Top 5 Analytical Thinking Patterns to Maximize Thesis.io

Thesis.io is more than a search box. It is your DeFi business-intelligence (BI) surface for reasoning across on-chain and off-chain data. The secret to getting real signals is not a magic prompt but an analyst-grade structure. Use these five patterns to turn vague asks into crisp, action-ready queries.

Don’t just prompt. Think like an analyst.

Pattern 1: Ask with context

Because Vague in → vague out. Make the ask specific on time, scope, characteristics, and output shape.

Instead of

“Give me alpha.”

Try

“List alpha candidates on Solana from the last 24 hours with market cap < $1M and rising unique buyers.”

Sample Template

[Task] on [chain/segment] during [time window]

where [constraints] and [evidence signals].

Return [columns]. Sort by [rank rule].

Pattern 2: Clarify your definitions

Words like “manipulation,” “whale,” or “momentum” hide assumptions. Make them measurable.

Instead of

“Any whale manipulation today?”

Define first

Manipulation = abnormal trade burst: 30-minute volume z-score > 3 and at least 2 whale clusters provide > 60% of window volume.

Then ask

Detect potential manipulation events on Hyperliquid for today.

Trigger Conditions (must both be true):
1. Trading volume in a 30-minute window has a z-score > 3 (≥ 3σ above baseline).
2. More than 2 distinct labeled whale clusters together account for over 60% of the window’s total volume.

For each triggered case, return:
• token → traded asset (e.g., ETH-PERP, SOL-PERP)
• window_start → timestamp for the start of the 30-minute window
• volume_z → z-score of the window’s trading volume
• clusters → list of whale clusters involved and their volume share
• tx_count → number of transactions in the window
• wallet_links → relevant whale wallet explorers

Explain briefly why each case triggered.

Pattern 3: Cluster and Segment like a Charm Analyst

To deal with massive data, one tool rises as an analyst’s best friend: clustering. It turns noise into patterns and reveals hidden structures. Use wallet clusters, cohorts, and buckets to transform chaos into insight.

Aliases

The “Clustering” concept has similar concepts but different names: Grouping (databases), or cohort analysis (marketing), or bucketing in querying/BI, or segmentation (computer vision).

What to cluster

• Wallets by behavior: hold time, trade frequency, size, venue preference.
• Tokens by fundamentals: market cap buckets, liquidity tiers, sector tags.
• Events by timing: listing cohorts, news windows, or funding rounds.

Example:

Bucket tokens on Solana by MC: nano (<$10M), micro ($10–50M), small ($50–300M).

Within nano and micro, surface top 15 by 24h velocity and holder growth.

Return tables per bucket and one-paragraph cohort insight.

Pattern 4: Compare, rank, and explain

Relative context beats absolute numbers. Always ask for a baseline, ranking, and attribution.

Example:

Compare ORAI vs top 10 AI tokens over the last 30d on:

- marketcap, volatility, MA 7D

- liquidity depth at 1% price impact

Rank by volatility. For the top 3, add one-sentence attribution for each.

Pattern 5: Decompose the task (plan the steps)

Hard questions become tractable when you split the work and show the trail.

Clarify the Plan

Screen → Diagnose → Validate → Rank

Example:

Step 1 (Screen): AI-sector tokens, last 7d, top decile in market cap

Step 2 (Diagnose): Attribute moves to whales/news/DEX listings. Show evidence links.

Step 3 (Validate): Social mentions 3d MA vs 14d baseline.

Step 4 (Rank): 0.4 growth + 0.3 whale_inflow + 0.2 social + 0.1 liquidity depth.

Return the final table and brief notes per token.

— -

Bottom line

The best analysts thrive by reducing uncertainty and eliminating ambiguity. Context cuts through uncertainty in massive data; clear definitions resolve ambiguity in concepts. With these five patterns: context, definitions, clustering, comparisons, and decomposition, you can turn Thesis.is into your sharpest DeFi assistant.