The definitive guide to categorisation and enrichment

What is transaction categorisation and enrichment?

At its core, transaction categorisation and enrichment is the translation layer of modern banking. Without it, customers would only see the amount spent – a fairly useless piece of data on its own. Categorisation and enrichment involves adding contextual information, such as the merchant name, location and  spending group, for instance: groceries, to every income payment and spending transaction, in order to make the data useful.

Not all models operate in the same way, but transaction categorisation and enrichment by Moneyhub (also known as the Engine) uses proprietary machine learning models to clean, identify, and label raw banking data, turning it into actionable insights with profound accuracy.

What makes a categorisation and enrichment solution market-leading?

1. Accuracy
2. Coverage 
3. Taxonomy 

Accuracy

Accuracy is a measure of whether the categorisation and enrichment information applied to a transaction is correct. What percentage of the time is the transaction in the right category, with the right merchant data, geolocation data? 

Accuracy is critical to any categorisation and enrichment solution because it’s a risk liability. Miscategorising a high-cost loan repayment as leisure, for example, compromises credit risk models and affordability assessments. It leaves institutions making decisions on incorrect assumptions, and therefore 

Moneyhub offers a categorisation accuracy of 98% and a merchant identification accuracy of 99%.

Coverage

Coverage refers to the percentage of transactions that a system can successfully identify, map, and enrich with meaningful information. 

It’s an integral feature of any categorisation and enrichment solution because it’s a key performance metric, measuring how reliable the system is. Without high coverage, a large proportion of transactions are unable to be categorised and enriched, leaving useless data in the hands of banks and their customers. 

Therefore, the higher the coverage, the better the categorisation and enrichment solution. Moneyhub offers a coverage of 99%. 

Taxonomy

Taxonomy refers to the hierarchical naming system that institutions use to categorise their transactions. Generally speaking, the more categories and levels a system has, the better. 

That’s because a deeper taxonomy enables: 

• Fewer transactions uncategorised: less ‘other’ or ‘miscellaneous’, and more true categories
• More useful in-app experiences: customers have a clearer overview of their spending for budgeting and financial wellbeing purposes
• More contextual insights: analysts can accurately segment cohorts of similar behaviour thanks to the depth of insight

Moneyhub offers three levels of categorisation as standard for any transaction. But one of the ways in which the solution is market-leading is through a fourth level of granularity across lending transactions in particular. 

This is beneficial to banks and lenders because it provides the necessary detail to segment customers on an individual basis – ensuring teams can confidently work from real-time spending behaviour instead of predictive or assumptive demographic markers. 

Moreover, we offer far more categories than a standard bank would, with 65 categories at our level two layer. Without such depth, many organisations struggle to identify the niche or new retailers that make up a large portion of daily spend, causing confusion for the customer.

For example:

What does categorisation and enrichment help banks to achieve?

There are several strategic use cases, dependent on the capabilities of the exact product you choose to build or buy.

Here is how leading institutions are deploying the Moneyhub Engine:

1. Reducing the customer ‘confusion tax’ by presenting recognisable transactions 
2. Identifying fraud in real-time to maintain customer relationships
3. Driving customer savings opportunities
4. Personalising product offers at higher conversion rates
5. Feeding the data into the product roadmap for better targeting

1. Reducing the customer confusion tax with recognisable transactions

Approximately 20% of customer phone calls to a typical bank are due to unrecognised transactions. When customers only see a string of raw merchant data along with the price, they tend to fear fraud or attempt a chargeback. Both of these outcomes are negative for the bank.  

By displaying clear merchant names, logos, and maps, you allow users to recognise their spending immediately. This reduces inbound enquiry volumes and the operational overhead of managing disputed transactions. Not only are customer relationships managed more effectively, but the operational overhead associated with dispute investigation and admin is also significantly lowered.

Learn more about how to help your customers recognise their transactions.

2. Real-time fraud identification

Transparent, accurate data is the foundation for security. Categorisation and Enrichment solutions work in real-time to feed into fraud detection algorithms, empowering customer and fraud teams to protect their customers with confidence. 

