Discussing Membership Churn and how to deliver valuable Membership Insights with Data Science.
*Transcript from cloudThings recent Membership Sector Digital Conference
Today I’ll be talking about reducing Membership Churn and how to deliver valuable Membership Insights with Data Science.
Hopefully you’ll come away with an overview of how Membership Organisations (or any organisation really) can make use of technology, Data Science and a data led approach to deliver a really modern approach to managing Membership Churn whilst maintaining (and growing) engagement and how all of that should fit into the broader context of emphasizing data driven decision making in an ongoing strategy.
I’ll start by discussing the type of data you should be tracking and why that’s important before walking through some concrete examples of how data can be used to build a Churn management program.
Finally, the crucial piece of this entire topic must be organisational ‘buy-in’ and how to build a positive culture around the proactive use of data to drive decisions at a strategic as well as a tactical level.
Membership Churn is the likelihood of an individual ceasing to engage or pay dues with a Membership Organisation for a variety of different reasons which results in the Membership Organisation having to spend extra resources to attract new members.
It’s also known as Membership Churn Risk, Member Attrition Risk or Member Turnover Probability.
There are a few different ways of tracking Churn, which I’ll talk about later but the most straightforward formula to calculate Membership Churn is:
I’d like to really emphasize just why Membership Churn retention is such an important topic, especially for membership organizations.
As paid for (PPC) channels become the default standard, the cost of new member acquisition is just going to go up and up and as more and more organisations embrace modern multi-channel engagement patterns, the competition for ‘eyeballs’ on your content and adverts is going to increase as well; making proactive membership retention strategies critical to strategic thinking organisations.
This applies across all sectors, but especially organisations that are dependent on a recurring Membership model as their primary revenue stream.
The first piece of the puzzle you’ll need to solve is in understanding what's actually going on in your organisation on a day to day basis.
And really, that all comes down to data.
The more complete a picture you have the better equipped you’ll be to make impactful decisions in developing a winning strategy.
You may have heard the ‘Single Customer View’ referred to by different names… The Single User View, The Single Member View… Whatever you call it though, it’s a data object that should be at the heart of your strategy
It doesn’t matter whether it's a glorious, all singing, all dancing enterprise data warehouse with dozens of real time data feeds coming in or just a centralized set of spreadsheets that you update manually; the crucial point is that the data is consistent and correct.
Maintaining accurate (correct) data isn’t easy though.
Often, you’ll have lots of different systems, all talking to each other, making it much more difficult to maintain a single version of the ‘truth’, but consistency and accuracy have to go hand in hand as maintaining a Single Customer View is all about one central place where everything matches and gives you full Omni-channel view.
This also means you need to learn to see it as a continually evolving project instead of just a one off, ‘Big Bang’ get everything in one place and it’s done job.
That mindset will allow you to create an asset that you can use to drive your strategic goals forwards.
You can start this process with just some very basic data that’s easy to get a hold of.
Even that basic level of data is valuable information that can help you make more informed decisions if viewed correctly as it starts to hint at what things drive people’s behaviour, what drives people’s engagement or makes them to churn or not churn.
Once you have a good process established for gathering this level of data (and more importantly for using the data in your day-to-day decision making) you can start prioritising other streams of information to pull in; more aspects of your Members that you want to understand.
Acquisition Source is important one to understand and can probably be accessed fairly straightforwardly. Having access to that data will allow you to segment your churn rate by source, allowing you to better understand performance and ROI of different acquisition channels.
Outbound Engagement is always going to be a central source for any marketing team.
It doesn’t matter whether this is just email opens and clicks or if you can analyse the data right down to fine-grained engagement metrics, it's not only a useful way of digging into the performance of your outbound activities but can also reveal critical early churn signals.
Inbound Engagement can be defined as people reaching out for support or about accessing services or having questions about those services.
This is where things are likely to get a bit more difficult in terms of accessing data.
Inbound Engagement is the area where most organisations will have a multitude of disparate systems that need tying together. Unfortunately, a lot of the time, those systems won’t be set up to make this an easy task but will offer a wealth of information when done correctly.
Touchpoint Tagging is the Holy Grail of a Single Customer View.
Once you have all of this data around who people are, what they've been doing, how they've been engaging with you and the type of thing they've been engaging with, you can start to think about categorising those touchpoints in terms of what that engagement actually means for your organisation.
