Questions to answer
- How confidently can you predict customer retention or churn based on product usage?
- Can we based on product usage, predict with 90+% confidence, cohorts of customers who will churn/retain?
Building product usage data-model
It is important to break product usage analysis into phases based on the customer and product lifecycle. To keep things simple focus on the following phases:
- Initial acquisition or trial without prior usage history
- Beyond the initial phase
The "aha" use case metric: initial acquisition phase
The aha or the Atomic Use Case (AUC) is an indicator of customer usage. This use case(s) is focused on initial experience with the product. Pick a use case that you hypothesize is a very strong predictor of customer retention or inversely churn. Examples of such use cases are listed below:
- For LinkedIn, it could be getting to X connections within certain days of registration,
- at InVision, it could be about sharing a project within the first 30 days
- at boohma, it was submitting a proposal for fulfillment within 2 weeks.
- for nutrition app, it could be logging 3-5 days for 2 weeks straight within 1 week of downloading the application.
- for operations orchestration, incident remediation/enrichment automations deployed and savings displayed within 30 days of download.
- for a digital advertising platform, it could be submitting a proposal for a buy within 14 days.
How do you pick the "aha" use case?
First and foremost, you should have a hypothesis for this use case based on user/product research you have done. For example:
during my time managing the Orchestration product at HP, our use case was: an incident remediation/enrichment automation deployed and savings displayed within 30 days. Since this was an on-prem solution, it included all steps:
- flow development
- flow deployment
- flow execution, and
For boohma, an outdoor advertising platform, the use case was campaign fulfillment within 2 weeks. This included:
- audience and campaign goal analysis
- media availability
- fulfillmentWord of caution here, the level of confidence associated with the use case will vary depending on the product's life cycle and prior knowledge. This is why treating this as a hypothesis."We predict that if customers successfully complete the Atomic Use Case (AUC), our churn and LTV models will be accurate by 90%"What do we measure?Goal metrics * -- active users who *successfully perform specific use cases x times in the last alpha # of days* [time]*.Signal metrics Think of signals as preceding steps that users must take to get to use case completion. When it comes to retention: executing of these tasks will not be a strong predictor of retention but are important intermediate steps and predictors of use case completion.
Staying with the two use cases, examples of signals can be:
- with Operations Orchestration: it was workflow development or modification of pre-supplied flow templates.
- with Boohma, it was the geospatial analysis of media opportunities within a market.Your instrumentation and measurement system needs to support the collection of both goal and signal metrics.
Beyond the initial use case
This phase assumes that the customer has moved on from the initial acquisition stage. The goal at this stage is to classify users as:
- Heavy users
The definition of what each stage means is going to be specific for the product. It is recommended to start with the mathematical model of the definition. For example, it could be:
Active users who visited the product *x or more times *during the last *ɑlpha days *(frequency over time).
Use cases completed over time: active users who successfully complete specific use cases x or more times during the last alpha days
This metric is very similar to the "aha" or the Atomic Use Case but now the use cases are expanded to reflect broader or deeper product usage.
Why should you care?
Product usage analytics combined with other predictors such as demographics and behaviors in B2C, business verticals, size of the company in B2B, customer satisfaction data can be used in fine-grained segmentation at each stage of the customer buyer journey:
Trial to purchase
- which of my trial customers are most likely to convert to a paying customer?
- What would be the expected Life Time Value for these cohorts?
- What product usage, behavioral, demographic attributes best describe the users that churn?
Initial use to adoption
- which of my customers are predicted to be loyal promoters?
- Which of the customers is going to automatically renew? And which of these are at risk?
- Which of the customers are great candidates for up-sell and cross-sell?
Last but not least: Smarter acquisition strategies
- the analysis can be used to guide customer acquisition strategies: what is the customer profile where the product is a better fit and results in low churn.
- Sell the sticky feature or module that results in better retention
- Sell to the most valuable customer segment maximizing Customer Life Time Value (LTV)
- Analyze the effectiveness of customer acquisition channels through the lens of churn, retention and Life Time Value. After all, it is more expensive to acquire a new customer than to sell into an existing one.
How can we help?
- Our team can help in the analysis of your product usage metrics and customer interactions to predict loyalty, segments most likely to renew or buy additional products.
- Develop an actionable roadmap for your churn and customer loyalty models through data analysis and cost-benefit analysis