We're constantly improving Basiq Enrich so you can understand customer spending and deliver better financial solutions. Recently, we retrained our expense categorisation model with +3M training records to expand coverage and increase accuracy to +90% so you can get accurate spending insights across more Australian and New Zealand merchants.

The prep work

Being able to accurately classify and categorise payment transactions is the first step to understanding customer spending and delivering better financial solutions. But transaction data is often messy, lacking information and not standardised.

Because data isn't standardised, each institution presents a new set of challenges, with every institution describing transactions in a unique way.

As an example, a purchase from Tony Bianco could look completely different across the Big 4:

Bank Transaction Descriptor
NAB

V6092 29/05 SP * TONY BIANCO COLLINGWOO 74617630150

WESTPAC  SP * TONY BIANCO COLLINGWOOD VIC 
ING SP * TONY BIANCO - Visa Purchase - Receipt 125282In COLLINGWOOD Date 29 Dec 2019 Card 462263xxxxxx4392
CBA  SP * TONY BIANCO COLLINGWOOD VI AUS Card xx4832 Value Date: 16/10/2020 

A unique strength of Basiq’s categorisation engine is that we can standardise and aggregate transaction data from +100 institutions across Australia and New Zealand before before sending it for categorisation. Each transaction is classified based on whether it's a debit or credit:

Debit
  • bank-fee: a fee incurred by the user from their

    bank e.g. ATM withdrawal fee
  • payment: payment made to a merchant
  • cash-withdrawal: funds withdrawn via ATM facility
  • transfer: funds transferred to an account
  • loan-interest: interest charged on a loan account
Credit
  • refund: funds returned due to refund
  • interest: interest earned
  • transfer: funds received from an account
  • loan-repayment: loan repayment credited to a

    loan account

Once a payment is identified, Basiq Enrich cleans and enriches transactions to reveal the true nature of spend:

Expense categorisation with +90% coverage and accuracy

An important part of our enrichment process is the ability to enhance Payments with multiple attributes such as merchant details and expense category.

Recently, we processed +3M transactions to retrain our expense categorisation model for the below results:

  • Coverage: +90% of of Australian and New Zealand merchants recognised
  • Accuracy: +90% of merchants correctly categorised

For every transaction categorised, we provide detailed industry classification at 4 levels, so that you can pan out and zoom in as necessary. For example, if a customer shops at Tony Bianco, "Retail trade," only tells you part of the story, but to understand what your customers are really spending their money on, you may need more granular detail such as "Footwear retailing" ⬇️

   },
        "category": {
            "anzsic": {
                "division": {
                    "code": "G",
                    "title": "Retail Trade"
                },
                "subdivision": {
                    "code": "42",
                    "title": "Other Store-Based Retailing"
                },
                "group": {
                    "code": "425",
                    "title": "Clothing, Footwear and Personal Accessory Retailing"
                },
                "class": {
                    "code": "4252",
                    "title": "Footwear Retailing"

By retraining our model we have also reduced false positives so that we can make improvements to our expense categorisation more quickly and frequently to incorporate the latest merchant information and industry changes.

For more product news head to the Product News section of the blog, and stay ahead of updates by subscribing to the Basiq newsletter.

If you want to incorporate spending insights into your product, get in touch with our team to see how Basiq can help.