The psychology of super fund member segmentation
Super fund members aren’t all the same; but neither are they all entirely different. The more a fund can understand about its members, the more it can personalise its engagement to each of their needs and preferences. The question is how to recognise relevant differences between fund members while also accounting for the practical limits to how much about its members a fund can realistically know.
One current problem is that some very useful information about members is not being used enough by some funds. The funds that fail to sufficiently personalise their member engagement shouldn’t be surprised when their members ignore their communications. But on the flip side, there are some pieces of information that can lead funds to believe they understand their members better than they actually do. Attempts at personalisation that don’t match a member’s actual needs and preferences don’t help either.
A member’s age – a good start
If I could have only one piece of information about a member that would help me to understand their needs, it would be their age. For a start, it would tell me how a bunch of important rules apply to that member, such as whether they were eligible to draw a pension.
A member’s age also provides a guide to their likely retirement planning and broader financial requirements – by giving a clue to their remaining life expectancy, their likely remaining period of employment, and their life stage (and therefore, for example, the potential for them to have family commitments and/or a mortgage).
A member’s age is also useful for understanding some of the key psychological issues that are likely to be most relevant when engaging with each member cohort about their superannuation. For example, how can a fund make the idea of retirement salient for younger members for whom it is (understandably) likely to feel very distant? And how can it engage with elderly members when some of them might be suffering from cognitive decline?
Unfortunately, some funds don’t send different communications to different age groups, even when age is clearly a relevant consideration. This has resulted in 80-year-old members receiving fund communications about planning for retirement (when they are already retired), and to 30-year-old members being told about the eligibility rules for contributing to super in their 70s (which is far too distant to be seen as relevant for them). These funds should at least start the journey to personalisation by tailoring their communications for different age cohorts.
A member’s account balance – becoming increasingly relevant
Like a member’s age, their account balance can also be directly relevant in applying the law – both for members with small accounts (for whom their fees might be capped) and for those with large ones (who might approach limits on how much can be transferred into a retirement account). By segmenting based on account size, funds can communicate about these issues only to the members for whom they are relevant.
When combined with a member’s age, their account balance also allows a fund to get an idea of each member’s projected retirement income. Funds could use this information to compare against different measures of adequacy and to nudge those with lower projected retirement incomes to contribute more if they can, or to invest more aggressively, for example.
Of course, members with lower account balances might not be poor. They might have a second (larger) super account that their fund is unaware of, or an unusually large amount of non-super assets. However, with increasing super consolidation, with the introduction of stapling, and with super projected to make up a larger proportion of people’s retirement wealth for younger cohorts, these problem should diminish (albeit not entirely) with time.
A member’s contributions history – pretty good too
As with age and account balance, a member’s contributions history speaks directly to the application of different super-related laws that will be relevant for some members. Funds can use a member’s contribution history to gauge whether they should be talking to each member about approaching or exceeding various contributions caps, or about the possibility of accessing a government co-contribution, or of receiving a tax benefit.
There is no point in funds suggesting that their members do things that the fund should have reasonable grounds (on the basis of each member’s past contributions history) to believe the member would not benefit from (or even be eligible for).
Telling members that ‘eligibility conditions apply’ might be factually correct and might satisfy a fund’s lawyers, but leaves members wondering whether whatever the fund has suggested is even worth their trouble investigating. Sure, we might not be able to avoid these types of disclaimers, but to prevent members throwing their hands in the air in exasperation, let’s at least send our communications to those members who we believe would benefit, and explain to them why we think it’s relevant for them.
Each of the three measures discussed above – a member’s age, their account balance and their contributions history – is imperfect. While the inferences a fund can make about its members are likely to be right in most cases, it’s easy to imagine exceptions; a younger member could have terminal cancer (and therefore a short remaining life expectancy), while an older member could be starting a new family (and need to top up their insurance).
Surely there is more to know about members than just their age, account balance and contributions history? Each member has their own personality, goals, aspirations and circumstances. So can we do better?
Maybe; but to do so we need to enter the danger zone. When moving on from these easier-to-measure, relevant-for-most-people, big-ticket issues, we will need to tread carefully. There is a danger that our good-intentions will actually making things worse.
A member’s gender – men and women are different, sometimes
What if you know a member’s gender? The good news is that this should allow a fund to refine some of the estimates they’ve made already based on the member’s age, account balance and contributions history. For example, women will have longer remaining life expectancies compared with men of the same age. So on that basis, theoretically the fund should encourage a female member aged 40 to choose the same investment option as, say, an otherwise equivalent 36-year-old male.
