IQ Dimensions vs. Regex: When to Use Which?
One of the most common challenges is extracting clean data from messy dimension values (like campaign names with complex separators).
Use IQ Dimensions when the value doesn't exist or needs transformation/enrichment.
Example: Translating values, fixing capitalization, adding spaces, or categorizing sentiment (e.g., turning "angry_review" into "Negative Sentiment").
Use Regex (Regular Expressions) with "Condition" rules when the value already exists in the dimension string, and you just need to "pluck" it out without changes.
Example: Extracting "Retargeting" from VG | Retargeting | Product_Name.
Why: It is significantly faster than AI for large datasets (50k+ rows) and prevents loading delays.
Tip: If you need a Regex code but don't know how to write it, ask a tool like ChatGPT or Gemini: "Give me a Regex with group matching to extract the second element between pipes from this string: [Example String]."
Crafting the Perfect Prompt
AI performance is only as good as the instructions it receives. To move beyond "generic" summaries and get "state-of-the-art" reports, follow these best practices:
Step 1. Provide Real Examples (The "Rule of 20")
Don't just say "Categorize these reviews." Instead, provide ~20 manual examples.
Show the AI: "If the text says 'shipping was slow,' categorize as 'Logistics'; if it says 'great quality,' categorize as 'Product'." Once it sees the pattern, it will handle the next 5,000 rows with much higher accuracy.
Step 2. Iterate and Review
Never expect the first result to be perfect.
Prompt: Run your initial instruction.
Preview: Use the "Preview" function to check the data.
Refine: If it makes a mistake, tell it: "This specific value should have been 'X' instead of 'Y', please adjust the logic."
Finalize: Apply once the preview looks correct.
Tip: Avoid "Rigid" Constraints. When asking for categories or groupings, give the AI some breathing room. Instead of saying "Split this into exactly 4 categories," try "Split this into 4 to 6 categories." This allows the model to make smarter assumptions based on the actual data.
Advanced Report & Widget Creation
Whatagraph IQ has recently been updated to follow complex, multi-step instructions more precisely.
Multi-Tab Planning: You can now provide a detailed list of instructions for an entire report. For example: "Tab 1: Line chart for Sessions. Tab 2: Table for Top Landing Pages." IQ will now read these individual widget instructions more carefully.
Custom Text Summaries: Stop writing manual summaries. Use the IQ Text Widget to analyze data.
Advanced Strategy: You can paste your brand's voice guidelines or specific business logic (e.g., "If CPC is above $2.00, flag it as a critical warning") into the prompt.
Visual Themes: When prompting for themes, upload an image of your brand guidelines for the best results. Be aware that theme prompts currently work "collectively"—asking to change just the footer might result in the AI suggesting colors for the whole report to keep it cohesive.
IQ Chat: Your On-Demand Data Analyst
The IQ Chat (outside of the report) is now more "agentic." It doesn't just look at one table; it can perform multiple "tool calls" to investigate your data deeply.
Deep Audits: You can ask: "Audit the last two years of data, perform pivots by country and device, and identify where we are losing performance."
Source Groups: For the fastest performance, use IQ Chat with Source Groups. Because the data is pre-organized, the AI can fetch and analyze it much quicker than live API calls.
Summary Table: Quick Reference
Feature | Best For... | Key Tip |
IQ Dimensions | Transformation, translation, and enrichment. | Provide at least 20 examples for complex logic. |
IQ Text Widgets | Automated insights and client summaries. | Paste your "Business Logic" into the prompt for tailored advice. |
IQ Chat | Cross-source analysis and deep-dive audits. | Be specific about metrics (e.g., "Analyze Conversion Rate by City"). |