The alternative? ‘Garbage in, garbage out’ – a phrase that accurately represents how annoyed customers become when their accounts are locked and payments frozen when fraud controls are unnecessarily applied, due to poor data accuracy.  

Read about identifying fraudulent transactions in more detail.

3. Driving assets under management through customer deposits

29 million UK adults are ‘current-account coasting’, leaving an estimated £520 billion sitting idle in low or no interest accounts. This inertia costs UK consumers roughly £20 billion a year in lost interest. 

The Moneyhub Categorisation and Enrichment Engine provides an opportunity for banks to a) identify those with savings capacity, and b) target these customers with appropriate messaging to generate deposit action. 

By analysing recurring income versus expenditure, banks can use the data to identify customers with excess funds (that aren’t earmarked for upcoming bills), and trigger a real-time nudge. 

This technique, combined with the use of Variable Recurring Payments (VRPs), was the foundation for Paragon Bank’s incredible results with Spring Savings. Month-on-month, the volume of daily open banking payments deposited into user accounts doubled. 

Learn more about driving deposits through savings goals.

4. Personalising product offers at higher conversion rates  

By understanding a customer’s full financial picture, including their recurring obligations and external spending, institutions can identify the next best step at the exact point of need.

Banks and lenders therefore use the Categorisation and Enrichment Engine to: 

1. Segment their customers
2. Assess cohort behaviour (including buying patterns)
3. Create targeted product offers

For example, from the data derived by the Categorisation and Enrichment Engine, a lender can identify that the customer is paying high interest on a high cost short term loan offered by another provider. With information on the payment amount, number of repayments and merchant name, the provider can accurately nudge the customer with an offer for a more affordable internal consolidation product. 

Instead of fighting for market share, cross and upselling becomes helpful. 

Learn more about how to improve cross and upselling in banking.

5. Feeding data into the product roadmap for better targeting

Many product teams feel like they’re marketing in the dark. Stuck with proxy data, such as ONS surveys or unreliable credit scores, it can be difficult to know what your customers actually need. 

The Categorisation and Enrichment Engine bypasses this issue by replacing proxy data with the real deal – accurately enriched transactions grant access to true behavioural insights. By harnessing these insights, you stop wasting budget on proxy guesses and start acting like a proactive financial coach who understands a customer’s world before they even have to explain it.

To learn more, read about how transaction analysis can inform product marketing.

Core features and capabilities

Beyond accuracy, coverage and granularity, there are some core features and capabilities that we recommend you don’t compromise on. Whether you’re building a categorisation and enrichment solution in-house or outsourcing to a trusted provider, these include: 

There are also some more optional features that may matter to you, depending on your main use case, such as: 

• HMRC tax mapping: Designed for UK tax return solutions, this feature matches transactions to designated HMRC tax categories for self-employed and SME users
• Benefits accuracy: for UK banks in particular, key use cases like this discounted phone tariff for Voxi may depend on the ability to accurately predict benefits
• Regularity detection: subscription and cashflow management tools alongside predictive analytics depend on the ability to precisely spot regular incoming and outgoing payments, helping institutions to meet Consumer Duty demands and customers to avoid overdraft fees

Build vs buy: navigating the technical challenge

When implementing a transaction enrichment solution, financial institutions must decide whether to allocate internal resources to build a proprietary system or partner with a specialist provider. While building in-house may appear to offer more control, it often results in a recognition gap that takes years and millions of pounds to close.

The true cost of in-house development

Developing an enrichment engine is not a one-off project; it is a permanent data-science commitment. Fine-tuning a legacy system to reach market-leading accuracy typically takes three to five years and can cost upwards of £6 million in dedicated developer and data scientist salaries.

As merchant data degrades, and rapidly, databases require constant updating. Thousands of new businesses are established every month, and transactions tagged with these merchants will simply fail to be fully categorised and enriched, leaving your customers confused and support teams burdened. 

Finally, most banks do not place a core focus on merchant identification, but solving operational challenges and improving customer engagement. With limited budgets, institutions should consider whether the ongoing commitment of categorisation and enrichment is really worth it, or whether their capital, both monetary and human, could be better spent on customer-facing innovation.