Skip to main content

Breadcrumb Prompting Guidelines

These guidelines help you craft effective prompts for Breadcrumb.ai to analyze your data and produce the right visualizations and insights.

What Breadcrumb can do​

  • Analyze data across multiple sources
  • Visualize results as charts, tables, dashboards, and reports
  • Answer follow-up questions and remember context
  • Brainstorm data-driven business decisions

How to write effective prompts​

Breadcrumb understands everyday language, but giving it a bit more direction helps it deliver exactly what you need.

Include these elements in your prompt​

  • Business goal: what decision or question are you answering?
  • Metric(s): e.g., revenue, orders, MAU, conversion rate
  • Filters: e.g., region = US, channel = paid, status = completed
  • Time range & grain: e.g., last 12 months by month; last 30 days by day
  • Breakdowns/grouping: e.g., by region, product, channel
  • Output type: chart, table, or KPI (and preferred chart type if any)
  • Formatting/units: currency, percent, decimals, top N, sorting

The more concrete you are on the 6 items above, the less clarification is needed.

Ask a clear question​

Start with what you want to know or measure.

Examples:

  • “Show total sales by month.”
  • “What’s our average order value this year?”
  • “Which customer segment brings in the most revenue?”

Avoid vague prompts:

  • “Sales report”
  • “Customer data”

Describe how you’d like to see the result​

Breadcrumb automatically picks a visualization, but you can request a specific format.

Examples

  • “Show this as a line chart.”
  • “Make a bar chart by category.”
  • “Create a table with totals.”
  • “Put this into a dashboard comparing all regions.”

Ask follow-up questions​

Breadcrumb remembers the context of your last question. You can keep refining your analysis naturally.

Example flow

“Show sales by product.”

“Now break that down by region.”

“Add a trend line for the past 6 months.”

You don’t need to restate everything — Breadcrumb will build on your previous results.

Quick-start prompt templates​

  • Trend over time
    • "Show [metric] for [time range] by [grain], filtered by [filters], as a [chart type]."
  • Category comparison
    • "Compare [metric] by [dimension] for [time range], top [N], sorted [desc/asc], as a [bar/column] chart."
  • KPI
    • "Show KPI for [metric] in [time range], with unit [currency/%] and delta vs [previous period/same period last year]."
  • Table
    • "Create a table with columns [col1, col2, col3] for [time range] filtered by [filters], sorted by [col] [asc/desc], limit [N]."
  • Investigation
    • "Diagnose the change in [metric] from [period A] to [period B] by [dimension], and visualize top drivers."

Let Breadcrumb Guide You​

You don’t need to know the perfect question to begin. Breadcrumb can suggest ideas and ask clarifying questions to help you find what matters. You can structure your prompts as questions, and Breadcrumb will help narrow them down.

“What interesting things about our product data can you look into?”

→ “Do you want me to focus on usage, revenue, or retention?”

“Let’s take a look at retention. What can you do about it?”

→ “I can show retention trends, segment users by activity, or highlight where drop-offs happen. Which would you like to start with?

Advance prompting techniques​

Break big questions into smaller steps​

Complex analysis often works best when you guide Breadcrumb step by step. Start simple, then refine and broaden the scope.

Example: Understand why Q3 revenue dropped

  • Start simple:

    “Show total revenue by month for 2024.”

  • Spot the issue:

    “Highlight any drops or anomalies.”

  • Drill down:

    “Break down Q3 revenue by product category.”

  • Compare:

    “Compare that to Q2 performance.”

  • Find drivers:

    “Which regions or products contributed most to the decline?”

  • Summarize:

    “Give me a summary of key takeaways.”

Set up space knowledge​

Space knowledge allows you to define custom instructions and data context that Breadcrumb will use automatically for all prompts in a workspace. This is especially useful if you want consistent terminology, metrics definitions, or company-specific rules applied.

What you can define

  1. Metric definitions
    • Clarify what metrics mean to your organization.
    • Example: “Revenue is net revenue after discounts and refunds.”
  2. Data sources
    • Name and describe key datasets or tables.
    • Example: “Orders table contains all completed orders with columns: order_id, customer_id, product_id, order_date, revenue.”
  3. Business rules
    • Any special calculations or logic.
    • Example: “Churn rate is defined as customers active last month but inactive this month.”
  4. Units and formats
    • Default units, currency, or date formatting.
    • Example: “All monetary values should be in USD with 2 decimal places.”
  5. AI behavior
    • Instructions on how the AI should interpret prompts or respond.
    • Example: “Always suggest top 3 insights first. Provide charts only when meaningful.”

How to use

  • Once space knowledge is set, you can reference terms naturally without redefining them in every prompt.
  • Breadcrumb will apply these rules automatically, reducing errors and improving clarity.

Use group filters​

Group filters allow you to apply a single filter across multiple widgets or visualizations within a group. This makes it easier to explore data consistently without updating each chart individually. Group filters are expressed in natural language, so you can describe the condition just as you would in a question.

How it works:

  • When you set a group filter, every widget in the group automatically respects that filter.
  • Filters can include dimensions, metrics, time ranges, or conditions.
  • Breadcrumb updates all related visualizations in real time when the filter changes.

Examples:

  1. Basic filter
    • Filter: “Show US region only”
    • Effect: All charts in the group now only show data for the United States.
  2. Multiple conditions
    • Filter: “Region = US and Channel = Paid”
    • Effect: Every widget shows only paid channel performance in the US.
  3. Dynamic time filter
    • Filter: “Last 30 days”
    • Effect: All visualizations automatically adjust to display data from the past 30 days.
  4. Top-N filter
    • Filter: “Top 5 products by revenue”
    • Effect: Only the top 5 products appear across all charts in the group.