KNIME
Getting Started: KNIME
Our KNIME Primer for Spreadsheet Users
1. Introducing the KNIME Analytics Platform
Who this guide is for
People who do their analysis in spreadsheets and are:
– Still using large, fragile spreadsheets for recurring tasks
– New to KNIME and want a simple, practical starting pointMost of the people we train work in finance and audit, so the examples use general ledger exports and trial balances. The same steps work for any analysis you currently do in Excel.
This guide covers the basics of KNIME. Using AI in KNIME comes later, once you can build, read and rerun a workflow.
What you’ll be able to do after this guide
– Install KNIME and understand the main screen
– Build a first workflow that replaces a spreadsheet task
– Know where to look for logs, errors and performance issues
If you’re brand new to KNIME: Do sections 1–4, then jump to 7 to build your first workflow.
If KNIME is already installed: Skim 1, then go to 3, 4 and 7.
The Transition from Spreadsheets to Workflow Automation
If you do monthly reporting in Excel, you already know the problems. Formulas break when someone adds a row, nobody is sure which copy of the file is current, and it is hard to show a reviewer how you got from the source data to the final number.
KNIME is a free, visual workflow tool. You use it to turn a manual spreadsheet process into a workflow that records every step and calculation. Next month you run it again with one click instead of rebuilding it.
We've trained hundreds of finance and audit professionals on KNIME. This guide covers what you need to get started.
Figure 1
What is the KNIME Analytics Platform?
The KNIME Analytics Platform is free, open-source software for building data workflows. Instead of writing formulas or scripts, you connect nodes. Each node is a small block that does one job, such as reading a file, filtering rows or joining two tables.
You drag nodes onto a canvas and connect them in the order the data should be processed. Work that people often split across spreadsheets, SQL, Python and BI tools can be done in one workflow.
What it can do:
Read data from almost any source (files, databases, APIs, cloud services).
Clean, reshape and validate data with no-code or low-code logic.
Build machine learning models and evaluate them.
Automate repetitive data tasks.
Produce reports, dashboards and services that others can reuse.
Why we use KNIME
Most teams now have access to a large language model, and the obvious move is to paste the data in and ask for the answer. For analysis that someone has to sign off, this causes two problems. You cannot see how the model reached its answer, so you have nothing to show a reviewer who asks how a figure was calculated. The model is also probabilistic, so the same file and the same prompt can give a different answer the next time.
Asking the model to write Python or SQL instead leaves you with code that nobody on the team wrote and few of them can read. Someone still has to fix it when the source file changes and explain it when the logic is questioned.
In KNIME each step of the analysis is a node you can open, and you can see the data before and after it. The same workflow run on the same data gives the same result every time. The workflow is the record of what was done, so you can hand it to a reviewer.
That is why we start here. A team that can build and read KNIME workflows already has a visible, repeatable record of its analysis. When you add AI later, it goes into that workflow as one defined step, and the steps around it stay as they were. You can see where the model was used, check what it produced and show a reviewer both. Without that foundation, AI output is hard to check and harder to defend.
From our training sessions: The audit trail benefit is undersold. When we train internal audit teams, the first "aha moment" is usually when they realise they can show exactly how a number was calculated. Excel hides that logic across cell formulas in multiple tabs. KNIME makes it visible in one workflow.
When should you use the KNIME Analytics Platform?
Use KNIME when you keep returning to the same spreadsheet for a semi-automated task, such as a reconciliation, and want to stop the errors that come from manual copying, pasting and editing.
Great use cases
KNIME works well when your data comes from several sources and has to be combined the same way each time, or when a spreadsheet has become too large, slow or fragile to manage. It suits tasks you repeat, such as monthly reporting or routine data preparation, because you build the steps once and rerun them. Anyone can follow the logic by reading the workflow. It is also a practical way to try machine learning without a full coding environment, and to share prototypes with colleagues.
