Coffee Analytics recently carried out their first research in Russia on how companies manage their customer LTV (lifetime value). Andrey Muratov, Managing Partner of Coffee Analytics, told how an ML-based LTV forecast can be used to improve marketing through customer lifecycle management. It allows you to attract new customers more effectively, engage loyal customers more actively, and even manage your product range. Andrey […]
6 Strategies to Use LTV Metrics for Managing Your Customer Lifecycle
Coffee Analytics recently carried out their first research in Russia on how companies manage their customer LTV (lifetime value).
Andrey Muratov, Managing Partner of Coffee Analytics, told how an ML-based LTV forecast can be used to improve marketing through customer lifecycle management. It allows you to attract new customers more effectively, engage loyal customers more actively, and even manage your product range. Andrey used the performance of companies like the Ozon e-commerce platform, the Eldorado tech store, and others to illustrate his point. Georgy Chibisov, Growth and Performance Marketing Director of Lamoda, also shared his expert knowledge on the subject.
The 6 strategies described in the article are displayed on a graph illustrating the customer lifecycle: from the reach and the first purchase stages to the retention stage.
This is how the 6 strategies for managing LTV are distributed throughout the customer lifecycle
The main business goal at this stage is to find and attract the most valuable audience with minimum effort. The cost-income ratio is conveniently measured by the total customer acquisition cost (CAC) to customer LTV. Although there is no single standard and each industry will have its own guidelines, ideally, CAC to LTV should be at least 1 to 5 (i.e., ROI = LTV/CAC — 1, equal to or greater than 400%).
When a company knows its CAC to LTV ratio, marketers can redistribute the budget among various channels: Google AdWords, Yandex.Direct, social media, referral program, etc. This allows the company to spend its marketing investments in an optimal way based on expected ROI for the entire customer lifecycle rather than on the profit gained from the first order or the first month of customer engagement.
In this example, Google is the most effective channel
Here, Google AdWords has a lower ROI to ARPU ratio compared to the referral program. So, if the company were to keep only one channel, the referral program would be the best choice. However, Google AdWords would surpass it by 28% (434% compared to 406%) in the near future, therefore, it would be more beneficial to keep Google AdWords.
First, you need to select the strata of customers that are in your Top 20% list (their LTV is 80% or more of the maximal value). Then upload this segment to your ads dashboard to create a lookalike audience. Facebook, for example, allows you to choose how similar the existing, and the lookalike audiences will be.
Now you need to estimate how effectively this campaign attracts a new audience compared to other campaigns launched. Here you might need to use LTV again. To get a broader picture, you can also compare LTV dynamics per month to see how fast your campaigns lose their potential.
Let’s suppose that marketers carry out a promotion and analyze its results. The CAC, CPC, and ROI show that the promotion was successful. Yet, it will not be obvious without taking LTV into account that launching the promotion made no sense, in fact: the campaign attracted bargain hunters with low LTV.
In 2018, we decided to switch from multichannel attribution to LTV. Today, we estimate all the efforts spent on attracting new customers as well as customer retention, and return based on their LTV within the cohorts.
This approach allowed us to restructure marketing campaigns, which now focus entirely on growing the customer base. We estimate the number of marketing investments throughout the whole customer lifecycle. It is worth noting that LTV allowed us to flexibly manage our investments into the customer base. Now the investment doesn’t stop only at the stage of attraction, it continues for every stage of the customer lifecycle.
Thanks to LTV, we were able to optimize our marketing costs on platforms, channels, and sources of customer attraction in all the regions where the company operates.
Clusterization is an approach to segmenting a customer base that involves dividing customers into homogeneous groups. A classic example is to divide customers by age, gender, etc. Yet, such division does not always give desirable advertising results.
Another standard approach is an RFM analysis when customers are segmented by recency, frequency, and monetary sum of the items bought. It works quite well. However, the results are rather rough as the approach uses pre-defined segmentation criteria that are based on transactional data only. Our experience shows that ML-based clusterization using all the customer data available is more effective when the algorithm chooses the segmentation criteria. Still, RFM can be effective, which is proven by Ozon’s experience.
Ozon’s marketers used RFM analysis to split the customer base into 27 segments, and calculated rate multipliers for each segment based on monetary (income gained) to LTV (income expected) ratio. This allowed them to define more valuable segments as the browsing priority in remarketing campaigns. As a result, the conversion rate doubled and the ROI increased by 55%.
Another downside of the RFM approach is the large number of segments with similar LTVs. ML, on the other hand, singles out fewer clusters with more distinct boundaries.
What customer segments are there? What is the number of customers in each segment? What is the organic growth of the segment?
The number of customers in each segment and its organic growth for August 2021
Besides, you don’t have to use LTV metrics only for clustering. For example, Eldorado took the probability of purchasing into consideration in order to divide its customer base into 10 segments. As a result, its ROI from contextual ads grew 2.5 times.
LTV allows a company to monitor the most valuable segments, to analyze how customers migrate from one segment to another monthly, and how their migration affects the company’s monetary outcome.
It makes sense to give higher priority to the most valuable segments. For example, inviting such customers to private sales or giving them early access to new collections. If a company interacts with valuable customers effectively, their average LTV grows, and the number of valuable customers grows as well. This will ultimately lead to the increase in total LTV of all the customers, and the company’s financial performance.
LTV allows you to see which product categories your customers prefer. In other words, which categories contribute to the LTV of each segment over time. These products are often fast-moving consumer goods, such as shirts in a clothing store.
It is important to track LTV change for each segment specifically, taking into account the exact periods where the segments grew. Here is an example. Let’s suppose that last month the company noticed a rapid fall in LTV of the customers who have children aged under 5 and 10:
This might have occurred due to a change in the product grid. In our example, we are referring to the range of products for children, which was reduced. This means that a product can be low-margin, however, you cannot stop selling it as this will result in customer churn.
The graph below displays this correlation schematically:
In this example, Category 4 is not that important when analyzing the items that the cohort has bought within 3 months. However, the situation changes over 24 months: Category 4 contributes more and more to LTV growth.
It is crucial to know which product categories impact LTV when considered over long periods. Customers will be retained by keeping important products on offer. An omnichannel loyalty program allows you to collect all the necessary data to see a broader picture.
A Machine Learning-based model that comprises all the data on customers and transactions allows for the factors affecting the LTV of a cohort or a segment to be defined. A company can use these factors to work out hypotheses of why LTV grows. Let’s analyze the decomposition tree below:
If we analyze the purchases broken down by days of the week, the tree will show that weekend sales decrease the impact of purchases on LTV compared with weekdays. This means that it makes more sense to actively attract customers on weekdays. However, such hypotheses, of course, need to be verified by A/B testing.
To conclude, you can maximize your company profit on every step of the customer lifecycle — from attraction to re-activation — by calculating your LTV and actively using it in your strategies.