New business ‘wins’ tend to attract the limelight, but there is great value in retaining customers and increasing the probability of renewal too, particularly for SaaS companies.
Optimising customer renewal rates is an under-appreciated competence that tends to be given more focus during recessionary periods. But for Permira and its Portfolio Group customer retention and renewal is a crucial lever for the portfolio companies to drive revenue growth, whatever the wider business climate.
On top of this, we are living through a period of rapid innovation in Machine Learning (ML) and Artificial Intelligence (AI), offering the prospect of radical increases in predictive power. A thoughtfully applied ML model typically captures 70-80% of previously unidentified customer churn and down-grading of packages. In addition, the latest modelling techniques are capable of much more than predicting at-risk customers.
The evolution from heuristics and statistical methods through to decision trees and regressions, and on to deep learning and explainable AIs, allows for much greater granularity of insight, including the identification of underlying risk drivers and preventative protocols.
In other words, SaaS business leaders should be targeting models that tell them not just where churn might happen, but also what is causing it.
Permira recently hosted a webinar for data, analytics and AI leaders within the Permira Funds’ portfolio, looking at customer renewal optimisation, with a focus on predictive analytics and converting that insight into pragmatic and actionable customer success motions.
As a follow up, we asked two portfolio companies to share their experiences in this area with our Advanced Analytics and AI community, and we summarise some of their key observations below.
FullStory: Engaging Customer Success Teams
Digital experience intelligence platform, FullStory, has developed a proprietary customer health algorithm focused on minimising churn, called the “Customer Love Score”.
To develop and select the right solution for its specific context, FullStory evaluated a variety of models, focusing on the twin goals of accuracy and actionability. The final two contenders were a random forest ML model and a linear points-based model. The former had greater accuracy, but insights from the latter proved significantly more actionable, without a major depletion in accuracy.
This focus on actionability is key. The customer success team was integral to the development and selection process, ensuring that its outputs feed directly into the playbook of those on the front line of customer engagement.
At a strategic level, the model has also revealed several false assumptions, such as showing that certain factors thought to be key drivers of churn were, in fact, insignificant. It also uncovers opportunities for expansion.
In relation to customer segments, FullStory has found that for high-touch accounts, the model provides the customer success team with additional context; for other customers it reveals ‘quiet risks’ and potential expansion opportunities; and for the mass-market, it helps to identify outreach priorities.
Informatica: Driving Sustained Change
Informatica, a US-based enterprise cloud data management leader has evolved its business model from license to subscription, and now to consumption-based pricing under Permira’s ownership.
Informatica has transformed its Customer Success function through the development of a proprietary AI Platform (“Cassini”). Cassini is a series of predictive models that process signals on customer behaviour and health to drive next-best-actions for the customer success team that optimize for outcomes such as renewal, upsell and cross-sell.
Since deploying its AI-led customer success approach, Informatica has seen its subscription renewal rate climb from 84% to 94%. Key to this success has been the approach of building a series of small, highly-tuned predictive models that provide not one health score, but rather a suite of real-time metrics along the customer’s lifecycle.
Informatica’s model starts before a customer is even acquired and provides a ‘deal-grade’ quality score for deals in the sales pipeline. After a customer signs a contract this allows the team to apply appropriate levels of customer success resource from the very start of a customer’s journey.
As the customer onboards and begins to use Informatica’s product, Cassini collects and processes data points on customer usage, engagement and satisfaction. This data feeds a Machine Learning based ‘next-best action’ recommender for customer success teams, which flags cross-sell and up-sell opportunities which have a high probability to close.
Finally, Informatica’s ‘customer health score’ provides an ongoing churn risk score that acts as an early warning system for customer success teams so that they can pre-empt and mitigate churn. For the renewal rep, the health score enables them to engage clients several months before the contract expiry date and optimize the renewal offer.
Informatica’s Cassini platform helps Informatica to play offense in opening up customers for expansion, and defense by deploying an early warning system to detect adoption/consumption risks. Critically, this also aligns the Customer Success and Renewal teams to a common goal.
Key Take-aways For Customer Retention Success
- To be effective, the impressive predictive power of the latest ML and AI models must be balanced with what is actually actionable by a customer success team – and this can mean sacrificing some accuracy for the transparency that results in actionable insight.
- The development and selection process of a model should begin with the end-users – the customer success team – and make them integral to its development. Involve the customer success team throughout and encourage playbook/call-to-action development in parallel.
- Insights around customer retention can be structured to inform new sales, before they happen, as in the case of Informatica.
- The data required to predict customer retention overlays neatly onto that required for further sales expansion opportunities, for new or upgraded products.
- The most valuable data a business has for predicting customer behaviour is its own internal data (e.g. customer usage, engagement and satisfaction data). Focus on mining this proprietary data first, rather than investing in third-party data which tends to deliver only marginal gains in predictive power