top of page

Afterpay & Credit Risk Concerns


All companies can produce gross returns from taking on risk, but it is whether a company can accurately price the risk inherent in gaining the return that determines the competitive advantage and long-term value of the company. From a reporting perspective this then means that the most important cost when determining the quality of a company in the business of credit is the bad debts expense. It is for this reason that Afterpay consumer credit losses are the subject of this report.

Executive Summary

The following report has been written with the aim to aid the reader better understand Afterpay Touch Group’s (APT) ability to determine the repayment risk inherent in its trade receivables. It highlight’s my concerns regarding the company’s ability to determine repayment risk and how it reports losses using non AASB measures. This has been done through an examination of APT’s reported Net Transaction Loss (NTL) expense and its components. After a thorough analysis of the underlying components of the NTL it is apparent that the NTL is a particularly inaccurate and misleading metric in determining the losses of APT’s Afterpay consumer credit product.


The Net Transaction Losses

The NTL is calculated by subtracting late-fees from gross losses, better illustrated and broken down by the following equation:

Net Transaction Losses = Gross Losses - Late Fees


  • Gross Losses = Receivables Impairment Expense

  • Late Fees = Revenue recognised upon charge to a user at certain time points where late fees become applicable and are expected to be recovered.

The Components

What is required next is a breakdown of the two components (three components when expressed as a percentage) of the NTL that were found in the above equation:

  • Late-Fees

  • Underlying Sales

  • Receivables Impairment Expense (Gross Losses)


Late-fees by their nature are paid by APT’s ‘bad’ customers. These are customers who fail to: pay APT back on time, pay for one of their instalments, pay for 2–3 instalments or fail to pay for their entire balance. Although late-fees are capped at a maximum of $68 on a single instalment plan it is still unavoidable that due to the nature of who is paying the late-fees that the ‘worse’ the customer the more late-fees they will pay.

This as a result increases the relative weighting of each of the ‘worse’ customers as a proportion of APT’s late-fee revenue. The result is that the ‘worse’ the customer is, the more late-fee revenue is earned. While the risk that late-fee revenue will not be collected is higher with the more late-fee revenue earned, the more late-fee revenue earned the smaller the NTL figure.

As late-fees are recorded when charged, and not when paid back, the reality is that the amount of late-fees collected in cash from ‘bad’ customers in subsequent reporting periods will be less than the amount recorded as revenue in the current reporting period. It is the amount recorded as revenue in the current reporting period that is used in the calculation of the NTL.

The second timing difference caused by using late-fees to calculate the NTL is caused by the fact that APT’s merchant fee revenue is earned over the weighted average time outstanding of the receivable[1], as per AASB 15. Though in the case of late-fees, they are recorded on an accrual basis only in cases where the trade receivable has not been recouped in cash.

As the cash has yet to be received it cannot yet be redeployed in the funding of further receivables. This gives rise to a continuous timing difference between modelled accrual return on capital and actual cash-based return on capital, on which APT’s return on capital ultimately depends.

The impact of ‘worse’ customers is also just as material in this second timing difference, with the worse a customer, the longer they take to make a repayment of both the original receivable and any accrued late-fees in cash that have been charged and recorded on an accrual basis.

The use of late-fees in the NTL measure is misleading due to the inherent timing difference outlined above and the significant risk of accrued late-fee revenue being overstated relative to late-fees that are eventually collected as cash. Though APT presents late-fees as reducing the risk of losses on its consumer credit product, it is apparent that late-fees increase accrued revenue on riskier users and defers bad debts to subsequent reporting periods in which underlying sales have grown significantly, both of which are not accounted for in the NTL.

The above deconstruction of the NTL is to give the reader an understanding of simply how aggressive the accounting methodology is of non-AASB measurements at APT.

Underlying Sales

The underlying sales figure represents the total gross volume of transactions processed by APT over the reporting period. Due to the rapid growth of APT the amount of underlying sales within the first month of the financial year will be significantly less than the underlying sales in the last month of the financial year.

Due to the seasonality of retail, APT’s underlying sales will have their most rapid growth in the boom retail month of December, the Christmas shopping period. This results in a higher proportion of ‘young’ trade receivables rather than ‘old’ trade receivables being reported on the 31st of December, APT’s H1 reporting end-date.

