Customer Lifetime Value (LTV) is a core metric that measures the health and sustainability of SaaS product business models. This paper takes subscription-based SaaS products as the starting point and discusses in detail the statistical methods, practical points, and the importance of multi-dimensional analysis of LTV.
The background of the article is that assuming a subscription SAAS product, the payment model is monthly, quarterly, and yearly. The project has gone through multiple stages from the relative blank market to the encirclement and suppression of competing products.
The article is a process of thinking, which may not form a theoretical guidance for everyone, but it is still acceptable as a way to develop ideas.
Statistics on LTV of subscription-based SAAS products
First of all, customer lifetime value (LTV) is the core indicator of the health and sustainability of SAAS products, but LTV statistics are a complex project that requires deep collaboration between multiple departments.
1. LTV calculation method based on SAAS products
The core purpose of statistical LTV is to predict the net income that customers bring to the company throughout their life cycle. The more commonly used method is based on cash flow forecasting, which seems to be more closely related to the characteristics of SAAS products.
The core formula is: LTV=(average monthly recurring revenue) * gross margin % * average customer lifetime (months)
For the calculation of average monthly recurring revenue, it is more recommended to calculate LTV by customer group/level (e.g., by acquisition month, package level, and customer size), and it is more meaningful to calculate LTV by grouping.
To achieve these three challenges, product managers will only continue to appreciate
Good product managers are very scarce, and product managers who understand users, business, and data are still in demand when they go out of the Internet. On the contrary, if you only do simple communication, inefficient execution, and shallow thinking, I am afraid that you will not be able to go through the torrent of the next 3-5 years.
View details >
In terms of gross profit margin, generally speaking, gross profit margin = (revenue – variable cost of directly serving the customer) / revenue * 100%. However, for software products, there will be some variable costs, such as servers and other infrastructure, which can be shared according to user usage. What is more troublesome is the customer’s support cost, which is mostly human cost, which is difficult to accurately separate, and can only be roughly allocated according to the attributable part.
This does not include fixed costs such as R&D costs and general management costs (such as the cost of non-front-line personnel, etc.), and LTV here pays more attention to the gross profit contribution brought by customers.
The harder part to predict is the average customer lifetime, which is usually the average customer lifetime (month) ≈ the churn rate of 1/month. Monthly customer churn rate = total number of customers lost in the month / total number of customers at the beginning of the month * 100%. Of course, the churn rate should still be calculated according to the customer base as much as possible, because the churn rate of new and old customers and different packages may be very different.
2. Statistical practice of LTV
The practice of statistics mainly involves two issues, one is the sorting and integration of data, and the other is the key points of statistics.
2.1. Source and sorting and integration of data
CRM (e.g., Salesforce): Customer basic information, contract value, subscription start/end date, add-on/down/churn records.
Subscription/Billing System: Precise MRR, ARR, transaction flow, renewal status, refund history. This is the core source of revenue data.
Product analytics tools: User behavior data (activity, feature usage depth), which are important indicators for predicting churn and buy-in potential.
Financial system: cost data (server fees, payment fees, customer success/support cost sharing).
Customer support system: Support the volume/type/resolution time, customer health score.
2.2. Key points of statistics
Customer grouping: Group statistics by customer acquisition time, customer acquisition channel, initial package, customer size, industry, etc. Too much pursuit of average LTV data is not of great value.
Distinguish between gross loss and net expansion: When calculating churn rate, distinguish between customer churn and revenue churn, and clarify the calculation of NRR.
Dealing with variable contracts: For customers with high fluctuations in usage billing or contract values, MRR needs to be calculated more granularly (e.g., rolling averaging).
Cost Allocation: Establish clear, consistent, and auditable rules for distributing variable costs to customers or customer bases. Avoid over-complexity, but also ensure rationality.
1. LTV is a strategic action that requires a high degree of collaboration between multiple departments
In fact, many people have this concept, whether it is statistics on LTV or statistics on data from other dimensions, but the final implementation road is often full of twists and turns, resulting in a no-brainer. This is a work that requires a high degree of coordination from top to bottom. The key points are:
Unified data caliber: All departments agree on the definition of key concepts such as “customer”, “MRR”, “churn”, “additional purchase”, “cost sharing”, etc., if there are multiple interpretation methods for a data, the final result will inevitably be distorted.
Absolute single source of truth: The source data must be consistent to ensure that all departments see calculations based on the same set of source data.
Clear processes and responsibilities: Clarify the responsibilities and processes of each department in data entry, cleaning, calculation, reporting, and application.
Regular communication: Establish cross-departmental meetings to share conclusions and conjectures, discuss data anomalies, and jointly formulate optimization strategies.
