"To BNPL or not to BNPL, that is the question"
Updated: Nov 3
BNPL is the acronym for "Buy Now, Pay Later". It is an up and rising consumer financing scheme with several variants. The simplest form of BNPL splits the price of a good into several payments - hence the "Pay Later" tag line. In recent years, and with interest rates hovering over 0%, alternate consumer credit schemes such 3as BNPL became a prominent choice for vendors to spur consumer enthusiasm.
<Disclaimer: the analyses in this articles are presented for information purposes only>
This definitely gained steam. BNPL Transaction volume in the US skyrocketed from $2bn in 2019, through $8.3bn ion 2020, to $24.2bn in 2021, and the number of transactions rose accordingly.
Just over a year ago, Tomio Greon published in Protocol an article titled "Invest now, win later: Inside the ‘buy now, pay later’ gold rush".
Not all that glitters is gold
The challenge leading to this analysis is the question whether an operator should or should not enter the space of BNPL lending.
We analyze a range of outcomes as multivariate function of volatile variables, notably Cost of Capital and the Rate of Default.
Critically, I show that not all possible outcomes end in positive ROI for BNPL lenders.
I use a static sensitivity analysis, then I employ a probabilistic model to help management evaluate the inevitable gamble they are taking.
All data presented in this post are for information only. The analysis is simplified and abstracted, and should be regarded as instructional only.
Merchant - Revenue and Cost Drivers
Merchants, obviously, want consumers to spend more, frequently, and repeatedly. This is how they make money, recoup their fixed and variable costs, and generate profit. To this end, they want to reduce friction.
Consumers make purchase decisions as they intuitively manage their finances, for the better or worse. Research has shown that the subjective perception of one's available resources has a measurable impact on their willingness to buy. An example is a research by Michaela Pagel, showing that "individuals, across the income spectrum, spend more on discretionary items... on days they get paid". Another work by Caixa Bank also shows the link between payday and increased consumer spending.
The other side of the coin to this notion is the impact of a particular purchase on the individual's finances. If a particular purchase strains their current finances, they are bound to postpone, or avoid it.
This phenomenon is all too visible in e-Commerce context: Cart abandonment rates are 80% - 90%, showing that sometimes buyer remorse precedes - and averts - the purchase.
To alleviate this pain, and reduce friction to purchasing decisions, merchants have been offering credit in all forms, splitting payments included, as it postpones part of the cost to following paydays and replenishing balances.
Klarna, a major BNPL provider, cites the following results for merchants offering this payent scheme:
41% increase in average order value (AOV)
30% increase in conversion
25% increase in return on ad spend (ROAS)
These translate into up to 60% increase in merchant revenue.
The costs merchant see are usually a combination fees:
Processing: paid whether the transaction involves credit or not
Credit related: Merchants pay lenders whenever BNPL credit is extended
Market data shows that BNPL processing fees amount to 4-8% of the transaction cost.
A simple calculation shows that rise in merchant revenue is well worth the cost:
Assume 35% operating margins; 6% BNPL fee
Cost of Revenue
In this example, merchants are better off by 11.4. Interestingly, the higher margins are commanded, the higher is the benefit.
Crucially, the BNPL context allows merchants to pass credit risks to the lender:
Should a payment delay or default, it is the lender's burden to collect. The merchant is entitled to its share as soon as the transaction is executed.
Lenders - Revenue and Cost
Using the example above, the revenue part is simple: For every sales of 160, BNPL providers commands 9.6 - 6%.
The credit extended, however, does not come without cost. Let us ignore the operational and transaction costs for a moment, and focus on two factors:
Cost of Capital: Simplifying again, let this be the interest rate at which the BNPL borrows.
Rate of default: (Simplified, the percentage of loans that consumers - creditors - will not repay.
A base-case scenario models revenue, costs, and BNPL lender margins as follows:
We assume the following:
Borrowing interest: 2% annually
We account for 4 payments, the first upfront, 3 bi-weekly consecutive payments - six weeks in total.
Default rate: 5% (excluding late payments for simplification)
Default cost is not only lost revenue, but also loss of the outstanding balance, assuming 50%.