Typical workflow moments
Workflow Stage | How KNIME Helps |
|---|---|
Start of a project | Explore messy or unfamiliar data, test ideas and build understanding before committing to deeper analysis. |
Daily or weekly cycles | Clean, validate and distribute information consistently, reducing repeated manual work. |
Advanced analytics | Build predictive models, apply algorithms, evaluate results and keep everything documented in a clear workflow. |
Operationalisation | Move analytics into production, integrate with enterprise systems and automate processes for reliable long-term use. |
Who can use KNIME?
You do not need to write code to use KNIME. If you can build a spreadsheet, you can learn to build a workflow, because the tasks you already do in Excel become nodes you can see on the canvas.
More technical users can add Python or R code and connect KNIME to databases, cloud systems and APIs, and the logic stays visible in the workflow. Because one person's workflow can be read and rerun by another, KNIME suits work where someone has to check the result, such as audit, regulatory reporting and quality assurance.
The official KNIME website lists the platform's full capabilities and further learning resources.
Want help rolling this out to your team?
We run KNIME training for finance and audit teams using your own processes and data. Get in touch to talk about a short pilot or a training cohort.
2. Installing the KNIME Analytics Platform
You need the KNIME Analytics Platform installed on your machine before you can build a workflow. Most people can install it without help. This section takes you from the download to opening KNIME for the first time.
If you’re on a locked-down corporate laptop, you may need IT to install KNIME for you. In that case, send them this section or point them to the KNIME installation page.
Follow these steps:
Figure 2.1: Download Landing Page
Choose a release line. The Standard Release is updated every six weeks and has the newest features, so it suits most users. The Long Term Support Release is updated less often and suits teams that put stability first. The differences are explained in the release notes FAQ.
Figure 2.2: Releases
Scroll down to the section for your operating system (Windows, Linux, macOS).
Figure 2.3: OS Installation Packages
Click the installer type you want to start the download:
Windows options:
Installer: Extracts files, adds a desktop icon and suggests memory settings.
Self-extracting archive: Creates a folder with the installation files and needs no extra software.
Zip archive: Can be downloaded and extracted anywhere you have full access rights.
Once the file is on your machine, run the installer.
On Windows, open the .exe file and follow the standard installation prompts.
On macOS, open the .dmg file and drag KNIME into your Applications folder.
On Linux, extract the downloaded archive and place the folder wherever you prefer.
Most users finish in a couple of minutes.
Launch KNIME from your applications list. The first thing KNIME asks for is a workspace. A workspace is a folder where your workflows and settings will live. You can accept the default option or choose your own. Once you have chosen one, KNIME opens with an empty canvas ready for your first workflow.
Figure 2.4: Workspace Preview upon Launch
Advanced Options (You Can Skip These on Day One)
As your work grows, you may want to add extra capabilities such as Python integration, database tools, machine learning libraries or cloud connectors. To install these, open the Menu and select Install Extensions. Choose what you need and KNIME will set it up for you. You can add extensions whenever you like, so there is no need to install everything at the start.
Figure 2.5: Install Extensions under the Menu option
If you work with large datasets, you may want to increase KNIME’s memory limit. This is done by editing a small configuration file called knime.ini located in the installation directory. Raising the value of the -Xmx setting allows KNIME to use more memory. Most users do not need to change this at first.
KNIME is now ready to use, with no activation step. You can start a workflow, open the example workflows or try rebuilding a task from your own work.
3. A Quick Tour of the KNIME Interface
When KNIME opens you see the user interface, where you build workflows and inspect data. Each area of the screen has one job, so the layout is quick to learn.
This section explains each part.
Figure 3.1: KNIME Modern UI General Layout (application tabs, side panel, workflow editor and node monitor)
Application Tabs
The row of tabs at the top of the window. Each tab is either the entry page or an open workflow. They work like browser tabs: click one to switch to that workflow and close it when you have finished. This lets you keep a reference workflow open while you build a new one.
Side Panel Navigation
The panel on the left of the screen has four tabs:
Description (Info): Shows information about the selected workflow or component, including what it does and any documentation its author added.
Node Repository (Nodes): A list of all available nodes. You pick the nodes for your workflow from here. Each node does a specific task, such as reading a file, filtering data or writing to a database.