Integrate LTV into decision-making: Statistics that simply stay at the data level are meaningless, either abandon LTV statistics or integrate and intervene in decision-making.
Leadership support: It’s very simple, the emperor is not in a hurry and the eunuchs are in a hurry to do things.
Dimension of LTV statistics
The background of the problem is to discuss whether LTV should be for a specific customer, a certain channel customer, or a customer in a certain time period. Taken together, LTV is not an either/or statistical significance, it can and should be calculated and analyzed in different dimensions, each serving different business purposes.
However, no matter what business purpose it is, the core foundation is customer grouping.
Calculating the exact LTV of a single customer directly is neither realistic nor necessary in most scenarios. Unless it is a very small number of ultra-high customer unit prices and highly customized enterprise customers.
The significance of customer grouping is that customers in the same group have great similarity characteristics, such as customer acquisition time, channel, scale, etc., which determines their retention rate, purchasing power, and revenue model are also similar. Calculating their average LTV is more instructive to the business.
In addition, based on the analysis of the historical behavior of similar groups, the results obtained are more accurate in predicting the future, which is much higher than the prediction based on the behavior of individual customers, after all, the behavior of individual customers is extremely accidental.
2. LTV calculation and significance in different dimensions
It is more recommended to calculate LTV by time period/customer group, the average LTV of all customers as a group in a specific period of time. By comparing LTV over different periods, the long-term value impact of market strategies, product iterations, and changes in the competitive environment can be evaluated.
At the same time, by continuously observing the trend of LTV, the overall health status and growth quality of the project can be judged.
The biggest role of LTV statistics according to the customer acquisition channel dimension is to optimize the allocation of marketing budget. Calculate the LTV/CAC ratio based on the customer acquisition cost of the channel. Channels with high ratios indicate high return on investment and should increase investment; Channels with a low or even less than 1 ratio need to optimize or cut their budgets.
Of course, this also helps to better understand the value of the channel. Customers attracted by different channels may have significant differences in product usage, retention rate, and additional purchase ability. LTV can tell you which channel brings in truly high-quality, long-term profitable customers, not just sign-ups or first orders.
3. The data distortion problem caused by the single attribution of LTV statistics to the channel
This is also a very common problem, when users are grouped according to channels and LTV statistics are carried out, if they encounter changes in the industry market environment or the strong impact of other external factors, it will lead to serious distortion of LTV data, resulting in misjudgment of channel value.
To put it simply, it is because of attribution bias that the channel takes the blame.
For example, a company acquires customers through channel A and competes to launch a comprehensive free subsidy policy, resulting in a sharp increase in the churn rate of channel A. If you only look at the LTV of channel A at this time, the conclusion may be that the quality of channel A has declined, but the real reason is that the churn rate of all channels has increased indiscriminately. It is very likely that this will lead to wrong strategies, such as blindly reducing the budget of channel A and ignoring the real threat: the change in business strategy of the competition.
3.1. Solution: From single-channel attribution to hybrid attribution
First, the types of churn should be distinguished. For example, if the customer intention is poorly matched, then the channel factor is greater. If it’s a churn caused by product experience, it’s better to optimize the product instead of effort on the end channel. If it is a change in the market pattern of the industry caused by the competitive business strategy, what needs to be done is to adjust the business strategy in time and decide whether to follow up or turn to counterattack.
Secondly, the contrast of the time dimension can be increased to exclude external environmental noise. For example, you can calculate the historical average churn rate of the channel and monitor the degree of deviation of the current churn rate relative to historical levels.
Generally speaking, churn rate offset = (current churn rate – historical average churn rate) / historical average churn rate, under this calculation formula, if the offset of all channels expands simultaneously, it can basically be determined that it is caused by external environmental factors. If it is only an increase in the offset of a single channel, it can be preliminarily determined that it is a channel problem.
4. Users’ cash retention and behavioral retention
Suppose the renewal cycle of a subscription product is monthly, quarterly, and yearly. For users who choose quarterly payment for the first time, they should see that the user has prepaid the service fee for 3 months, and its renewal behavior needs to be defined in layers, rather than simply equivalent to waiting for a second renewal for 3 consecutive monthly payments.
The so-called cash retention is to see whether the user pays again at the end of the current service cycle, which is mainly used to predict revenue and financial accounting. Behavioral retention depends on whether users continue to use the product until the next cycle, which is particularly important for determining product health.
4.1. Two analytical perspectives on user retention
Assuming that user A paid Q1 fees on January 1, how should they be evaluated for retention?
