Skip to main content
The quality of your prompts directly impacts the quality of your insights. Here’s how to ask questions that unlock FlipsideAI’s full potential.

​
The fundamentals

​
Be specific

Vague questions get vague answers. Specific questions get actionable insights. 
Tell me about swaps on Blur.

​
Define your terms

Don’t assume FlipsideAI knows what you mean by “whales,” “recently,” or “active users.” Define it explicitly.
  • Whales: greater than $100K holdings
  • Retail: less than $1K holdings
  • High-value transactions: greater than $50K
Use explicit dates instead of relative terms:
  •  “2024-01-01 to 2024-03-31”
  • L “recently” or “this quarter”
Exception: phrases like “last 30 days” or “past week” are fine.
  • Quality users: score >4 on reputation metric
  • New addresses: first transaction within date range
  • Active addresses: >5 transactions in period
  • Power users: >20 transactions or >$10K volume

​
Build context incrementally

Start broad, inspect the results, then narrow your focus. Think of it as a conversation, not a single query.
1

Explore

”Show me DEX activity on Base"
2

Narrow

"Focus on the top 3 protocols by volume"
3

Deep dive

"Break down Uniswap swaps by token pair for the top 10 pairs"
4

Analyze

"Which pairs have the highest concentration of whale trades?”

​
Advanced techniques

​
Provide examples

Show FlipsideAI what you’re looking for: 
Find protocols similar to Uniswap (automated market makers with liquidity pools)
that launched on Arbitrum in 2023.

​
Specify output format

Tell FlipsideAI how you want to see the results:
  • “Show this as a time series line chart”
  • “Create a bar chart comparing the top 10 protocols”
  • “Format this as a table with columns for: date, protocol, volume, unique users”
  • “Make this Twitter-ready with a clean, simple visualization”

​
Request validation

Ask FlipsideAI to double-check its work:
  • “Validate these results using an alternative calculation method”
  • “Cross-check this TVL against DeFiLlama data”
  • “Show me the query you used and explain your methodology”

​
Break complex questions into steps

For multi-part analyses, guide FlipsideAI through the process: 
Let's analyze Aave lending behavior in three steps:
1. First, show me total deposits and borrows over the past 90 days
2. Then, break down deposits by asset (top 10 assets by volume)
3. Finally, identify addresses that deposited >$100K and took a loan within 24 hours

​
Common patterns that work

​
Comparison questions

Compare [metric] across [entities] for [time period]
Examples:
  • “Compare DEX volumes across Ethereum, Arbitrum, and Optimism for the last 30 days”
  • “Compare average transaction fees for Uniswap v2 vs v3 this month”

​
Trend analysis

Show me the [metric] trend for [entity] over [time period], broken down by [frequency]
Examples:
  • “Show me the daily active user trend for Aave on Ethereum over the past 90 days”
  • “Show weekly stablecoin transfer volumes on Solana for Q1 2024”

​
Cohort analysis

Find [user type] that [action] during [time period], then track [metric]
Examples:
  • “Find addresses that first used Uniswap in January 2024, then track their monthly trading volume”
  • “Find wallets that bridged to Arbitrum in Q1, then show their DeFi adoption rate”

​
Anomaly detection

Identify [entities] where [metric] is [threshold] compared to [baseline]
Examples:
  • “Identify pools where liquidity dropped >50% in a single day over the past month”
  • “Find tokens where trading volume spiked >10x compared to their 30-day average”

​
Working with specific data

​
Reference contract addresses

When in doubt, use contract addresses instead of token symbols: 
Analyze transfer activity for 0xA0b86991c6218b36c1d19D4a2e9Eb0cE3606eB48 (USDC on Ethereum)
This eliminates ambiguity and ensures you’re looking at the right asset.

​
Specify chains explicitly

Many tokens exist on multiple chains. Be explicit: 
Show me USDC transfers on Arbitrum (not Ethereum or Polygon)

​
Use table names when you know them

If you’re familiar with Flipside’s data structure, reference specific tables: 
Query the ethereum.defi.ez_dex_swaps table for Uniswap v3 activity
Type @ in the chat to browse available tables.

​
What to avoid

Don’t do this: - Assume FlipsideAI remembers context from previous conversations (it doesn’teach chat is isolated) - Use ambiguous terms like “popular,” “big,” “good,” or “active” without defining them - Ask multiple unrelated questions in one prompt (break them into separate messages) - Request real-time data without specifying “use live query mode”

​
Pro tips

Don’t just accept the numbers. Dig in:
  • “Why is this number so high?”
  • “What might explain this trend?”
  • “Are there any data quality issues I should be aware of?”
Upload an example image or describe the style you want:
  • “Make this chart look like the ones in Dune Analytics”
  • “Use a dark background with purple accents”
  • “Create a minimalist chart suitable for a presentation”
Save your favorite styles as Rules for consistent output.
If something looks wrong, say so: - “This number seems too low. Can you double-check?” - “I expected to see more activity in December. Did we miss something?” - “This chart is hard to read. Can you simplify it?”
For structured, methodological analyses, use Workflows:
  • Type @ to browse available workflows
  • Workflows provide pre-trained, quasi-deterministic approaches to common analytical questions

​
Example prompts to try

Get inspired by these real-world prompts:
  • DeFi
  • NFTs
  • Cross-chain
  • User behavior
  • “Compare lending protocol TVL across Ethereum Layer 2s (Arbitrum, Optimism, Base) for Q1 2024. Include market share breakdown.”
  • “Find all addresses that provided >$100K liquidity to Uniswap v3 ETH/USDC pools on Arbitrum, then track their impermanent loss.”
  • “Analyze liquidation events on Aave v3 for the past 30 days: total value liquidated, top liquidated assets, and liquidator concentration.”

​
Next steps

Customize with Rules

Save your preferred prompt styles and output formats

Use Workflows

Apply structured methodologies to complex questions

Work with data

Validate, download, and manipulate results

Analysis modes

Understand how FlipsideAI processes different questions