Cost of interest
Cost of default
Lender 's margins
Sensitivity analysis I: Static
In a static world, where all assumptions hold for eternity, we're golden: At scale, a solid 3.3% profit margin would easily recoup operational costs, generate share holder profit.
Alas, since we live in a less favorable universe, we must account for assumption volatility and brace ourselves for lesser outcomes.
To the rescue comes sensitivity analysis: The first version I demonstrate is a what-if data table, where we plug in low-to-high range of interest and default rates:
Do note the following assumptions:
Borrowing interest rate increments from 2% to 8%
Default rate increment from 5% to 11%
Default balance increases from 50% of debt to 66%
I have used Excel What-if Analysis Data-table tool to draw the potential outcomes, and conditional formatting to show positive and negative results.
As you can see, some outcomes look less promising now: Increase in borrowing rate is not that dramatic, but the combination of rising default rates, and the relative defaulted debt can become devastating.
The above analysis draws several potential outcomes for BNPL under different economic parameters, and factors in their potential impact relative to our initial base-case scenario, suggesting that profit is not guaranteed at all scenarios.
But what are the probabilities of each scenario?
Sensitivity analysis II: Probabilistic
The above data table is agnostic to the probability of each potential scenario. Managements, however, require a finer understanding of a volatile future.
A way to generate a probabilistic scenario would be a Monte-Carlo simulation that would generate a large number of possible outcomes. It is a two-step process:
1. First, we run the process a large number of times (in our example: 5,000), In each run of the simulation, the arguments of the formula generating the result are randomly generated, and fed to the model.
Note that the distribution of the outcome depends on a number of assumptions (namely distribution type) and data (Minimum, expected, and maximum limits). The resulting distribution may be skewed.
In this stage, the BNPL lender's income is 6%, and the cost drivers are randomized to generate a data table with 5,000 results.
Note that the function I used for the interest: =ABS(NORM.INV(RAND(),D15,1)) shows only positive numbers, as we do not want negative interest rates.
The function for the rate of debt default is: =NORM.INV(RAND(),D16,1)
In both cases I have kept the standard error to 1 to dictate a normal distribution. You may refine the model, and select other types of distributions.
Note that this function is volatile, so every time the data table (labeled "Trial", and "Profit") fills down, the interest and debt-default rates vary before they are fed into the model.
Note that for this run, the risk of loss is 51.06%
Then, an average is drawn for this run, and assessment of the loss is made.
2. Now we repeat the simulation another 500 times (This is an arbitrary decision for this example).
We are interested in the average of the averages.
Although the distribution of each run may vary, the distribution of the averages is normal.
The volatility is reduced (due to the large number of runs: 1,250,000 expected profit margins)
The average shows that risk of loss for this endeavor is 100%.
Conclusion and Aftermath
The first stage of this analysis was done in the beginning of the year 2022. Some economic parameters were still rosy - as reflected by central bank annual interest, and strong sentiment in the market due to COVID recovery.
At the time, the possibility of loss on BNPL was remote. Even as interest rates were to rise from their historic low, the model was pretty resilient, had it been the only parameter at stake.
But then, several things started to accumulate:
In November 2021, Klarna, one of the harbingers of BNPL, reported a huge loss, even as demand for its financing surged. This raised brows among analysts: Is this model tenable, or, the more you lend, the higher the losses that will eventually emerge from your financials?
One hypothesis is that the COVID crisis resulted in augmented debt default, taking quite some time before it emerged in its financial reports.
In February 2022, Russia invaded Ukraine, resulting in surge in inflation, driven by food and energy cost
Inflation compound by supply chain woes reminiscent from COVID and exacerbated from China's zero COVID policy
Households in major BNPL market are starting to feel the burn, and this will compound with higher base interest rates, but also in higher commercial interest rates, aimed to compensate for default risk.
In this new climate, BNPL lenders would do best to re-evaluate their premises with eye on expected debt-default rates, and keep an eye on varying market characteristics, both in the legal status of payment debt, as in the relative strength of consumer segment in each economy.