From a financial continuous perspective, just wait until the end of the quarter to observe whether there is a second renewal. The advantage is that it is in line with the revenue recognition logic and avoids false retention. The disadvantage is that the lag is relatively serious, and Q2 is needed to judge the retention of Q1 users and delay decision-making. For example, users are likely to have stopped using it in Q1, but it was not discovered until Q2.
From the perspective of user activity, it is necessary to monitor whether users continue to use each month (assuming that continuous use is defined as > 5 days of use per month). The advantage is that the risk of user churn can be discovered in real time, and the product value delivery can be deeply bound, but the disadvantage is that it cannot directly reflect the financial results, because even if the user is active every month, there is still the possibility of refusing to pay for the second time.
4.2. Solution idea: hierarchical evaluation of the dual-track system
First of all, two concepts should be clarified: cash retention rate and behavioral retention rate.
Cash retention is the percentage of quarterly paying users who pay for the next cycle after the service ends, and is calculated quarterly. Behavioral retention rate is the proportion of monthly active days of quarterly paying users, which is calculated on a monthly basis.
The second is the dynamic monitoring of user behavior, such as when the user is inactive for two consecutive months, the operation department must intervene in time. If the monthly activity drops by more than 50%, start a survey on user satisfaction.
Finally, at the end of the quarter, the cash renewal is verified and the behavior model is calibrated. If it is found that the behavior is retained but the cash is not renewed, it is necessary to determine whether it is an abnormal action of a competing product or a problem with the function of the product itself.
5. How to view the channel and LTV value of non-subscription products (i.e., one-time purchases).
Now we can jump out of subscription-based SAAS products and look at those industries where there is almost no repurchase (such as wedding services, real estate agencies, study abroad consulting, B2B large-scale equipment sales, etc.), and their business logic is fundamentally different from subscription-based SAAS. The core of the optimization of this type of business is to maximize the difference between the value of a single transaction and the cost of customer acquisition, and at the same time make up for the lack of repurchase through the extension of customer value (to distinguish that users have no possibility of repurchase, and the repurchase of the retail industry is not included in this scope).
5.1. Focus from LTV to Value per Transaction (STV)
For this type of industry, our focus should be from LTV to STV. Normally, STV = (customer unit price × gross profit margin) + derivative value.
For the optimization of customer unit price, the common practice is to dynamically price according to channel sources and customer attributes (only from the perspective of operational technology, excluding external interference factors such as relevant regulations), such as the same product or service, a premium of 15% for customers in high-end channels.
Of course, it can also be bundled to increase the unit price of customers by adding high-margin additional services, such as real estate agency + decoration package design.
As for derivative value, it needs to be weighted. For example, the number of new customers brought by customer referrals (referrals) × average STV × conversion rate, 1 wedding customer referral = create an additional value of ¥20,000.
5.2. The evaluation of channels should focus on the marginal contribution rate rather than the retention rate
Generally speaking, the marginal contribution rate (MCR) = (STV-CAC)/STV, when the MCR > 40%, the budget should be expanded immediately to increase the effect. When the MCR is between 20% and 40%, it is necessary to focus on optimizing the channel conversion link to increase the conversion rate. When the MCR is less than 20, it is necessary to reconsider the channel value, reduce investment or retry other positioning to see if there is a turnaround.
For example, a study abroad institution found that the MCR of Xiaohongshu channels is 35%, and the MCR of other information flow advertisements is only 12%, which must tilt the budget towards Xiaohongshu.
5.3. Operational optimization ideas based on STV
For this business model, operators can try to optimize it from two levels.
The optimization of the transaction layer should focus on the extraction of one-time value, and maximize the single-transaction value under the conditions permitted by the rules. You can try to implement it through various recommended add-ons, bundling strategies, etc. For example, after the wedding customer signs the order, the honeymoon travel package is automatically pushed.
At the level of derivative value, it is mainly customer referral and assetization of customer cases. Under the same customer acquisition cost, the average input cost decreases exponentially with each additional referral, which has a great impact on the overall project as a whole, especially for some industries and re-customized ultra-high customer unit price industries.
Finally, there is the problem of the statistical cycle of MCR, which focuses on how to strike a balance between the reliability of data and the timeliness of decision-making during the statistical period.
When the sample is too small, it will cause data distortion and seriously affect the judgment of the channel. For example, in some industries with sparse transactions, if only 2-3 orders are traded, it will cause the MCR to be excessively inflated.
Of course, there is another situation, such as the real estate industry, which leads to a decline in MCR when encountering policy fluctuations, and if there is no reasonable judgment system, it will kill potential channels by mistake. Another example is that study abroad customers may only recommend new customers 2-3 months after signing a contract, and the MCR of the month is not reflected, which may underestimate the long-tail value of the channel.
There are many problems like this, and you can think for yourself to see what solutions you have.