Receivables Impairment Expense

APT’s receivables impairment expense measure is based upon the trade receivables recorded on its balance sheet at reporting date. The receivables impairment expense is calculated in accordance with AASB 9. This is a forward-looking expected loss approach (ECL) and is calculated by the company as follows:

“For the Group’s Afterpay consumer receivables, the Group applies the general approach in calculating the ECLs. ECLs are based on the difference between the contractual cash flows due in accordance with the Afterpay terms and all the cash flows that the Group expects to receive. Due to the short-term nature of the Afterpay receivables, the ECLs are based on the lifetime expected credit losses.”[2]

Lifetime ECLs are not calculated on a 12-month basis but calculated based on the trade receivables recorded at reporting date, or better defined by PWC as:

“Lifetime ECLs are the expected credit losses that result from all possible default events over the expected life of the financial instrument. Expected credit losses are the weighted average credit losses with the probability of default (‘PD’) as the weight.”[3]

Based on the above disclosure by APT and the definition of Lifetime ECLs by PWC its can be seen that the company’s reported receivables impairment expense (gross losses) are based on trade receivables at reporting date (the financial instrument). Whilst being in line with AASB guidelines, and accurate regarding reported trade receivables, the impairment expense used in the NTL does not accurately consider that APT cycles its trade receivables on an average of about 6.2 times a year. This is represented in the below table of Underlying Sales divided by Trade Receivables at year end:

However, the receivables impairment expense (gross losses) relative to gross trade receivables (receivables before the impairment expense) is still not an accurate measure for APT’s bad debts.

This is because the amount of trade receivables recorded on the company’s balance sheet at reporting date have a significantly higher risk profile than the company’s average underlying sales processed during the year. The higher risk profile is caused by receivables in arrears being recorded in a higher proportion in the gross receivables amount than in underlying sales.

The trade receivables in arrears may still be repaid, however, they naturally have a higher risk profile and a longer lifetime on APT’s books, opposed to underlying sales that are paid on-time and spend less time recorded on the balance sheet.

This is evident in the below table that shows the receivables impairment expense (gross loss) as a percentage of gross receivables:

This ultimately means that we cannot simply apply the above average of 8.42% to APT’s underlying sales to calculate the company’s credit losses. This will result in an inaccurate reporting figure that overstates losses to bad debts.

The above is the main and most important point of this report. Underlying sales and receivables impairment expense are not to scale and cannot be used as an effective measure. Receivables impairment is measured on trade receivables at reporting date, it is divided by underlying sales, a product of multiple receivables turnover, to paint a rosier picture of APT’s bad debts expense.


What is the most accurate measure of APT’s credit losses?

The most accurate measure for APT’s credit losses, as found in the company’s financial statements, is the Total Allowance for Doubtful Debts as a percentage of Total Receivables (receivables after impairment charges). The reason for using this measure is the fact that the majority of the trade receivables impairment expense (gross loss) is allocated to the trade receivables in arrears (implied by the receivables breakdown screenshots below).

The receivables are broken down into categories of age, as seen in the below screenshots:

Note 9: Annual Report 2018
Note 9: Annual Report 2019

The above breakdown of aging trade receivables implies that every 30 days roughly 4% of trade receivables, after write-offs, fall into arrears. This has translated, as can be seen below, into a provision for impairments that has averaged around 5.2% of Total receivables recorded on the balance sheet.

The above provision ofTotal receivables toTotal allowance for doubtful debts is as close a risk profile as one can expect to obtain of the company’s underlying sales, given current APT reporting methods. This is due to the fact that the majority of the provision for bad debts has been applied to receivables that have fallen in arrears. Leaving only the youngest receivables in the amount forTotal receivables.


Impact on APTs Profitability

Any reduction of the NTL causes an equal increase in APT’s Net Transaction Margin (NTM. The NTM is used on a compounding rate to determine APT’s return on capital employed.

Any slight change in either the NTL or NTM has a material impact on APT’s return on capital due to the fact that the return on capital earned is equal to the NTM compounded by up to 11–17 times (theoretically) , which is the amount of time APT can theoretically redeploy its capital per year. This is depicted in the below equation:

Reconciling Modelled Losses with Competitors

Zip Co:

Zip Co presents its cost of sales in a ‘cash cost of sales’[4] figure which is presented as a percentage of average receivables for the reporting period. This figure is comprised of interest, bank fees, data costs and bad debts written off.