Space Explorer (Explorer): A file browser that lets you navigate your KNIME Hub spaces or your local KNIME workspaces. Use it to open and organise workflows, components, data files and KNIME Hub content.
Workflow Monitor (Monitor): Lists the nodes in your workflow that have errors or warnings, so you can find the step where something went wrong.
Note: K-AI is the AI assistant built into KNIME Analytics Platform. It can answer questions (Q&A mode) and build workflows (Build mode). You do not need K-AI for this guide. We build everything by hand so you can see what each step does.
Workflow Editor
The large canvas in the centre of the screen. It shows your nodes and the connections between them, and it is where you build and run your workflow.
Help, Preferences and Menu
Menus at the top right of the screen. They give you access to help, extension installation, preferences, themes and update settings.
Node Monitor
A panel that shows the output of the selected node. Use it to check what your data looks like at each step, including tables and flow variable values.
Figure 3.2: User interface elements (workflow toolbar, node action bar, rename components and metanodes)
Workflow Toolbar
A toolbar at the top left of the canvas with common actions such as save, undo and redo, plus commands for the selected node.
Node Repository Result
The filtered list of nodes shown in the side panel after you search the Node Repository. It shows only the nodes that match your search term or chosen categories.
Rename
Lets you rename metanodes and components on the canvas. A clear name tells others what the metanode or component contains and what it does.
Zoom Controls
Buttons to zoom in, zoom out or fit the whole workflow to the screen.
Node Action Bar
A bar that appears next to the selected node, with buttons to configure, execute or reset it, or open its view.
Try it now: Open KNIME, create a new empty workflow, click a node in the Node Repository and drag it onto the canvas. Watch how the Node Monitor updates when you select different nodes.
4. The Workflow Editor: Your Main Canvas
The Workflow Editor is the canvas where you build workflows. In Excel the logic is hidden in cell formulas. In the Workflow Editor each step of your analysis is a node, and the connections between nodes show the order the steps run in, so anyone can follow and rerun what you did.
When you create or open a workflow, the Workflow Editor fills the centre of the screen. You drag nodes onto it from the Node Repository, configure them and connect them. The canvas grows as you add nodes, so there is no limit on the size of a workflow.
Figure 4.1: Workflow Editor elements (node description panel, workflow canvas, port outputs and node execution)
Panning Across the Canvas
To move around your workflow, hold the Space bar, then click and drag the canvas in any direction. You can also press P to switch to pan mode, and V to switch back to selection mode. The cursor changes to a hand while you are panning. You will need this once a workflow is larger than your screen.
Zooming In and Out
The Zoom Controls appear in the bottom-right corner of the Workflow Editor. Click the plus (+) button to zoom in or the minus (−) button to zoom out. Scrolling the mouse wheel up zooms in and scrolling down zooms out. The Fit to Screen button adjusts the zoom so the whole workflow fits in view.
Ctrl+2 (Windows/Linux) or Cmd+2 (macOS) fits the whole workflow to the screen, which is useful before you show it to a manager or reviewer.
5. All About Nodes
Selecting Nodes
Click a node to select it and KNIME highlights it. To select several nodes, click an empty part of the canvas and drag a rectangle around them. You can also hold Ctrl (Windows/Linux) or Cmd (macOS) and click nodes one at a time to add them to the selection.
You can then drag the selected nodes as a group, and their connections move with them.
Understanding Node Anatomy and Structure
Every node has the same layout, which tells you what it does, whether it has run and what it connects to.
The Node Body
Figure 5.1: Anatomy of a node
Each node is a coloured square with an icon. The node name, such as Excel Reader or Row Filter, appears above it, and an annotation you can edit appears below it. The colour shows what kind of job the node does: nodes that read data are orange, nodes that change data are yellow, nodes that write data are red and visualisation nodes are blue.
Status Indicators: Understanding the Traffic Lights
The traffic light under each node shows its state:
Figure 5.2: Traffic light status
A node can show a green light with a small warning triangle. This means it ran but hit a problem that did not stop it, for example missing values, or rows it skipped because of a format problem.
Check every warning. Hover over the warning triangle to read the message, or open the Workflow Monitor in the side panel to see all the warnings in the workflow.