But while this difference in life expectancy is important at a population level, at a member level this difference is arguably swamped by the uncertainty about how long each individual member will actually live. Yes, women might live a few years longer on average, but there’s a lot of variability around that average, for both men and women. Given this uncertainty, it seems unreasonable to expect a segmentation model to be so granular that you would communicate differently to 40- and 36-year-olds about their investment options (on the basis of their differing life expectancies).
But the benefits of knowing a member’s gender don’t stop there. This knowledge might also enhance a fund’s understanding of a member’s likely future income too. If women are more likely to have their income disrupted by family responsibilities, or tend to work in industries with less opportunity for income growth, then there is more risk of women with small account balances retiring with inadequate savings and financial security (compared with otherwise equivalent men). Arguably, again, there is therefore more reason to encourage these women to contribute more to super if they can, or to invest more aggressively for the long-term.
To be clear, this gender overlay is a refinement to the segmentation model based on age, contributions history and account balance, rather than a replacement of it. If you know that a member is a high income-earning 50-year-old with a large account balance, the fact that she is also a woman is likely to be less relevant to her future financial security.
But what about the fact that woman have different personalities? Or that woman are less confident managing their finances? Or that woman tend to have lower financial literacy than men?
Sure, each of these things is correct on some measures, in some contexts, to some extent. But the problem is that while these types of differences are measurable at a population level, there tends to be a large overlap between the genders. Yes, woman might be less confident overall, but there are plenty of men who lack confidence in managing their financial affairs, or who have low financial literacy. And some women, of course, have higher financial literacy than most men.
While there is more to gender differences than just the things mentioned above, when it comes to super the fact is that most people have limited knowledge, regardless of whether they are male or female. If we engage with all members using good behavioural principles, including using simple language, clear visual representations, appropriate layering and chunking of information, designing good choice architectures, and creating frictionless processes with just-in-time action-oriented financial education, then all members can benefit.
A member’s risk aversion – important, but often too unreliable
Unfortunately, while a member’s risk aversion is theoretically fundamental to their investment choice, it is practically difficult to measure. For a variety of reasons that I’ve discussed in the chapter ‘Risk tolerance, perceptions and profiling’ in my most recent book, even many of the risk-profiling questionnaires used in face-to-face financial advice conversations are inadequate. Across the broader industry, there is no shortage of ways to improve how we understand and communicate risk.
A member’s risk tolerance can be thought of like a personality trait. But even if you knew each of your members’ personalities (as you would if each member had completed a personality profiling questionnaire), the correlation between the personality attributes and those members’ behaviours would still be modest. As Kahneman et all point out in their recent book, Noise, ‘the validity of broad traits for predicting specific behaviours is quite limited; a correlation of 0.30 would be considered high.’
Decision-making research demonstrates that in the face of unreliable information we should be hesitant in moving away from what is likely to be best for most members (sometimes referred to as ‘statistical base rates’). When it comes to risk, generally speaking younger members should invest more aggressively, older members less so. Given the consequences of deviating from this maxim can be extremely dangerous (at both ends of the spectrum), only on the basis of reliable information (preferably in the context of a financial advice conversation) should we be comfortable to suggest otherwise.
A member's communication preferences – still need to give options
What if we know that some members have visited our web site, and that some have called our call centre, and that some others have clicked on a link in the most recent email we sent them? Can we use this information to communicate with our members via their preferred channel?
We sure can. However, we’d also have to recognise that people’s preferences are likely to vary at different times, for different types of information, in different circumstances. Put differently, people’s preferences and behaviour are often context-dependent (and, as psychologists would say, ‘unstable’). Rather than preferring one channel or another per se, perhaps a member likes doing some on-line research via a fund’s web site before giving the fund a call if and when needed. Despite noticing that this member rarely calls them, it would be prudent for their fund to still offer them both on-line and in-person engagement options.
But what if some members had not engaged with their fund via any channels at all? Perhaps we could assume that these members feel overwhelmed. We could try to reduce this feeling by giving them only one thing to think about at a time, and by making things simple to understand and easy to do. This sounds reasonable; but shouldn’t we do that with all members anyway?
A member's retention risk – not as obvious as it might seem
Finally, what if a fund has noticed a cluster of member activity that has historically tended to precede those members rolling into a new fund? Should the fund then identify other members who display a similar set of activity, anticipate that these members might be thinking of rolling out also, and give them a call? Perhaps these members falsely believe that they need to change funds in order to switch their investment options, for example, a false belief that a call could disabuse them of, thereby preventing their move.
Maybe; it depends.
Firstly, how reliable is the prediction that the member will roll into a new fund? Will 90% of the members who are identified as a retention risk change funds in the near future? Or will it be only 10%? While 10% is a lot higher than the chance of a randomly selected member changing funds, it would still mean that 90% of the calls the fund made were to members who weren’t actually going to change funds at all.
