CreditMatch — How To Estimate Customer Payment Times

Cash is the lifeblood of any business.

It is especially precious to the early-stage, bootstrapped company.

For companies that have established product/market fit and are relying on future sales for their expansion plans, forecasting cashflow is essential to ensuring the efficient allocation of scarce resources to maximise runway.

In an ideal world, an invoice would be paid within the timeframes set out in that agreement.

In practice however, this is rarely the case.

Variance in between actual payment and agreed payment can occur as a result of life’s events.

It can also occur as a result of an individual’s personality. Brown & Taylor (2014) (amongst others) have confirmed the existence of links between the ‘Big Five’ personality traits and an individual’s attitude towards their finances.

Discussing money is not something most people are comfortable with. It can become especially uncomfortable in the context of late payments.

Given our interest in accurate forecasting, Brown & Taylor’s insights, the value we place on our customers and our belief in the explanatory power of A.I., College Connections has set out to investigate our data’s ability to predict a customer’s likely payment date.

First, we applied a certain Natural Language Processing algorithm to historic anonymized customer data.

Then, using the output from this algorithm as independent features, we ran a multivariate linear regression algorithm to explore the explanatory power of the NLP features with regards to our dependent variable, namely: days to payment (calculated as (difference between invoice issuance date and payment date)).

Having identified a strong explanatory power between the independent variables and the dependent variable, as expressed by a relatively low Root Mean Squared Error, we are now able to predict when a customer will pay their invoice with some degree of accuracy.

The value of this is profound: it helps provide a more humane and efficient approach to credit control.

Without this tool, credit control requires a “one-size-fits-all” approach that views all late payers as equal. Each late payer is contacted one by one, without necessarily any insight into why they are late paying.

However, with CreditMatch, we can gain an insight into why a customer is late paying. Are they late paying because of their personality, in which case this requires a more jovial conversation around the late invoice? Or does their personality not explain their late payment, which would require a more frank conversation.

With CreditMatch, we can prioritise credit control resources to focus on following up with customers that are not within the range of payment variance explained by their personality.

We are delighted with this novel insight that we have gained. If you would like to pilot CreditMatch within your credit control team(s), please do not hesitate to write to us at info [at]