A progress indicator appears while a node is running. On a large dataset this can take several minutes.
Each node has small connection points called ports. They set what a node can receive and what it passes to the next step, and they are how data, models, database connections and other objects move through a workflow.
Once you can read the ports, you can tell what a node needs as input and what it produces as output.
Input and Output Ports
Data ports are the small triangles on the sides of a node. You connect two nodes by drawing a line from the output port of one to the input port of the next.
Input ports are on the left of a node and receive data. Most nodes have one input port. Some have more: the Joiner node, which merges two tables, has one input port for each table.
Output ports are on the right of a node and pass its results to the next node. Some nodes have more than one output port, for example when they split data into two groups.
Both input and output triangles point right, the direction in which the data flows.
Port Shape Conventions: Triangles vs. Squares
Some ports are triangles and some are squares. The shape tells you what the port carries.
Triangular Ports: Data Streams: Triangular ports carry data tables, shown as black triangles. When you connect two of them, the whole table passes from the first node to the second, which processes it and passes its own output on. Data moves from left to right across the canvas.
Square Ports: Objects and Resources: Square ports carry a single object that is passed on whole: a database connection (red square), a model (grey square), a report (blue square) or an image (green square).
Port Multiplicity: Single vs. Multiple Connections
Single Input/Output: Most nodes have one input port and one output port. A Row Filter has one black input, which receives the table to filter, and one black output, which passes the filtered table on.
Multiple Inputs: Some nodes have several input ports because they combine or compare data from different sources. The Joiner node has two input ports for merging tables, and the Concatenate node can take several tables and stack them. Comparing budget with actuals, for example, needs two inputs: the budget and the actuals.
Multiple Outputs: Some nodes have more than one output. A Row Splitter has two output ports: one for rows that match your criteria and one for rows that do not. Each output can feed its own branch of the workflow, and branches can be brought back together later with a Joiner or Concatenate node.
Optional Ports: Some input ports are optional and have a dashed outline. The node runs without them, but connecting one adds a feature. For example, many writer nodes have an optional flow variable input at the top. Without it, the node writes to the file path you typed in. With it, the node can build the file name for you, such as "GL_Report_January_2025.xlsx", so you do not have to change the path each month. You can ignore optional ports until you need them.
Port Colours and What They Mean
Each port type has its own colour and shape:
Figure 5.3: Different Port Types
Data Table Ports (Black Triangle Ports): These carry data tables, with rows and columns like an Excel sheet or a database table, and they are the ports you will use most. When you read a general ledger export with Excel Reader, filter it with Row Filter, total it with GroupBy and save it with Excel Writer, every connection uses black ports. Most workflows are a chain of nodes joined this way.
Database Connection Ports (Red Square Ports): These carry a live connection to a database such as SQL Server, Oracle or your company's ERP system. The connection node sets up the link but reads no data. You plug the connection into other database nodes, which read or write the data. If your data lives in a database, you can bring in only the rows you need without exporting to Excel first, and you always work with current data.
Flow Variable Ports (Red Circle Ports): Flow variables are settings passed between nodes, such as a file path, a date range or a threshold. Their ports are red circles at the top of a node, separate from the data ports on the sides. Instead of typing "January 2025" into several nodes, you can set it once as a flow variable at the start of the workflow and have each node read it. Next month you change it in one place. You will not need flow variables at first, but they are what let a workflow run month after month without editing.
Tree Ensemble Model Ports (Grey Square Ports): These carry a trained machine learning model, in this case a tree ensemble. One node trains the model on historical data and outputs it through a grey port, and a predictor node applies it to new data. In finance this could be used to forecast revenue, predict which receivables will go overdue or flag transactions that look unusual. You can come back to this once the basics are familiar.
Report Ports (Blue Square Ports): These carry a report while it is being built. KNIME can combine tables, charts, text and images into a formatted report and export it as a PDF or HTML file. Your workflow can import the data, run the calculations, create the charts, assemble the report and save or email it without manual steps.
Image Ports (Green Square Ports): These carry images, such as a chart produced by a node. You can pass a chart into a report, save it as a file or combine several charts into a dashboard. For example, a budget against actuals bar chart can go straight into your monthly variance report.