Secondly, how effective are the calls at preventing members who would otherwise change funds from doing so? And conversely, how often does a call to a member who wasn’t going to change funds actually lead them to do so? … perhaps by prompting them to crystalise a decision and to take action, or by making them feel uneasy or suspicious about their fund stalking their on-line activity.
You cannot simply rely on the model to tell you the answer to these questions; by its nature that model is likely to be ‘over-fit’ to match historical patterns of member behaviour. This means that to some extent it is actually less able to predict future member behaviour than it seems. The fact that a member ultimately didn't switch funds could be the result of the call the fund made to the member, or it could be the result of the model misidentifying the member as a retention risk in the first place.
To determine if the retention strategy was effective requires calling a sample of members, not calling other members who display a similar pattern of behaviour, and comparing the results. In short, you need to run a controlled study.
I haven’t seen data from funds who have measured these types of initiatives in a rigorous way. However, from my experience in other financial services domains (with financial advice clients and with mortgage clients), the uncertainties related to identifying churning clients, and the cost and limited efficacy of the retention initiative can easily result in them causing more harm than good. As counter-intuitive as it might seem, sometimes it’s better to do nothing.
Alternatively, rather than doing nothing at all, funds could choose to call only those members for whom the quality of the prediction and the efficacy of the initiative are both high. At the extreme this could involve responding to members who request information about changing funds. Analysing the text of member conversations could help to provide the richness of insight needed to boost the reliability of the predictions. And of course, by providing member engagement that is relevant for each member’s needs and preferences more generally, funds can reduce the likelihood of members wanting to change funds in the first place.
The broad conclusion from each of these measures (beyond age, account balance and contributions history) is that they probably offer some benefits, but that each presents risks. There is a risk that they lead us to believe that we know more about the member than we actually do, and that we make unwarranted assumptions based on increasingly less reliable information.
More detailed segmentation models that incorporate a greater number of factors aren’t necessarily better. In fact, when talking about predictive models more generally, research shows that simple models are often the best. As Kahneman et al state ‘complex rules will often give you only the illusion of validity and in fact harm the quality of your judgments’. It’s not just that simplicity is a virtue, it’s that simple models are often more accurate too.
Partly this is because ‘the advantages of true subtlety are quickly drowned in measurement error.’ Asking a member’s age is easy, whereas measuring their individual preferences is tricky. Explaining to a member a complex set of factors that underpin an assumptions you have made about them might be trickier still.
Recognising these limitations, what else can funds do?
Just-in-time behavioural segmentation
If member behaviour is context-dependent, perhaps the most reliable approach is to engage with them in the specific contexts in which their funds see them behaving. So, for example, if a fund sees a member failing to complete a process, that fund could ask what support they could give to help the member succeed.
It might be difficult to predict whether a specific member has a mortgage and family (and therefore might need more insurance). However, if a fund sees a member attempt but fail to complete an insurance application form then the fund can make a strong prediction that the member has unmet insurance needs. Given that they are often a conduit to satisfying a member’s needs, the design of forms and the engagement around them rarely get the attention they deserve.
Alternatively, if a fund sees one of its member attempting to switch their investment options via the fund’s web site, the fund could ask what nudges or frictions it could apply to help the member to avoid making a rash choice which could potentially cause adverse long-term consequences. The 25-year old who is trying to switch to cash, for example, could be prompted to call the fund or to seek advice.
In each of these cases, members pop in and out of micro-segments in which they can benefit from very targeted assistance. Because that assistance aligns with the Design & Distribution Obligations, it is something that funds should already be at least preparing to implement.
Self-selection
Another approach that recognises the difficultly in knowing members’ individual circumstances, preferences and needs, is to make it easier for them to self-select the types of engagement that suit them. Members can be guided through choices with decision-trees, for example – whereby their responses to simple questions help them to navigate through complexity.
But too often funds provide their members with lists of features or lists of considerations, and leave the member to discern how to assimilate those disparate pieces of information into a decision that is relevant for them. When funds formulate this type of engagement a subtle but important mindset shift is required: from giving members the information required to make a decision, to helping members actually make the decision.
The broad conclusion from all this is that some funds haven’t gone far enough with segmenting and personalising their engagement, while others have potentially gone further than their knowledge of their members warrant. By focussing on the things we are most sure about we can maximise the benefit for the majority and reduce the risk of making unwarranted assumptions. And where gaps in our understanding of members inevitably persist, by being transparent about it, by giving members options, and by helping them to exercise choices that suit their individual needs and preferences, we can allow members to add the nuance that the segmentation lacks.
If you enjoyed this article, you might also like ...