This could be something domain specific like the types of content your members are engaging with or more generic like requests for membership info or even some kind of referral engagement or the offering of a discount leading to additional engagement.
Being able to aggregate by categories of engagement will help immediately show the impact of your strategy and help you to review actions taken and make sure that what you're doing is both working, impactful and delivering value for yourself and your Membership base.
In my view that’s what a Single Customer View should always aspire to look like.
Once you have some of that data in place you can start thinking about what you need to measure and (more importantly) how you’ll measure it.
That takes us to something called Key Value Measures…
The way to think about all that data you have in your Single Customer View is as ‘inputs’.
It’s the data that directly refers to touchpoints and is very tangible. It’s contextual, it's relatable you likely know what it means in relation to your organisation and what actions you can/should take to influence it.
The measurements that you have of your data… those are the Outputs.
They’re the things you want to influence with your different strategies by changing things which can affect the inputs.
That’s why it’s important to have good strategy for measuring each of these.
We’re now going to take a high-level look at some of the more important metrics, how to keep track of them and how to best calculate them.
What is Subscriber Churn?
Subscriber Churn is a way of measuring how many people your organisation loses. The simplest way of measuring that is to take how many members you’ve lost (or have churned) over a certain period and divide that by the total number of people you had at the start of the period you wish to measure.
The key point about this is that it's generic as it doesn't define a specific period and it doesn't necessarily define a member, allowing you to measure this at different levels. It means you could measure this on a weekly, monthly or yearly basis.
How you might break it down is governed not only by your organisational structure but also by what you're trying to measure. Measuring over a shorter timeframe can tell you how particular campaigns are affecting churn but are subject to some volatility. Measuring over a longer timeframe will smooth out this volatility and allow you to see the longer-term trends.
Lifetime Value is a phrase that should resonate with anyone involved in the Membership Sector and this form of LTV calculation is specific to a recurring model, so is measuring LTV in terms of Membership Churn.
To calculate LTV (Lifetime Vale) you just need to take your average margin over a specific period and divide it by your subscriber churn over the same period.
As before, calculating this over different periods means you can aggregate at different levels.
You can show LTV as a monthly view, you can have it by cohort view or by a segment view allowing you to consider trends different levels.
Next up is another classic metric known as Cost of Acquisition.
This is a fundamental calculation for the Membership Sector because its value comes from applying it in a segmented way.
COA (Cost of Acquisition) can be applied to all your different channels to get an idea of how they’re performing against each other or it can be used to measure different segments to see what sort of value and cost of opportunity you're getting for different activities.
Margin Churn, sometimes known as Revenue Churn or Reoccurring Revenue Churn, is a metric that has comes out of the SaaS software model of companies but it's really applicable to anywhere that has this kind of recurring structure
To calculate it you take the initial margin (the margin that you made in the previous period) and then look at the next period. You then need to ask how much margin should we lose through members churning and how much new margin did we generate from new acquisitions?
The point about measuring this as opposed to just subscriber churn is that it allows you to get a view of where you're losing or gaining more value.
It could be the case that you’ve noticed your subscriber churn rising quickly so you send a big 50% discount on renewals to those in some segment you've identified as being at risk. As a result of doing that you've potentially reduced your subscriber churn, but at a cost of people taking that discount when they weren't actually going to churn, meaning all you’ve really done is reduce your margin going forward.
Or in other words you’ve reduced Subscriber Churn by driving up Margin Churn.
Calculating and focusing on Margin Churn explicitly alongside Subscriber Churn and these other calculations will give you a view on what activities provide better value, especially when it comes to Member Retention.
I’ve gone into a fair bit of detail there on how to segment these different measurements and the different ways you can divide them up to get a granular picture of engagement, retention and churn across your database.
Dates were the most obvious metric to measure, by week, month or year (I’ll come on to the other metrics in a minute).
The granularity that each of those different things will give you will paint a different picture with more or less volatility and more or less granularity in terms of seeing the impact of small specific time boxed events.
MoM or YoY snapshots are another great way to get a comparison across time without having to measure your charts with a fine-tooth comb.
Let’s say you’ve noticed a yearly pattern on one of these charts. Taking the time to build that month on month picture within the year gives you a view of how it’s changed over certain periods allowing you to ascribe a RAG status to any trends that may need your attention.
Rolling averages are another way of good of looking at the medium-term picture while smoothing out some of that volatility that you're likely to see with more granular pictures.