From our deployments: New users overthink port colours. In 90% of finance workflows, you'll only use black data ports. We typically don't introduce database or model ports until week two of training. Focus on the black triangles first.
6. Working with Nodes: The Basics
To copy a configured node, select it and press Ctrl+C (Windows/Linux) or Cmd+C (macOS), then press Ctrl+V or Cmd+V to paste. The copy keeps the original node's settings.
Figure 6.1: Right-clicking a node opens the node context menu
This is useful when importing multiple similar files or applying the same filter criteria to different datasets.
Replace Nodes
To swap a node for a different one, drag the new node from the Node Repository and drop it onto the node you want to replace. KNIME keeps the connections where the ports match.
This is quicker than deleting the old node and connecting a new one.
Node Description Panel
When you select a node, the Description panel shows what it does, its settings, its inputs and outputs, and often examples.
Figure 6.2: Node information (left-click a node and select Info)
Read the description before you configure a node you have not used before. It tells you which settings matter.
Learning from Example Workflows
If you are not sure how to do something, search KNIME Hub for an example workflow, for instance "financial reporting" or "Excel import". Download it, look at how it is built and copy the parts you need.
Figure 6.3: Accessing Example Workflows via the Explorer
Because every step is visible, opening someone else's workflow shows you exactly how it works.
7. Build Your First Workflow (Adding, Connecting, Configuring Nodes)
This section covers the three things you do most in KNIME: adding nodes to the canvas, connecting them and configuring them. It ends with a first workflow you can build on your own data.
Adding Nodes to Your Workflow
There are several ways to add a node to the canvas.
Drag and Drop from Node Repository
The simplest is to drag a node from the Node Repository on the left of the screen.
Type in the search box at the top of the Node Repository, for example "excel reader" to import a trial balance or "groupby" to total transactions by cost centre. The list filters as you type.
Figure 7.1: Dragging and Dropping a Node to the Workflow Editor
Click the node, drag it onto the canvas and release the mouse button. The node appears with a red status light because it has not been configured yet.
This is the easiest method when you are looking through the nodes to see what is available.
Quick Node Search
If you know which node you want, quick search is faster.
Right-click an empty part of the canvas and select Quick Add Node, or press Ctrl+. (Cmd+. on macOS). A search box opens.
Figure 7.2: Quick Add Node Feature
Type the node name, such as "row filter". Use the arrow keys to pick a result and press Enter to add the node where you right-clicked.
Figure 7.3: Quick Add Node Search Feature
Once you know the node names, this is quicker because you do not have to leave the canvas.
Node Recommendations and K-AI
KNIME can also suggest which node to add next.
After a node has run (green status light), select it, right-click the canvas and choose Quick Add Node again.
Figure 7.4: Workflow Coach and K-AI
The list now starts with the nodes that other KNIME users most often add after that one. After an Excel Reader, for example, it might suggest Row Filter, GroupBy or Joiner.
Node recommendations (the Workflow Coach) and K-AI only work when you are signed in to your KNIME account. Signing in also gives you access to your KNIME Community Hub spaces.
Searching the Node Repository Effectively
The Node Repository search covers node names, descriptions and tags, so you can search for what you want to do as well as by node name.
For example, "sum" finds the GroupBy and Column Aggregator nodes, and "missing" finds the nodes that handle missing values. Partial words and small typos still return results.
Figure 7.5: Utilising Node Search
To see the full list again, delete your search text or click the X.
Connecting Nodes to Create Data Flow
Connections set the order in which data moves from one node to the next.
Creating a Basic Connection
Hover over the output port (the triangle on the right) of the first node.
Click and hold, then drag towards the input port (the triangle on the left) of the next node. A line follows the cursor.
When you reach an input port that can accept the connection, it is highlighted. Release the mouse button to connect the two nodes.
Figure 7.6: Drag and Drop I/O
KNIME draws a line from the first node to the second.
Understanding Connection Validation
KNIME only lets you connect ports of the same type: black data table ports to black ports, red database ports to red ports.
