMPU shows marginality per visitor (not buyer). By calculating AMPU, you can evaluate the effectiveness of your marketing campaign in every marketing channel.
The calculation for AMPU30 is as follows: 1% × 4€ = 0.04€.
The calculation for AMPU365 is as follows: 1% × 21€ = 0.21€.
Step 4: Comparing CPUser and AMPU
To determine how much you spent on user acquisition, select the user cohort you are interested in and calculate the user acquisition cost. In our case:
CPUser(365): 1€ > 0.21€
It turns out that we are only losing money when scaling. Do we have billions of investments to keep burning money for years?
You can calculate (CPUser — AMPU) in another way with Gross Profit to check if your calculations are done correctly.
Gross Profit — is the income from the user flow (gross revenue), excluding the costs of user acquisition cost and fix cost.
In your company, it might be called something different, and that’s normal. What matters is the practical meaning behind the concept. We are interested in the gross profit in the definition mentioned earlier — the profit that remains after subtracting variable expenses.
A good criterion for determining variable expenses is that they typically increase in proportion to revenue. Let’s consider a few examples:
- If you sell a mobile game through the App Store, you need to deduct app store commissions from your revenue, as they increase with sales. However, you don’t need to subtract the costs of the team developing the game. These are fixed costs and do not directly affect the cost of one copy of the product.
- If you develop and sell complex B2B software, and for a specific deal, a team of engineers works on integrating the product for the client, the expenses for this integration should be deducted from the revenue. However, once again, the costs for developing the software itself should not be considered, as they are not directly related to a specific deal and remain fixed expenses for R&D.
- If you sell mittens through an online store, the variable expenses within a specific deal will consist of the cost of purchasing the mittens, delivery, payment system commissions, and other associated expenses.
Gross profit(30) = AMPPU(30) × buyers = 3.6€ * 1,000 = 3,600€ Gross profit(365) = AMPPU(365) × buyers = 21€ * 1,000 = 21,000€
Gross profit(30) = revenue — COGS = 12,000€ — 8,400€ = 3,600€ Gross profit(365) = revenue — COGS = 70,000€ — 49,000€ = 21,000€
AMPU(30) = Gross Profit / Users = 3,600€ / 1,000 = 0.04€
AMPU(365) = Gross Profit / Users = 21,000€ / 1,000 = 0.21€
Note: A common mistake when calculating a unit economics is to calculate based on Revenue and not Gross Profit. This error will not be critical for some types of businesses since their revenue and gross profit are very close to each other (many SaaS businesses). But for many other services, this error can be very expensive since their revenue implies significant costs for its provision (Uber, Dropbox, and others)
Step 5: Some Conclusions from this Case
Playing with the Excel numbers, we can quickly understand that the economics starts converging to €0 at C1=5% and Av. Payment Count (or Lifetime in our SaaS case) = 7 months.
AMPU = 5% × €10 × 30% × 7 = €1.1
Additionally, we recall that CPUser = €1.
To be honest, achieving an Av. Payment Count of 7 is hard. The average number in SaaS is usually between 1.2–2.
What do figures reveal?
To understand this, let’s assume that we want to earn one million EUR with our music service.
For each year of our cohort’s life, our music product works with a coefficient of 0.21€ (AMPU). To earn one million, we need to attract 396,825 users every month, totaling one million in a year. As we recall, the cost of a team is enormous. This means that music services cannot afford paid channels. To be profitable, we can rely solely on organic traffic. If we have large SEO traffic, then maybe we can survive.
In 10 minutes, you realized
Let’s examine the economics of paid and free channels separately for B2C businesses.
Note: If your business has multiple segments (B2B, B2C, etc.), you must calculate each segment separately.
Sheet 2 in Excel. As a premium E-bike shop owner, you want to increase your profits. However, when you increase your ad budget, your profits actually decrease. You have analyzed the data using Cost per Click (CPC) on Google Analytics, but the calculations still do not align. As a result, you have installed a new analytic system in conjunction with a new CRM to obtain more reliable figures. You have obtained the following numbers for a 1-month cohort:
Given:
Let’s review some steps from the previous use case.
Step 1: First, we need to identify the business model
E-Shop is a model with costs per each sale:
AMPPU = (AvPrice — COGS) × AvPaymentCount
To calculate the AMPPU, we need to find the missing values. Let’s start with the average payment count within one month: Average Payment Count (Overall) = Buyers/Payments = 3300/4000 = 1.21
Average Payment Count (Organic) = 2700/3272 = 1.21
Average Payment Count (Paid) = 600/728 = 1.21
Note: This unit of economics is less frequent than taxi or food delivery services. Scenario frequency is one of the most important properties of a business.
