Key Metrics for Subscription Models
If you are running a subscription service, recurring revenue is without a doubt the most important metric to keep track of. What are the key contributing factors?
A decomposed recurring revenue, as illustrated below, provides a holistic view over the direct leavers of its growth. It allows you to:
- Track, analyse and optimise each element.
- Identify opportunities and pressure points so you can deploy resources most effectively.
In this article, we will use a monthly subscription service as an example and peel Monthly Recurring Revenue (MRR) like an onion into detailed and directly manageable metrics. We will also discuss various data projects that help optimise at each point.
MRR = Number of Active Subscriptions X Monthly Price
The first layer is simple. MRR is the direct product of the number of active subscriptions and their monthly price.
Number of active subscriptions can be first broken down into new and existing (excluding cancelled) subscriptions, each with its own tuning factors.
New Subscriptions = # Trials Started X % Converted
This metric sits on the acquisition side and is usually managed by the marketing team for consumer products and services. Below are some subcomponents to consider:
- Marketing ROI optimisation across different channels, campaigns and creatives. For more insights, please see our previous post here.
- Funnel conversion from landing to trial sign up. Map out the conversion flow from start to finish and track drop offs at each point. The two key things are to clearly highlight your value propositions and keep the sign-up process simple.
% Converted (From Trial to Subscription)
During the trial period, make sure all users get to experience what makes your product unique so they reach the Aha! Moment, when they realise the value it provides hence convert to paid subscribers.
One way to pinpoint the Aha! Moment is to conduct historical analyses over the behavioral patterns between trialists who do vs. do not convert to subscribers during the trial period. List out all your unique features and track their usage. Key metrics to consider per feature:
- % trialists used the feature at least once. If you observe a low value for certain features, it could suggest discoverability issues, bugs or it is simply not a feature users desire.
- Time to first usage. It shows how quickly users adopt the feature. You want to minimise this value. Here are some notable solutions:
- A well-designed onboarding flow explaining key features to new users.
- Timely and behavioral-driven tooltips. For example, when users land on a feature page for the first time, highlight and explain key functions or buttons.
- Usage retention and frequency. If you observe a high correlation between a feature usage and trial conversion, drill down further to the level of usage intensity. If the concept of your product or a particular feature is new to the market, it might require multiple experiences for users to fully understand its benefits. This metric will help you find the desired level. For example, Facebook discovered in the very early days that once users have at least 7 friends, they become hooked. Once you figure out this number, you can design flows and messages to encourage users to get this point faster.
Existing Subscriptions Retention
Once users convert to full paying subscribers, all efforts should go into retaining them and prevent churn. We discussed retention metrics in detail in a previous post here. Key things here are to:
- Continue to add value by refining existing and releasing new features. Create dedicated feature usage and retention dashboards to help guide your effort.
- Observe and encourage user behaviours that lead to higher retention and prevent those that lead to churn. Update your correlation analysis periodically for new insights and build churn propensity models for early detection and prevention.
- Automate insights extracted from data collected by your customer support team or other feedback channels (e.g. product reviews), especially from complaints. Natural Language Processing (NLP) models can help you quickly detect changes in sentiment and new trending keywords and topics.
Most services have a standard and fixed pricing structure. However, if you have a degree of flexibility to adjust your pricing dynamically (e.g. by market segments and / or time-based), explore different data strategies to optimise your results. If you have a good volume of traffic, conduct A/B tests for optimum price / conversion ratio. You can also use this approach during the pre-launch of a new country to determine the best fixed-price. Make sure your test traffic is significant in size and representative of your user base.
Multi-armed bandit (MAB) algorithm is another solution to maximise rewards systematically. Pre-define a list of pricing options and let the model find you the best outcome. We introduced it in a previous post here.
We decomposed recurring revenue to a list of components to help you understand the drivers of its growth and how data can help. We would love to hear your thoughts and suggestions, please get in touch. We are always happy for a chat!