If the types do not match, the input port will not highlight, or the node will show an error.
If you cannot connect two nodes, check the port colours and shapes. You may need a node in between to convert one type to the other.
Connecting to Multiple Input Ports
Nodes with more than one input port need each connection in the right place.
The Joiner node has two input ports on its left. The top port takes the first table, for example actual results, and the bottom port takes the second, for example the budget. The node needs both.
Hover over an input port to see what it expects.
Managing Connection Layout
In a large workflow, connection lines can cross and make the flow hard to follow.
Click a connection to select it, then drag the small handles along it (bendpoints) to route it around other nodes.
You can also move nodes, and their connections follow. Lay workflows out from left to right so the order of the steps is clear.
Removing Connections
To delete a connection, click it and press Delete, or right-click it and select Delete Connection.
If the next node had already run (green), it resets, because its input data is no longer available.
Figure 7.7: Deleting a node connection
An input port takes only one connection. If you connect a new line to an input port that already has one, KNIME replaces the old connection.
Configuring Nodes
Configuring a node sets what it does. Each node type has its own settings, but you configure them all the same way.
Opening the Configuration Dialog
Double-click a node to open its configuration dialog, which shows all of its settings.
You can also right-click the node and select Configure, or select it and press F6.
Figure 7.8: Node Dialog
The dialog stays open until you click OK (saves settings and closes), Cancel (discards changes and closes) or Apply (saves settings but keeps dialog open).
Understanding Configuration Tabs
Most dialogs split their settings across tabs. The first tab holds the settings the node needs to work.
The other tabs hold advanced and performance settings, and you can usually leave them at their defaults to start with.
Using the Preview Function
Many dialogs have a Preview button. It shows a sample of what the node will produce with your current settings, without running the node.
Figure 7.9: Preview Output of Selected Node
If the preview looks wrong, change the settings and check again before you run the node.
Example: Configuring Excel Reader
Most spreadsheet users start with the Excel Reader node, so here is how to configure it.
Double-click the Excel Reader node to open its dialog. The main tab has a file selection section.
Click Browse and navigate to your Excel file (for example, your general ledger export). Select the file and click Open. The file path appears in the configuration dialog.
Figure 7.10: Excel Reader configuration dialog with a file selected
If the workbook has several sheets, choose the one with your data from the dropdown menu, for example "GL Export" or "Trial Balance".
Tick "Table contains column name headers in row number" if the first row holds column headers, which is usual for system exports. KNIME then uses that row for column names.
If there are title rows or blank rows above the data, use "Skip first data rows" to tell KNIME how many to skip.
Click Preview to see the first rows of your data. Check that:
Column names appear correctly as headers
Account numbers, amounts and dates are in the right columns
Data types look correct (numbers show as numbers, not text)
If it all looks right, click OK. The status light changes from red (not configured) to yellow (configured and ready to run).
Configuration Best Practices
If other people will reuse or review a workflow, set values yourself instead of relying on defaults. When reading dates, for example, type in the date format instead of relying on automatic detection. The workflow is then less likely to break on new data, and a reviewer can see what you chose.
Where a dialog has a description or notes field, use it to record why you chose a setting, especially for filters and calculations.
Flow Variables Tab
Most dialogs have a Flow Variables tab. It lets a flow variable control a setting instead of a fixed value. You can ignore it for now.
Memory Policy Settings
Where a node has an Advanced Settings tab, you will find memory policy options there. The default setting keeps processed data in memory for fast access, which works well for most workflows.
For very large datasets (millions of rows), select "Write tables to disk" to reduce memory consumption. This is slightly slower but stops KNIME running out of memory.
Your first workflow: Filtering a GL export
Say you have an Excel export of 50,000 journal entries and need every transaction over £10,000 for audit sampling.
Drag in an Excel Reader node and point it at your file.
Connect it to a Row Filter node and set the condition Amount > 10000.
Connect that to an Excel Writer node to save the filtered results.
Next month, point the Excel Reader at the new export and run the workflow again. Anyone checking the sample can open the workflow and see the filter you applied.