We are provided with the sum of revenue and gross profits, which allows us to calculate the total Cost of Goods Sold (COGS):
Total COGS = Revenue — Gross Profit = €20,000,000 — €6,000,000 = €14,000,000
We also have the total number of payments, which enables us to calculate the average COGS per each sold item:
COGS per each sale = €14,000,000 / 4000 = €3,500
Now, we need to find the average price to calculate the AMPPU:
Average Price (all) = Revenue / Payments = €20,000,000 / 4000 = €5000
We can use the calculated COGS per each sale and average price for both organic and paid channels under the assumption that people are the same in both channels, and we don’t have accurate purchase data from each channel.
Therefore, the AMPPU (30) can be calculated as follows:
AMPPU(30) = (€5000 — €3500) * 1.21 = €1818
Step two: finding C1
In this step, we determine the conversion rate for the first purchase, denoted as C1.
C1 = Buyers / Visitors
We observe that our paid payment channel is performing less effectively compared to the organic one. Consequently, we need to review the advertisement settings to enhance its performance.
In the third step, we count AMPU
With AMPU, you see the effectiveness of each marketing channel.
AMPU30 = AMPPU30*C1
Fourth unit-economics step. Compare CPUser vs AMPU
Find out how much you spent on user acquisition
CPUser = Acq. Costs/Users
Some Conclusions from this Case:
With the figures at hand, we can now conduct a simple analysis.
3. Alternatively, we can consider abandoning paid channels and focus on developing the existing organic channel, where the economics is already positive.
But we have studied only one segment (b2c). It is quite possible that for users attracted from another channel or allocated according to other characteristics, the unit economics will not converge. Therefore, segmentation is an important task in calculating the unit economics.
Even if the unit economics does not converge on the sixth month for other segments, this may likely happen later, for example, in the twelfth month. It is possible that by this point, users from the cohort will bring in enough profit to pay off their attraction.
The only question is whether it suits your business. Are you ready to wait for payback, for example, within twelve months? But this is another question, the answer to which is not reduced solely to the calculation of the unit economics.
When you gather enough data for each user group, you can create tables to understand your clients’ behavior better.
Eventually, you will have data for each user group, product version, and time period. If you want to focus on unit economics and already possess a substantial amount of data, your unit-economics Dashboard (or Excel sheet) with Google User Acquisition (UA) raw data will resemble the example provided below. This dashboard will contain various segments, metrics, and slices, and you will require analytics to interpret the data and identify behavioral patterns and insights.
You can obtain such a template with explanations on my TG Channel, which includes anonymized real data from a large billion-dollar company with raw data implementation and cohort analysis.
You can replicate and adapt this unit-economics model and apply it to your own business based on the knowledge gained in this article.
If only a few weeks have passed since the acquisition of a cohort of users, then with a high probability, these users will bring revenue and profit for a long time to come. Therefore when calculating the unit economics, we are not interested in the current ratio of profit from the cohort to the cost of attracting it but what the ratio will be after a while.
To solve the problem of predicting future gross profit, you can analyze the dynamics of the indicator for past cohorts. On the basis of it, predict what will happen to the studied users.
Usually, you want to calculate the unit economics and make a big decision well before you get the data to calculate the actual profit from the cohort for the fifth or twelfth month.
Therefore, learning how to predict gross profit based on one or two weeks of actual data is important. In essence, we are talking about LTV forecasting here. This is not the easiest task, but solvable.
The answer to this question depends on the product and the problem being solved.
The answer to this question depends on the product and the problem being solved.
A unit can be either a new user, a paid user, or a subscriber to a trial.
A unit is usually chosen by new users. In this case, the unit can be both a user who has converted into a paying user and a user who has subscribed to a trial. However, there are some established conventions. In mobile games and applications, a unit is a new user. In SaaS, the paying customer. In e-commerce, the user who made the purchase.
Consider the problem of incrementality and features of the user attribution model to a particular source.
One of the difficulties that arise when calculating the unit economics is the task of correct user attribution of acquisition channels and the related incremental problem.
The attribution problem most often occurs with complex products with a long sales cycle, products where the acquisition funnel includes multiple devices or browsers, and products where the marketing mix includes many different channels. The problem also occurs in a number of channels like TV ads, YouTube, podcasts, blogging, and others.
Incorrect attribution leads to the fact that attracted users from paid channels are recorded in organic ones. This understates the profit you see as a result of an ad campaign and may underestimate the channel. That is, you will either refuse a marketing channel that works for you or you will not use it to the fullest.
Another common problem is the issue of channel incrementality: if we didn’t buy ads in this channel, then what part of the users would come through other mechanisms anyway. If you do not study the incrementality of channels, then there is a high probability of overestimating the channel, which will lead to a loss of money when scaling it.
In addition, channels can influence each other. Channels can bring back old users instead of attracting new ones.
All this is important to consider when calculating the unit economics for various channels.
Breaking down high-level metrics into components is a beneficial exercise. However, it is even more effective when performed in the context of a specific task rather than in isolation, attempting to fit all business models into a standard mold.
Remember, this framework alone will not create or develop a business for you. Instead, it serves as a tool to closely examine your business metrics, enabling you to identify both challenges and opportunities.
If you found the article helpful, please let me know, and feel free to ask any questions via Telegram.