Risk-based collections and the 3Ts

Collections management has often been treated as just another cost centre; even lenders that operate sophisticated risk analysis in other areas of their organisations can run out of organisational enthusiasm by the time they look at their collections operations. Especially when the economy is going well and all eyes have turned to growth. As such, good times can often lead to collections stagnation, which can be problematic when economic conditions turn as quickly as they can, say, in the wake of a global pandemic.

That’s why, as in all areas of credit risk strategy, we should always be looking for ways to use data and analytics to create risk-based strategies in collections. In episode seven of How to Lend Money to Strangers I speak to Terry Franklin, who’s built the second half of his career on risk-based collections and I won’t try to match his expertise here, instead you can treat this article as a quick introduction to the thinking that should underpin your first efforts

A risk-based collections strategy starts with the core belief that the decision to take any action should be informed by a mini cost-benefit trade-off; and that since not all customers will respond equally to any action and not all responses are equally valuable, a one-size-fits-all approach to debt management will always generate cases where too much is invested in collecting a bad debt alongside cases where a more effective approach would have been worth the extra spend.

Building a collections-specific scorecard includes some unique challenges, often deriving from the fact people in collections tend towards similarly bad scores in traditional account management scorecards thanks to the missed payments that got them there and the fact that ‘who will pay’ can become a more interesting question than ‘who will not’. But I’m going to cheat a little here and skip over all of that for now, so that I can focus on translating the risk into a strategy. I will feature propensity to pay scorecards in a future episode of How to Lend Money to Strangers, but for now, let’s just assume we can measure each customers’ risk.

With that in hand, a series of specific strategies can be created to optimise the cost incurred in attempting to collect funds against the amount likely to be collected from those efforts.

My direct experience in collections is limited, and over a decade old, so trust that there are more sophisticated approaches available, but I do believe that the 3T approach I outline below is still relevant today, at least when it comes to the high-level conceptualisation of a risk-based collections approach. And that is to think about your strategy in terms of the 3Ts of Treatment, Timing, and Tone—where treatment refers to the channel/ tool you use to make a contact, timing refers to when in the cycle you attempt to make that contact, and tone refers to the relative severity or conciliatory tone of the message’s content.

Treatment

The first step in designing a collections strategy is to decide on which treatment we want to deploy: do we want to contact a delinquent customer by mail, email, text messages, phone call, etc? To make this choice, we want to weigh up relative costs and benefits of each option, generally with the following trade-off relationship in mind:

Cost of the Treatment ÷ Average Success Rate < Average Value of Collection

So the cost of the treatment, adjusted for how often it works, must be less than the increased amount we expect to collect by employing that treatment. If making a phone call costs $5 and leads to a collection every second time, the adjusted cost of a phone call is $10. In this scenario, it would only make sense to use a phone call if we expected to recover at least $10 more in a phone call than we do in the next best strategy, which might be a much cheaper but also much less effective text message.

Many factors will influence this relationship, and in reality, a well-designed test-and-learn environment is needed to continually optimise the calculation, but the simplest factor is the total balance outstanding: if someone owes you $100,000 the increase in the effectiveness of any treatment is likely to outweigh the increase in the cost of that treatment; whereas if someone owes you $100 the extra few dollars collected from a phone call may not compensate the dramatic cost increase over a text message.

Let’s think about this as a simple two-by-two matrix. On the vertical axis, we have customers segmented by risk. Many low-risk customers will cure with little or even no intervention from us, so we can tolerate some level of ineffectiveness in our approach. High-risk customers, on the other hand, are both much more likely to roll onwards towards default and much more likely to have other creditors knocking on their doors for their own payments owed, in these cases, we want to prioritise effectiveness. But not if the cost at which that effectiveness comes outweighs the amount we expect to recover.

That’s where the horizontal axis comes in, here simplified by the balance due. If a customer owes us a lot of money, we are willing to pay more to collect on that debt than we are if they owe us a paltry amount.

If we combine that, then, we’d say that for a high-risk customer carrying a high balance (gold level priority) we’d look to deploy the most effective treatment we had available, costs be damned. But on the opposite quadrant where our low-risk customers with small balances fall (bronze priority), we’d focus on finding the cheapest possible solution — which might even be us nothing, and hoping they self-cure.

Timing

Having chosen how to contact a customer, the next step is to decide when to contact them. There are three timings to consider: the optimal time in the cycle to contact the customer, the optimal time in the month to contact the customer and the optimal amount of time should you allow the customer to act before re-contacting them.

You could argue that the time of day should also be a consideration here, too, but I think it is more of an issue of tone — contacting someone after hours or on week-ends communicates a serious tone more than it does anything else though of course in some cases it may be a decision based purely on availability — and that’s why I have included in it the third T.

Most important when building a collections strategy is the decision of when in the cycle to contact a customer. An aggressive strategy might contact a customer the day after a missed payment, while a more lenient one may allow a few days respite.

Two competing forces act in the period between the missed payment and the first contact. A longer period increases the likelihood of a customer making an unprompted payment and thus saving the bank the cost of an unnecessary contact. It also increases the likelihood, however, that a customer of limited means will settle other debt and thus be unable to meet their obligations to the bank. Therefore, the timing of contacts will tend to shift outwards as the perceived risk of a customer group decreases.

