Episode 16 - How To Prioritise Your Data Science Efforts

In partnership with

Did someone forward this to you? If so, click to subscribe 

Welcome to episode sixteen of the Retention Blueprint! 

We have migrated this newsletter to a new ESP; this comes with some cosmetic changes, a new email sender address and some new features, including:

  • A web version of this episode is immediately available here

  • All other editions are also on this link 

  • Ability to like or comment on newsletter posts 

  • Audio version so you can listen on the go (the voice is not me!)

  • New opportunities for sponsors 

  • A new referral scheme 

The main story of this newsletter will continue to dive into a specific subject area. This week, it is about prioritising your data science efforts.  

In addition, this newsletter will include a new AI for Retention & Marketing section every week šŸ¤–. This section will include useful links to the latest AI trends, new AI products, AI experts to follow, AI articles to read, and more.

šŸ¤– Curated Collection of AI stories, links & prompts for Retention, Subscription & Marketing Teams šŸ¤–

  • Andreessen Horowitz recently launched its top 100 Gen AI Consumer Apps by Monthly visits. Click here to learn more. 

  • The Financial Times released a report suggesting that 56% of Fortune 500 companies see AI as a risk to their business, rising to 90% amongst subscription media and entertainment businesses. Click here to read

  • Here's a great LinkedIn post from Jeremy Grandillon with 17 ChatGPT prompts to help you stand out from your competition. 

  • Ben Bites is an excellent resource for all things AI; he has a newsletter, too.

Ready to revolutionize your workday with AI?

Discover the key to unlocking unparalleled productivity with HubSpotā€™s free guide to using ChatGPT at work. Youā€™ll find practical insights, useful integrations, and 100 prompt ideas to help you unleash the power of AI for a more efficient, impactful professional life.

šŸ“° TOP STORY

How to Prioritise Your Data Science Efforts 

Jim Gray, Turing Award Winner said in 1998  ā€œData Science is the 4th paradigm of science after empirical, theoretical and computational science. Everything about science is changing because of the data delugeā€. 

Despite the ā€˜data delugeā€™ we have seen over the last 25 years, many brands continue to:

  • Lack a proper understanding of their customer base 

  • Be unsure where to focus data science efforts 

  • Struggle to use data science to drive business outcomes 

  • Employ insufficient data science resource to meet business needs

  • Have issues with commercial and data teams ā€˜speaking the same languageā€™ 

So while the ā€˜data delugeā€™ has brought benefits to brands who are able to harness it well (think Amazon, Netlfix, Spotify), most brands are still not using data science effectively.

Data science is often used to refer to predictive analysis (i.e. churn models, cross-sell models etc.), but data science has 4 dimensions:  

  1. Descriptive: What is happening/has happened

  2. Diagnostic: Why it happened 

  3. Predictive: What will happen 

  4. Prescriptive: What we should do 

There is a common belief that predictive data science provides more business value than descriptive data science.

Executives often get excited about how data can be used to predict who will churn or how to maximise revenues through cross-sell and upsell modelling. 

In reality, predictive data science is additive to descriptive data science once you have reached maturity in descriptive data science. However, a greater return on time/investment typically comes from correctly conducting descriptive data science. 

Here are some basic descriptive data science techniques you should use before you start to build a predictive model. 

Deciles  

Deciles is a method whereby the customer base is grouped into 10% buckets by revenue. This type of analysis can help to understand whether the top revenue drivers by decile are due to higher percentages in higher tiers, longer tenure, higher a la carte revenue per unit, high average order frequency, or higher average order value.

The table below illustrates this: 

Understand Customer Cohorts Movements Through Time

The type of analysis shown below can help you to understand

  • How seasonality affects the retention curve when comparing different cohorts to one another 

  • How price changes, e.g. drops or increases, can affect the retention curve and, therefore, total revenue (e.g. of course, higher prices and lower retention can still be net positive) 

This graphic illustrates this:

Cohort Decomposition 

Many subscription services offer bolt-ons, which can be purchased on a one-time basis in addition to the subscription.

Comparing different cohort bolt-on revenues can be helpful when positioning these services.

Analysing different cohorts (by interest area, time acquired, channel, or another dimension) across bolt-on purchases can reveal significant differences in revenue and drivers of a la carte product or service purchases.  

This graphic illustrates this:

Descriptive Analytics at Moments of Truth 

Basic descriptive analytics at moments of truth allow you to understand where to focus your retention efforts and trends across different cohorts.

Over time, you can build predictive data science models for moments of truth that represent the biggest revenue opportunities based on the descriptive data science work you have conducted.

How to Prioritise Data Science Efforts

Broadly across all subscription business types, data science activities should be prioritised as follows: 

  1. Basic subscription revenue calculations, including ensuring you measure churn metrics correctly, including splitting early life and in-life churn (see episode 3 of this newsletter). Plus, fundamental descriptive churn cohort analysis, including deciling, cohort movements through time and bolt-on purchase decision trees (as above).

  2. Incrementality enabling measurement of the impact of an action on behaviour (details also in episode 3).

  3. Building an accurate predictive CLV model (see episode 1 of this newsletter) 

  4. Descriptive data science to understand which actions in early life drive long-term tenure (e.g., frequency of product use, getting a specific ā€˜resultā€™, using a specific ā€˜featureā€™, etc).  

  5. Profile / interests / behavioural segmentation for marketing / content / messaging.

  6. Identifying drivers of free trial retention

  7. Price elasticity for conversion initiatives, churn save initiatives and in life upsell / cross-sell initiatives.

  8. Predicting likelihood to churn inc. holistically, based on service issues or usage drop.

  9. Cross-sell / up-sell propensity modelling.

Summary

While it can be tempting to jump straight into predictive data science, the real power lies in mastering the fundamentals of descriptive analytics first.

This is where you uncover the core insights that drive retentionā€”understanding your customer cohorts, revenue deciles, and critical moments of truth.

These foundational efforts guide your strategic decisions and help identify where to apply predictive data science.

Until next week,

Tom

P.S. What did you think of this episode?

Login or Subscribe to participate in polls.

Daily News for Curious Minds

Be the smartest person in the room by reading 1440! Dive into 1440, where 4 million Americans find their daily, fact-based news fix. We navigate through 100+ sources to deliver a comprehensive roundup from every corner of the internet ā€“ politics, global events, business, and culture, all in a quick, 5-minute newsletter. It's completely free and devoid of bias or political influence, ensuring you get the facts straight. Subscribe to 1440 today.

How you can work with me

When you are ready, get in touch to discuss consulting, courses or my fast track retention accelerator.

Reply

or to participate.