Another way you could segment Membership data is by cohorts…
Cohorts refer to one contiguous set of people.
They let you look at the behaviour of people when grouped by when they first appeared in your database or by what channel or campaign they first came through, to get a view of your channel performance or even more generically by the category of the first touch points on which you encountered them.
(This works even better if you have these touchpoints tagged and categorized as I mentioned earlier).
The third way you can segment member data is by Segments.
Segments are very distinct from Cohorts in that, with a Cohort, a person stays in the same Cohort forever.
A Segment however, is defined by what they do going at a particular time. This means people can move around between different segments based on their behaviour, making measuring the size of your Segments and how people move between them is a critical activity.
Typical types of Segments include level of engagement - based on how many things someone interacts with in each week/month etc.
Content Preference - This could be channel preference or a content category, identifying people by the sort of thing they interact with/how often they interact with it.
Segments are an area where Data Science can really help out with cool things like clustering algorithms as something like Content Preference is going to be really fuzzy measure but it's still useful to be able to aggregate those fuzzy things into concrete and actionable business intelligence.
The final thought I’d like for you to takeaway around Segmentation is that it’s most valuable when you start applying domain/sector specific knowledge to build Segments around data that's important to you rather than, perhaps, some of the generic ideas mentioned here.
In a perfect world you’d end up with a Single Customer View, which contained everything you could want to know about your Members and all the channels they've interacted with throughout their history with you.
You’d then be able to apply that data to any of the Key Value Measures mentioned above, segment, split or group it by any Segment or other dimension that might apply to your organisation and by doing that get to a point where you have an easy way to perform an in-depth analysis that will explain explicate behaviour patterns.
I’m going to take a moment now to talk about how to actually get to grips with all that data you’ve collected and analysed and what kind of actions to take from it.
I’m also going to touch on what causes Membership Churn in the first place, how you can predict it, and then what sort of actions you can take to reduce it.
The first key point is to understand how to predict Membership Churn before it ever happens.
I'm going to walk through one example of how you might do this, how you might build a Membership Churn signal predictor using an engagement score.
The above diagram shows a typical customer journey with a Membership Organisation.
You can clearly see someone joining and then engaging, but eventually also churning with engagement split out by channel and by month.
In the first month you can see their sign up, they’ve received their welcome email, they're reading loads of your content and they're even attending events.
Over the next few months engagement is fairly consistent. They may be engaging slightly less but you'd expect that because people have lives to lead.
Eventually some of that engagement starts to drop off.
They might sign up to an event which they then don’t attend, they’re reading less content per month, maybe they're not engaging with your emails as much and then by month six… BOOM… they've churned.
They’ve cancelled their subscription and you've lost them as a Member and they go back to the top of the funnel as someone that you need to think about acquiring again.
So where did it all go wrong?
Finding and calculating an engagement score can help you understand that question.
You could say that it went wrong as soon as they didn't attend an event or you could say it went wrong as soon as their engagement dipped down from where it previously sat.
The best way to define that score however is by accounting for all channels of activity and using that data to build a score.
For instance, you could say looking across all these channels, the engagement score starts off high.
They’re highly engaged with your activities but then… oops… You start to see some signals indicating they're not quite as fully engaged as they were and that pattern continues, getting worse and worse and Boom… They’ve churned.
Even from this high level analogy you can start to see if you calculate something as simple as the number of newsletters opened or events attended, minus the number newsletter sent but not opened or events not attended but signed up for you can start to see that there are key turning points in their Membership journey and because you're reducing it to an actual data problem you can then say, “OK, we've got this. Once people reach a yellow engagement score, we need to start taking action.”
You may feel that there's no need to do anything when someone's on a green engagement score and that by the time someone's on a red engagement score it's already too late to do anything.
The real (data led) solution you’re looking for is the best way of targeting each of these zones whilst still being responsive to the preferences of individuals, providing great value for all your Members and not just giving everything away for free.
To do that you’ll need to calculate the opportunity costs of different Churn Management Strategies.
Let’s say you’ve identified a Cohort with a hundred of your members in it who have low engagement.
They’re in the yellow zone, about to crossover to the red zone and past data tells us that when that happens, about 20% of those members can be expected to churn within the next three months.
So, what’s at risk here?
From the 20% figure you can calculate the potential Margin Churn, allowing you to make a calculated decision on incentives to keep that segment.
“We've got this pot of money.
It's at risk.
