From raw transactions to real customer understanding
Enrich turns raw transactions into clean, structured data with categories, merchants, processors, locations, regularity, and more. Our market-leading enrichment powers customer-level signals for your apps, teams and AI models to depend on – built to work with messy data from any source, at enterprise scale.
Trusted by over 400 institutions globally
Transactions are the richest source of customer knowledge, and the least used
Every payment says something about a customer: what they earn, where they spend, which bills they pay, and who else they bank with. Yet much of that data is never enriched at all. Card payments get tidied up for the app, while transfers, direct deposits, business accounts, historical records, and data from other banks stay as raw codes no one can read.
Of customers found a transaction they didn’t recognise in the last month (Bud research)
Disputed a transaction they didn’t recognise in the same period (Bud research)
From a single transaction to a complete customer view
Enrich reads each payment in the context of everything else the customer does: income, bills, transfers and products, and turns it into customer-level signals your team can use straight away. With Drive, the same signals scale across every customer in your base.
- Transaction level. Category, merchant, processor, location and tags on every transaction, in milliseconds.
- Customer level. Regularity, income and cash flow, transfers between the customer’s own accounts and products held elsewhere, across all their accounts.
- Across your base, with Drive. Thousands of characteristics, scores and segments for every customer, ready for campaigns, next best actions and analysis.
Every transaction
- Categories across three levels
- Consumer and business taxonomies
- Merchants, logos and websites
- Processors and marketplaces
- Clean descriptions
- Confidence scores
Context
- Locations and coordinates
- Online payment detection
- Merchant tags
- Your own attributes, such as transaction types
- Essential and non-essential spending
Every customer
- Regularity and next payment date
- Income and cash flow, from salaries to gig work and benefits
- Internal transfers between own accounts
- Financial products held elsewhere
- Debt and high-risk transactions
With Drive
- Thousands of characteristics
- Product usage
- Loyalty and share of wallet
- Churn risk and financial health
- Customer segments and next best actions
Deeper than a category and a logo
Five families of enrichment, designed to work together.
Categories with real granularity
A hierarchical taxonomy of more than 300 categories across three levels. It tells you not just that money went to credit and lending, but whether it was a BNPL loan, a cash advance or a payment to a debt collector.
- Three levels, one shared language for every team
- Kept current as behaviour changes, with stable mappings so your reporting doesn’t break
- Use Bud’s taxonomy, map it to your own, or add your own attributes, such as transaction types
- Dedicated categories for consumers and businesses
Merchants, marketplaces and processors
Many transactions involve more than one business. Enrich tells the merchant apart from the processor, marketplace or delivery app that took the payment, and resolves each one to a clean name, logo and website.
- More than one business per transaction: a Deliveroo order from a restaurant shows both
- Processors such as PayPal, Square and Stripe identified separately, with their own names and logos
- 30+ million entities, 100+ thousand logos, with websites and contact details
Locations down to the street address
Location is resolved on its own, independently of merchant identification, through a hierarchy from country to street address. Coverage holds up even for the local businesses no merchant database knows.
- Full street addresses and coordinates for around 30 million points of interest
- Hierarchical: when an exact address isn’t available, Enrich still returns the city, county or postcode
- Online payments detected and flagged, so they’re never mapped to the wrong location
Regularity, whatever the rhythm
Regularity detection identifies each payment’s own rhythm from the transactions, without fixed rules or set periods, and doesn’t depend on any other enrichment – so rent paid to a landlord is caught just like a Netflix subscription.
- Whatever rhythm the payments actually follow: weekly, monthly or quarterly, and unusual cycles such as every 15 days that fixed rules tend to miss
- How much each payment varies, and when the next one is due
- Missing payments highlighted, and customer corrections respected every time data is re-enriched
Advanced enrichments for the questions that matter
On top of the core, Enrich adds the signals product, risk and marketing teams ask for most: the ones that turn a list of transactions into an understanding of someone’s finances.
- Income in detail: salary, gig work, benefits and irregular deposits
- Transfers between the customer’s own accounts kept out of income and spending
- Financial products and debt: mortgages, loans and cards held elsewhere, repayments and collections
- Merchant tags such as luxury brand or fast fashion, high-risk transactions, and essential vs. non-essential spending
Not all enrichment is built the same
Enrichment can look like a commodity until you compare results transaction by transaction. Differences in accuracy, coverage and speed show up in every product built on top, which is why Bud builds, trains and runs its own models, end to end.
Proven at bank scale
Built for billions of transactions a day: more than 100,000 transactions per second at peak, under 5ms per transaction and 99.99% uptime. Real time, streaming or batch, scalable and explainable.
Accuracy you can explain
Models balance coverage and accuracy, so results are repeatable and come with confidence scores. When a model isn’t confident, it doesn’t guess, and you decide how far to push coverage for your use case.