The optimal time of the month to contact a customer may initially seem to be closely related to the optimal time in the cycle to contact a customer. And they can overlap, but the underlying reasoning behind each is different. The optimal time in the month to contact a customer is based on customer-specific events like the day on which they’re paid. In a big lender with daily or near-daily cycles, a customer’s payment may fall due on the 5th of a particular month or the 10th or the 15th. If they are paid on the 25th of every month, though, the day in which they are most able to meet their obligations will be then, and independent of the cycle date.

This sort of strike date optimisation is more relevant in some markets than others, and generally more associated with higher risk populations, so I’ll leave it out of the general discussions going forward. However, if this is part of your collections or pre-collections toolset you can use the same approach to design it into a risk-based collections strategy.

The final aspect of timing to consider relates to the amount of time a customer is given from when they are first contacted to when they are expected to act. This can include decisions about how long we wait after sending a reminder text message to making a phone call, as well as decisions about the allowable duration of promise-to-pay agreements.  

When choosing the optimal duration of a promise-to-pay it is important to consider the due date of the payment not just as an opportunity to receive a payment but also as an opportunity to gather risk-indicative information. Each time a customer misses a promised payment, the lender learns more about that customer and is presented with an opportunity to adjust their strategies. So, where a promise-to-pay falls due at the same date as the next instalment falls due, an opportunity for extra information is lost. Wherever possible, therefore, it is preferable to set promise-to-pay arrangements so that they fall due between contractually pre-existing agreements. This is not always easy, because customers often only receive funds once a month, but it is worth pushing for.

On the same two-by-two matrix, the timing decision can be thought of as a trade-off between initial contact priority on the vertical axis and ongoing contact frequency on the horizontal axis. A customer who is high-risk should be contacted early in the cycle, primarily because we expect there to be more than one obligation being missed, or at risk of soon being missed, and so we want to be making our case for payment while there are still funds available.

And while we’d want to be contacting all high-risk customers as often as we could, the fact is each contact comes with a cost so we’d limit the frequency-based once again on the value of the balance due.

Tone

Once we’ve chosen a treatment with which to contact a customer and the timing of that contact, we need to consider the tone of the message itself. This is something far harder to put numbers against, but we can think about it as a continuum that starts as a polite reminder helping a customer to avoid unnecessary fees and inconvenience that becomes a request that a demand that becomes a threat of legal action.

This change in tone is usually communicated through a change in language, but also by way of the name of the team making a contact, or even through changes in the packaging of the message. And the purpose of the change in tone is to emphasise the escalating severity of a growing default position, but it does come with a cost of sorts: the more aggressive the tone as perceived by the customer, the less likely they are to want to remain customers of the lender.

This is not always problematic, many customers in the later stages of collections are not welcome back, but it can spoil retrievable relationships earlier on in the process. In fact, the early stages of collections offer a unique opportunity to show customers your willingness to stand by them in difficult times, so managing the tone of a message is an important part of the broader brand strategy.

It is for this same reason that I have suggested the severity of tone move along the vertical, risk-aligned axis of our two-by-two matrix. Low-risk customers are those most likely to recover and to remain with the lender in the future, and therefore those whose relationships we least want to risk with a bad experience. High-risk customers, on the other hand, are most likely to default anyway so the downside is limited. Also, more cynically, they’re less likely to have competing offers and so even if they do recover, we can probably survive a bit of a knock in customer experience.

Then I mentioned in the timing section that the time of day at which a contact is made can also be used to communicate tone. In this regard, one would usually strive to contact a customer at a time convenient to them while the relationship was still good and the tone still light but not every customer is reachable during normal business hours. In these cases, it may be worth paying overtime rates to have a collections team making evening calls or calls on the weekend when the customer is less likely to be distracted. Admittedly, this is less of an issue in the mobile phone age than it was in the days of landlines, but conceptually I think it is still worth thinking about a horizontal axis that contrasts timing convenient to the customer from timing convenient to the collector.

 Overlaying the 3Ts, we now get a guiding principle that suggests for gold priority customers we should be prepared to invest in reliable treatments (even if those are more expensive); we should contact them early and frequently; and we should feel comfortable using a severe tone of voice at a time that is convenient to ourselves.

Whereas, for the bronze priority customers a strategy that is, first and foremost, cheap is preferable even if this means they are less reliable. Contact can be initiated late in the process, allowing the customer room to self-cure, and any contact we do make can be infrequent and with a friendly tone that protects the future relationship.

These guiding principles can then be translated into workflows. These workflows will vary tremendously, based on the size of your organisation, the regulations under which you operate, your budgets and available tools, etc.

That said, and just as an example, the gold priority approach might be translated into a workflow where on the first day after a missed payment a reminder text message is sent and a queue is created so that one of our collectors will call the customer within the week. If they contact the customer, they might ask them to make the missed payment within two weeks. The day before that promise to pay falls due a second reminder could be sent, and then if despite all of this the customer fails to honour the agreement, we might send our first official collections letter (often regulations require a certain number of specific actions be taken before an account can be written-off, and often a chain of legal letters is a part of that). In contrast, our bronze priority customers may be left to their own devices until late in the cycle, at which point we may only send out a text reminder.

A real strategy is more complicated than this. Much more complicated. There are exponentially more branches to most workflows arising based on whether or not a right party contact can be made, based on whether a promise to pay is initially agreed or not and then whether or not it is subsequently kept, based on the sophistication of collections scorecards and thus month-by-month readjustments of perceived risk, and of course based regulatory controls over the frequency of contacts and the ways in which vulnerable customers are handled. However, I think the 3T approach at least gives a framework and some structure to help make those decisions in a consider way, but if you have a better model, or if you think I may be missing out a key consideration or two, please feel free to reach out to the show.