So let's offer a discount on renewal to that whole Segment.”
Then, looking again at past data, you can see that 30% of Members who receive that discount will convert.
However, that means if it’s sent to a hundred members, you’re most likely giving that discount to ten people who weren’t going to churn in the first place.
Being able to calculate that will give you the ‘real’ cost of the discount you’re offering.
Giving away too much of a discount may not be efficient so calculating an accurate Margin Churn figure will allow you to fine tune the discount percentage offered, meaning you can calculate an ROI in a predictable manner.
And all of this can be calculated before you offer anyone the discount in the first place, thanks to the data you’ve collected.
Summing that up in a slightly different way; you now have the different segments of engagement score with green, yellow and red RAG status, and for each of those different segments you’ve identified the likelihoods of Members churning, expressed as a percentage.
That means if a Member has a green score, the likelihood is only 10% of them will churn in the next three months. With a yellow score that might rise to 40% and when it reaches red that might rise to 60%.
Now what you need to do is define the Opportunity Costs for a campaign directly targeting each of these segments.
You need to decide where’s best to deploy your resource to generate the most value.
You might decide that it's really not worth doing anything to try and stop your green segment from churning but that it's somewhat more worthwhile going after the members who are in the red zone.
But your data also shows that by the time people have ‘turned red’ they're not likely to respond to anything you do. It may even be they’ve set up a spam filter on your emails.
That means you still need to the ‘sweet spot’ to prevent Membership Churn.
For the Members in your Green Zone, you don't really need to worry about Churn Management.
You just want to think about how to get them more engaged so you can do really cool stuff with personalisation and segmentation.
Then, for the members where you have the highest likelihood of delivering value to yourself by addressing them as Churners you can focus on re-engagement campaign, building an actual content funnel to send people through with progressively better discounts to get them to re-engage.
And finally for members who you’re almost positive will churn no matter what you do you might start thinking about targeted outreach campaigns… picking up the phone, send a personal email or maybe even send a survey to better understand why those members churned allowing you to adjust your value proposition strategy based on the data you receive from it.
The cold start problem is when a member comes into your organisations ecosystem that you need to make recommendations too, but you don't know anything about them yet.
You don't know who they are or what they’re interested in, so you'll be leaping on the smallest of signals to make them recommendations.
All Recommendation Engines essentially work on one of two principles.
They’re either content based, which works on the principle of, “you've read X, X is related to Y so let's suggest that you read Y.”
Or they’re collaborative based, working on the principal of, “you’ve read X, someone else has read X who then also read Y… Let’s show you Y”
Both approaches have their pros and cons…
A concept-based filtering approach is very useful, but it requires having a lot of metadata attached to your content
Collaborative filtering can also be a useful approach as it will surface the similarities between your pieces of content.
The difficulty with more collaborative based approaches is that you need lots of tagging of touch points. You need all of that data to be available to you in a way that joins up with your other data (Our article on The Common Data Model might be worth reading here).
So in terms of solving the Cold Start Problem, most of the time a hybrid approach between the two will be required as neither solve this problem entirely on their own.
One possible solution to the Cold Start Problem is to take a heuristic approach, setting new members with a default profile based on content that other members have engaged with before.
You could also designate some content as featured content that you know existing members already engage with.
A third, and perhaps better solution to the Cold Start Problem, is one I’ve had a lot of success with in the past and that was by engineering ‘diving off points’. Central hubs where people only need to express a small bit of interest in a type of content before they’re directed down a well-made marketing funnel of content, almost immediately giving you data on the new member to understand what they'll engage with.
That’s just a couple of examples of the sort of thing you could try, and whilst there’s a lot more, what it really comes down to is having a joined up Omni-Channel view of your Membership Base.
The best Membership Churn strategy will come from taking all of that data that you've generated, taking all of those segments metrics that you’ve defined and feeding that stuff back into your engagement channels so that when you send out a new campaign, it's just the touch of a button to decide what will be of interest to which segment.
Ultimately, the value of data comes when you can turn it into actionable wisdom.
Going up from raw data in files and tables to aggregating that data in segments and cohorts and then using it to provide context to ongoing organisational strategies.
Becoming more and more mature at this process will allow you to do more automation at each stage, using data science, but it’s crucial to get the foundations solid before you start on that journey.
If you're a Membership Organisation currently struggling with Membership Churn then cloudThing would be happy to help. You know what to do...