For its Q4 period this figure was 7.7% of quarterly average receivables, largely in-line with APT’s equivalent ‘cash cost of sales’ using the modelled bad debts expense of 5.2%.

This metric is significantly less misleading than APT’s underlying gross loss measure due to the fact that Zip Co realises its receivables over a significantly longer time period, approximately[5] 7–8 months[6]. As the receivables are realised over a duration greater than that of the quarterly reporting period no adjustments to bring the provision in-line with underlying sales should be required.


Can APT Improve Their Ability to Assess Credit Risk?

The first line, of the first paragraph, of APTs Investor Centre landing page, describes the company as “a technology-driven payments company with a mission to make purchasing feel great for a global customer base.” If the company were to improve their ability to assess the credit risk of its users it would be through research and development applied to transaction and empirical user data that the company has collected.

This is a non-traditional approach to credit assessment that other companies, such as Klarna and Affirm, are also pursuing. However, unlike APT both mentioned competitors are also developing their non-traditional methods of assessment whilst also conducting traditional assessments, which involve enquiries about a user’s financial situation and credit score.

With Klarana raising $1.2 billion and Affirm raising $1 billion to date, it is evident that the development of non-traditional credit assessment methods is a complex and expensive research and development project. This would lead one to believe that Afterpay is applying similar resources to the development of its Transaction Integrity Engine (TIE).

With research costs being expensed and development expenditure being capitalised any resources dedicated to APT’s TIE will be found in any research expenditure line on its income statement and under intangible assets on its balance sheet.

If we look at the company’s balance sheet as of the 30th of June 2019, we will see approximately $90 million worth of intangible assets. It would be assumed that for a “technology-driven payments company” that the majority of this amount would be attributed to the company’s development of its non-traditional credit assessment method, however, this assumption would be wrong.

Most of the intangible assets on APT’s balance sheet arose from the merger of Afterpay and Touchcorp. The breakdown of intangibles as of the merger, on the 30th of June 2017, is found in the below table:

As of writing the company has approximately $89 million worth of intangible assets on its balance sheet. This means that from the funding, in excess of $550 million, that the company has raised, a maximum of $30 million has been spent towards and capitalised under the development of its TIE.

The inherited technology assets from Touchcorp related to assessing ‘transaction integrity’ are primarily in regards to detecting fraud. Fraud and repayment risk are the two major risk factors inherent in the Afterpay consumer credit product. This type of technology utilises data supplied by the card issuer, device, IP address and multiple other data points to detect fraudulent activity.

Detecting fraud this way is largely dependent on scale, and with Touchcorp serving other businesses this allows the company to gather more user data than would be available just through transactions made through Afterpay.

The scale that APT will be able to achieve through this approach to fraud detection is simply a drop in the ocean compared to that being achieved by other payment gateway providers, such as Stripe[7] and Braintree[8] (owned by PayPal), who also offer fraud detection services. Both Stripe and Braintree service millions of businesses and gather similar transaction data as APT from billions of transactions.

If APT are to improve their ability to asses credit risk they will be required to both adopt traditional methods of credit assessment alongside their existing checks, whilst also allocating more resources to the development of their non-traditional assessment methods. However, any proprietary fraud detection methods developed by the company are unlikely to be as effective as those developed by the stated competitors.


In Conclusion

It is apparent that APT’s credit losses from the Afterpay consumer credit product are not presented as clearly as possible. There are significant questions that remain and that should be raised by Shareholders with Management. These questions should primarily be regarding the composition of the company’s Cost of Sales, Research Expenditures, how bad debts are written-off throughout the year and the request for more concise details on the companies Transaction Integrity Engine (TIE).

It is also apparent when APT’s credit losses are compared to competitors that the company does not posses any competitive advantage in assessing the inherent risk of individuals seeking consumer credit.


[2] Page 28, Note B, Appendix Half Yearly Report 2019,

[3] Page 3, Moving from incurred to expected credit losses for impairment of financial assets is a game changer,

[5] The word ‘approximately’ is used as no analysis has been conducted to verify Zip Co’s claims of capital recycling. All things being held equal, I would confidently say it is likely that Zip Co’s actual receivable turnover is slower than the figure reported, as was found with APT.

1 comentario

07 abr 2020

Great insights here!

Me gusta
bottom of page