Skip to main content
Building powerful automations with Jsonify becomes much easier when you follow a few key design principles. This section outlines best practices to help you create workflows that are effective, reliable, and easy to understand.

​
1. Plan Your Goal Clearly

Before you use a single block, clearly define what you want your workflow to achieve. Ask yourself:
  • What is the ultimate output I need? (e.g., a list of company names and their websites, extracted product reviews, a summary of articles).
  • What are the starting points? (e.g., a list of URLs, a search query, a set of variables).
  • What are the main steps involved in getting from the start to the end, even if a human were doing it?
A clear goal makes it easier to choose the right blocks and structure your workflow logically.

​
2. Think Sequentially

Workflows in Jsonify execute blocks in a sequence, from top to bottom. The output of one block often becomes the input or context for the next.
  • Logical Order: Ensure your blocks are arranged in a logical order. For example, you can’t Extract Data before you Open Websites or Follow Links to a page.
  • One Thing at a Time: While some blocks can perform multiple internal steps (like Explore Content), it’s often clearer to break down complex tasks into a series of simpler blocks.

​
3. Choose the Right Block for the Job

Jsonify provides a variety of blocks, each designed for specific tasks. Understanding their purpose is key:
  • Need to input a list of URLs? Use Open Websites.
  • Need to click a button or type text? Use Interact with Page.
  • Need to find links based on a description? Use Find Links (followed by Follow Links).
  • Need to get specific data from a page? Use Extract Data.
  • Need to run the same logic for multiple inputs? Start with Open datasets.
Refer to the “Mastering Workflow Blocks” section for details on each block.

​
4. Build and Test Iteratively

Don’t try to build a massive, complex workflow all at once.
  • Start Simple: Begin with the first few core steps (e.g., opening a site, extracting one piece of data).
  • Test Often: Run your workflow frequently as you build it. This helps you catch errors early and verify that each part is working as expected.
  • Add Complexity Gradually: Once a section is working, add the next logical set of blocks.
  • Use Small Data Sets for Testing: When using Open datasets or processing lists, test with a small number of items first to speed up debugging.
  • Handling Paginated Data: To extract similar data from multiple pages (e.g., paginated search results or product listings), it’s often more reliable and less likely to trigger anti-bot measures to provide direct links to each page as input (e.g., using the “Add a numbered range of URLs” feature in Open Websites) rather than using a dedicated pagination block that simulates clicks on “Next” buttons. Processing each page URL as a new starting point can be beneficial because each new browser session initiated by the agent might use a unique IP address, making the activity appear less like a single bot rapidly clicking through pages from one IP.

​
5. Write Clear Instructions for the AI

Many blocks (like Extract Data, Explore Content, Find Links, Interact with Page) rely on your natural language instructions to guide the AI.
  • Be Specific: Vague instructions lead to unpredictable results. Instead of “Get info,” try “Extract the product price listed under ‘Our Price’.”
  • Provide Examples: For Extract Data, giving examples in the description field significantly improves accuracy.
  • Use Constraints: Tell the AI what not to do if necessary (e.g., “Do not extract text from the footer”).
  • Context is Helpful: Briefly stating the context (e.g., “You are on a search results page…”) can aid the AI.

​
6. Understand Data Flow

Be mindful of how data flows through your workflow:
  • Variables: Data from Open datasets is injected into subsequent blocks.
  • Block Outputs: Some blocks (like Find Links or Extract Data) produce output that is implicitly used by the next block (like Follow Links) or can be explicitly captured and used later.
  • Iteration: When using Open datasets or “Follow each link” in Follow Links, the subsequent blocks will run for each item/page, processing data in that specific context.

​
7. Keep It Readable

  • Name Your Workflows: Give your workflows descriptive names.
  • Logical Grouping: Arrange blocks in a way that visually represents the flow of logic.
By applying these principles, you’ll be well on your way to creating robust and efficient automations with Jsonify.