Built for messy data
Truncated open banking descriptions, core banking codes, aggregator feeds: Enrich is trained on all of them. Bring your own attributes, map results to your taxonomy.
AI that stays under control
Specialised banking models, trained on more than 50 billion transactions and human-labelled data, do the heavy lifting. AI speeds up learning and fills the gaps. Data and models stay in Bud’s environment, so new markets and taxonomies go live fast.
Categorisation accuracy (L1)
Latency per transaction
Transactions per second at peak
Entities, ~30m points of interest
Brand logos
What Enrich does for your bank
Experiences customers recognise and trust
Clean names, logos, categories and locations turn a cryptic statement into something customers understand at a glance – the foundation for every insight, budget and nudge in your app.
Of transactions enriched more accurately than with Fruition’s previous provider
Better transaction recognition with a merchant name and logo (Bud research)
Transactions behind Bud’s models
Fewer transaction dispute calls at First Direct
Institutions using Bud
“On average, we’re seeing that 42.6% of all transactions are being enriched more accurately with Bud than with our previous enrichment provider. Our partnership with Bud helps us deliver on our promise of financial understanding and actionable insights.”
Fits into any stack, at any volume
Any data source
Your own core banking data, core and digital banking partners such as Jack Henry, financial aggregators and open banking, or historical files, all enriched the same way.
Any delivery mode
Real-time API, streaming queues, micro-batches, or bulk re-enrichment for updates and migrations. Bud can store the enriched data with each transaction, or send it straight back.
API docs ->Your attributes and taxonomy
Map Bud’s categories to your own, pass in your own attributes and taxonomies, and benefit from a quick feedback loop whenever correction is needed.
Frequently asked questions
What is transaction enrichment?
Transaction enrichment turns raw transaction data, usually short, truncated descriptions full of codes, into structured information: a category, the merchant and any processor involved, a location, whether the payment is regular, and more. It makes transaction data usable for customers, employees and AI models.
What does Bud Enrich add to each transaction?
Categories across three levels, merchants with logos and websites, processors and marketplaces, clean descriptions, locations and online payment detection, merchant tags and confidence scores. At customer level, Enrich adds regularity, income and cash flow, transfers between the customer’s own accounts, financial products held elsewhere, debt and high-risk transactions, and essential vs non-essential spending.
How accurate is Enrich?
Enrich categorises transactions with more than 98% accuracy at the top level. Models are trained and validated with human-labelled data, every result comes with a confidence score, and you choose how to balance coverage and accuracy for your use case. The best way to judge is on your own data: ask us for a benchmark.
Can we use our own categories?
You can use Bud’s taxonomy as it is, map it to your own categories, or pass in your own attributes, such as transaction types. This can be adjusted at any point in time. Additionally, any feedback is welcome and can be instantly actioned.
Does Enrich work with open banking and other low-quality data?
Enrich is trained on first-party bank data, core and digital banking platforms, aggregator feeds and open banking data, including short and truncated descriptions. Because each enrichment works independently, a missing merchant match doesn’t stop a transaction from getting a category, location or regularity.
How fast is Enrich, and how much can it handle?
Enrich takes under 5 milliseconds per transaction and is built for billions of transactions a day, with peaks of more than 100,000 transactions per second. You can call it in real time, stream transactions through a queue, or send batches and historical files.
How does Enrich use AI?
Enrich runs on Bud’s own specialised models: natural language processing, neural networks and custom embeddings, trained on more than 50 billion transactions and human-labelled data. Generative AI is used to speed up the training process. Data and models stay in Bud’s environment, under Bud’s control.
Why not just use a general-purpose LLM?
General-purpose LLMs can read a transaction description, but at bank volumes they’re slow, expensive and don’t always return the same answer twice. Enrich’s specialised models give the same result for the same transaction every time, in milliseconds, with a confidence score, and they produce the structured, enriched data that makes your own AI models and agents more accurate.
Which markets does Enrich cover?
Enrich is live in the US and the UK, with merchant data for millions of businesses in each. Bud adds new markets and taxonomies quickly; talk to us about the regions you need.
How can we evaluate Enrich?
Benchmark Enrich on your own data, in the way that suits you. Choose a self-managed benchmark or one supported by Bud’s team, start instantly or tailor it to your data feeds and markets, then compare the results with your current provider, transaction by transaction, and see the difference in accuracy and coverage.
Is our data secure?
Bud is SOC2 Type 2 and ISO27001 certified, with at least three layers of encryption, zero-trust architecture and least-privilege access. Your data stays yours.
See what your transaction data really says
Benchmark Enrich against your current provider using your own data: self-managed or with Bud’s team, instantly or tailored to your data feeds and markets.
%201.webp)