Quick Reference: Node Operations
The most common operations:
Adding nodes: drag from the Node Repository, or right-click the canvas and select Quick Add Node
Configuring: double-click the node, right-click it and select Configure, or press F6
Connecting: drag from an output port to an input port
Disconnecting: select the connection and press Delete, or right-click it and select Delete Connection
Moving: drag the node to a new position, and its connections follow
Copying: select the node, press Ctrl+C, then Ctrl+V to paste a copy with the same settings
Replacing: drag a node from the Node Repository onto the node you want to replace
Reading node documentation: select a node to see its description in the Description panel
Learning from examples: search KNIME Hub for example workflows related to your task
8. KNIME Tables, Joins and Pivot Tables
This section is a deeper explanation you can come back to. If you just want to get KNIME up and working, you can skim and move on.
Data moves through KNIME as tables, with rows and columns like a spreadsheet or a database table. Each column holds one data type, and KNIME supports more types than Excel.
Figure 8.0: Table Preview/Output
Table Structure and Column Data Types
In Excel a column can hold a mix of numbers, text and dates. In KNIME every value in a column has the same type. This stops problems such as numbers stored as text, which in a spreadsheet can drop out of a total without anyone noticing.
Figure 8.1: Transform and Change Column Data type in File Reader Node Dialog
As well as text and numbers, KNIME handles:
Numeric types: Whole numbers and decimals
Text types: Text of any length
Temporal types: Dates, times and combined date-times
Complex types: Lists within cells, sets, JSON, XML and binary data such as images
To check how each column has been read, switch an output table to the Statistics view. It shows each column's type, the minimum and maximum of number columns and the distinct values in text columns.
When KNIME reads a file, the reader node sets each column's type from its contents. If it gets one wrong, convert the column with a node such as String to Number, Number to String, Double to Int or String to Date&Time.
If several columns have been read as text, the Column Type Auto Cast node can convert them all to the most suitable types.
If you are using the File Reader node, you can also adjust column types directly within its configuration dialog. Open the Transformation tab (Figure 8.1) and change the type of any column before loading the data into your workflow.
Sorting
Hover over a column name in the table view and click the arrow to sort by that column. This changes only the view, not the node's output.
To sort the output itself, use the Sorter node. To change the order of columns, use the Column Resorter node.
Rendering Columns
In the table view you can change how numbers are displayed, for example as percentages, at full precision or as bars. Click the caret in the column header and choose a renderer. Like sorting, this changes only the view.
JSON, XML and image columns have their own renderers. JSON and XML can be formatted for reading, and images show as thumbnails.
Table Operations and Transformations
Most workflows need to combine or reshape tables.
Combining Tables Vertically
The Concatenate node stacks tables on top of each other, adding the rows of one table to another. Use it for:
Figure 8.2: Concatenate Node
Combining periods, such as four quarterly files into one annual table
Merging datasets from different departments or regions
Appending historical data to current records
If the tables have different columns, you can choose:
Union: Keep all columns from all tables (missing values where columns don't match)
Intersection: Keep only columns present in all input tables
Combining Tables Horizontally
The Joiner node combines tables side by side by matching values in key columns. It does the job of VLOOKUP in Excel, and it can also match on several columns at once and keep rows that have no match. You can set:
Figure 8.3: Joiner Node
Join type: Inner, left outer, right outer or full outer join
Join columns: Which columns to match on (you can use more than one)
Column handling: How to manage duplicate column names
Row filtering: Additional conditions beyond the join keys
In finance and audit work, joins are used to match transactions to master data, compare budget with actuals and reconcile data between systems.
Pivoting and Unpivoting
The Pivot node turns rows into columns, for summary tables and cross-tabulations. The Unpivot node turns columns back into rows, which you often need when the source is a report laid out for reading, for example with months across the top.
Figure 8.4: Pivot and Unpivot Nodes
Both are useful when a chart or a regulatory return needs the data in a particular layout.
Storage Management
KNIME keeps small tables in memory and writes large ones to disk in the background, so most users never need to change anything. If you work with very large datasets, you can set individual nodes to write their output straight to disk (see Memory Policy Settings in section 7), raise KNIME's memory limit in the knime.ini file (see section 2) or switch to the Columnar Backend, which is faster on large tables. The KNIME Analytics Platform User Guide explains each option.
