How do we categorise bank transactions?

Our ML-based transaction categorisation is the heart of the re:cap platform.

It keeps your bank transactions clean and structured with almost no manual work. And whenever you do make a correction, the model becomes smarter over time.

1. What does automatic transaction categorisation do?

As soon as you connect your bank accounts to re:cap, our system imports your transactions and assigns them to meaningful categories, for example:

• Salary
• Rent
• Revenue
• Travel
• Marketing

Behind this is a machine learning model that detects patterns in your payment data, such as the payment reference, counterparty, amount, and recurring payments.

Your benefits

• You do not have to set up rules.
• You do not have to assign every transaction manually.
• You get a structured base for analysis, forecasting, and pre-accounting from day one.

2. Why bank transactions are the "heart" of the platform

Bank transaction categorisation is the foundation for almost everything in re:cap:

• Cash flow analysis: Where is your money going? What are your biggest cost drivers?
• Forecasting & runway: How do revenues and expenses develop over time?
• Open items & pre-accounting: Which invoices are already paid and which are still open?
• Investor and management reporting: Clean categories make your dashboards clear and comparable.

In short: For a cash-based view on your company, accurate transaction categorisation is key. Doing this by hand is unmanageable. That is why the ML model for bank transactions is the core of the entire platform.

3. Why our approach is better than competitors’

Many competitors rely on rule-based systems. That means:

• You must create rules yourself ("If payment reference contains X, then category Y").
• You have to maintain and update these rules when things change (new vendors, new tools, different payment texts).
• Over time you end up with a large and hard-to-maintain set of rules.

re:cap works differently:

• We use machine learning that learns from real transactions.
• You do not define rules. You only correct when needed. The model adapts and learns.
• The model recognises similar transactions and draws its own conclusions.

Example: You change a wrongly assigned transaction from "Other expenses" to "Marketing". Our model learns from this and applies the new pattern to similar transactions. For example, all future payments to the same vendor.

The result:

• Less manual work, because the model keeps learning.
• Fewer misclassifications over time.
• More consistency in your financial data.

Our customers: "You have the best automatic transaction categorization in the market."

4. Step by step: How it works in the app

Step 0: Connect bank accounts via open banking

Before the model can categorise your transactions, you need to connect your bank accounts:

1. Click on Data in the navigation menu. Then click on Add account to connect your bank accounts or payment providers.

2. Select your region.

3. Select your bank and connect it via the open banking interface.

4. re:cap automatically syncs your transactions.

Step 1: Default categories: Zero work

Once your bank is connected, the following happens automatically:

• New transactions are imported.
• Our ML model assigns each transaction to a default category.
• In your transaction overview, you immediately see how your operating cash in- and outflows are structured.

You do not need to configure anything.

The default categories are “zero work” and are set up to work well out of the box.

Step 3: Correct a category or add a new one

Our ML-model reaches 98.8% accuracy when categorising transactions. When something does not look right, you can correct it.

To change a category

1. Go to Analysis in the navigation menu and then click on Cash positioning.

2. Click on Transactions.

3. Click on the transaction you want to change.

4. In the Category field, select a different category.

5. Click on Save changes, and they will be applied to all reports in re:cap.

To create a new category

1. Click on Category in the transaction and choose Add category.

2. Select the top-level category and then give the category a new name.

3. Click on Confirm.

4. The just created category is now shown as "new".

Step 4: Train the model: Assign a few examples and you are done

Every correction helps the model learn. In practice it is often enough to reassign a few representative transactions:

• You change a few typical transactions of one type (for example all payments to one SaaS vendor).
• The model recognises the pattern (counterparty, reference, amount, frequency).
• Future transactions of this type are then categorised correctly.

Important:

• You do not need to start a separate "training" process.
• There is no complex setup.
• Learning happens continuously in the background.

5. Auto-matching: Invoices <> transactions

Using the same technology, re:cap also supports automatic matching of invoices and bank transactions.

Why this matters

Forecasting open items:

• We detect which invoices have already been paid and which are still open. This helps you plan future cash inflows and outflows more accurately.

Pre-accounting:

• You see at a glance which transactions belong to which invoices and which documents are still missing.

Curious about how our machine learning works? Read this article about Machine Learning at re:cap.