Table Manipulation Best Practices
These habits keep workflows fast and easy to follow.
Column Management
Use the Column Filter node to drop columns you do not need, especially after a join or an import that brings in extra fields. Fewer columns means less memory used, faster processing and tables that are easier to check.
Row ID Management
Every KNIME table has a RowID column that identifies each row. It is hidden in the table view by default. You rarely need it, but joins and some machine learning nodes use it to keep track of rows.
You can:
View RowIDs by right-clicking a column header and selecting "Show RowID"
Generate new RowIDs using the RowID node
Use an existing column, such as the transaction number, as the RowID
Handling Missing Values
Most real datasets have gaps. KNIME shows a missing value as a question mark (?) in the table view, so a gap cannot be mistaken for a zero or an empty text value.
The Missing Value node gives you several options:
Remove rows: Simple, but you can lose a lot of data if gaps are common
Forward fill: Use the previous valid value (useful for time series)
Column statistics: Replace with mean, median or mode
Fixed values: Insert specific placeholder values
Linear interpolation: Estimate missing values based on surrounding data
Each method can change your results, so choose deliberately and record your choice. In audit work, that record belongs in your working papers.
File Export Options
KNIME has a writer node for each common output format:
CSV Writer: For exchanging data with almost any system
Excel Writer: For colleagues who work in Excel
JSON Writer: For passing data to other systems
Parquet Writer: For very large datasets
Database Writers: Write straight to a database
Data Apps for Interactive Access
A workflow can also be deployed as a Data App, a web page on KNIME Hub where colleagues can:
Upload their own data for analysis
Adjust parameters and rerun analyses
Filter and explore results interactively
Download reports
Table Documentation and Quality
Column Descriptions
You can add a description to each column. It appears as a tooltip and can be exported with the data. Use it to record:
Business meaning of each column
Units of measurement
Data source
Transformation logic applied
Figure 8.5: Adding new names and descriptions in Columns
This helps when you hand a workflow to a colleague or come back to it months later.
Data Quality Validation
You can build data quality checks into a workflow with nodes such as:
Figure 8.6: Table Validator and Rule Engine
Table Validator: Checks that a table has the columns and content you expect
Rule Engine: Flags rows that break a rule you define
Put these checks early in the workflow so bad data is caught before it reaches a report.
9. Workflow States and Logs
Node Reset and Workflow Execution
Resetting a node (F8) moves it from executed back to configured and clears its output. If you save a workflow after it has run, KNIME saves all the output data too, which can make the file very large. If the source data is easy to get again, reset the workflow before saving it. A reset workflow stores only the node settings.
KNIME Log and Node Operations
Resetting a node does not remove the record of what happened. Configuration changes, executions, errors and warnings are all written to a file called knime.log. To find it, open the menu and select Show KNIME log in File Explorer. The file is in your workspace, in the .metadata/knime folder.
Logging
Each workflow also keeps its own log file in the workflow folder, which holds only that workflow's messages. This is useful when you want to look at one workflow on its own. Log files are rotated when they reach 10 MB, so they do not fill your disk. The KNIME Analytics Platform User Guide covers the other places KNIME writes log messages and how to change log levels.
10. What's Next?
You now know the KNIME interface, how nodes connect and how data moves through a workflow. From here:
Read our book: KNIME for Auditors goes deeper into practical workflows for audit and finance teams.
Practise: Download KNIME and rebuild a manual Excel process as a workflow.
Get training: We run online KNIME training for finance and audit teams, and deliver in-house programmes for firms like Verizon. We also lead open training courses hosted by KNIME, with participants joining from across the world.
Add AI: Once your team is building and reading workflows, we can help you add AI steps to them, for example reading invoices or contracts and pulling out the fields you need. Get in touch to discuss.
Explore: Browse the KNIME Hub for example workflows you can download and adapt.
Continue Learning
Explore more guides, tools, and insights to advance your journey and build lasting capability across your business.
