Going further with the MCP

You've got MCP connected and answered your first few questions. This page is for the next step: how to get faster answers, verify what comes back, work across marketplaces, and pair MCP with the rest of DataHawk.

Get faster answers

MCP is tuned for interactive use. A few habits keep responses snappy.

Narrow your date range

The shorter the window, the faster the answer. As a rule of thumb:

• 30 days: Your default for exploration
• 90 days: For quarterly trends
• 365 days: For year-over-year, but expect longer waits

When you're still exploring, start small. Widen the range only once you know which question matters.

Be specific about products and accounts

Asking about 20 specific ASINs is faster than asking about "all my ASINs". Same for "brand X" vs. "all brands". Filters do the heavy lifting. The more you give, the less DataHawk has to scan.

Reuse cached answers

DataHawk caches frequent queries. Ask the same question twice in a row and the second answer comes back faster.

One topic per session

Broad prompts like "do a complete audit of my business" can trigger 20+ tool calls and risk timeouts. Break it up:

• One conversation for sales
• One for advertising
• One for inventory

You'll get cleaner answers and won't hit rate limits.

Verify and reconcile numbers

Ask to see the tool call

For any number that matters, ask:

"Show me the DataHawk tool call you ran."

You'll see the exact filters, dates, and view used. Most "that number looks wrong" moments are explained at a glance; a marketplace you didn't mean to include, a date range that's off by a week.

Cross-check with the DataHawk app

When a number really matters (a board meeting, a supplier conversation), run the same query in Snowflake or BigQuery and compare.

Common gotchas when reconciling

• Currency: MCP normalizes to USD by default. Add "in local currency, broken down by marketplace" to see native amounts.
• Marketplace: Defaults to all connected marketplaces unless you scope. Always say "in the US marketplace" for accuracy.
• Refresh window: DataHawk refreshes once a day, at 9 AM PST. Queries run right at refresh time can catch a partial update. If something looks off, wait an hour and retry.

Work across Amazon marketplaces

If you sell on multiple Amazon marketplaces (US, UK, DE, JP, etc.), MCP queries every connected marketplace by default, which is rarely what you want for accuracy.

Always scope your question

Say which platform and which marketplace explicitly:

• "For Amazon US only…"
• "Across all my European Amazon marketplaces…"
• "On Walmart, last 30 days…"
• "Both Amazon and Walmart combined…"

A 5-second extra in the prompt saves 5 minutes of reconciling later.

Comparing marketplaces

Ask for side-by-side comparisons in one prompt:

"Compare revenue, units, ACOS, and conversion across my US and UK marketplaces, last 30 days."

Numbers come back normalized to USD by default. Add "in local currency" if you'd rather see native amounts.

Managing multiple DataHawk workspaces

If you run separate workspaces for different brands or clients, connect each to your AI client separately. MCP doesn't yet span workspaces within a single conversation.

Work with the investigation flow

For "why" questions ("why did revenue drop?", "what's driving the ACOS spike?"), MCP triggers a dedicated investigation flow. The AI builds hypotheses, gathers evidence across domains, and explains its reasoning instead of just dumping numbers.

When it triggers

Diagnostic questions, those starting with "why", "what's driving", "what changed", trigger it automatically.

How to redirect mid-investigation

If the AI heads down the wrong path, course-correct in plain English:

"Skip the catalog analysis. Focus on advertising changes."

When it's not enough

If a conclusion feels wrong:

1. Drop into focused mode: "Show me the raw sales data for those 7 days, day by day."
2. Run the same query in the DataHawk app to confirm the underlying numbers
3. Email [email protected] if the discrepancy persists

Combine MCP with the rest of DataHawk

Use MCP to explore, the app to share

MCP shines for ad-hoc thinking. Once you find an insight worth sharing, build a chart in the DataHawk app and send the dashboard link.

Paste screenshots from Seller Central

Claude and ChatGPT both accept image uploads. Drop a Seller Central screenshot into the chat and ask:

"Reconcile this against DataHawk data for the same period."

Use MCP as a first draft

Generate a weekly summary in MCP, then refine it in the app, in a Google Doc, or wherever you usually polish.

When to fall back to another tool

MCP isn't the right home for every job. Switch tools when you need:

• Custom dashboards shared with non-MCP users → DataHawk app or Power BI / Looker templates
• Heavy data joins across millions of rows → BigQuery export + your own SQL
• Scheduled reports with strict SLAs → DataHawk app's reporting features
• Years of history → BigQuery (MCP focuses on the most recent ~180 days)
• Large CSV exports for offline analysis → BigQuery export or the app

Not sure which to pick? Email [email protected] with what you're trying to do. We'll point you to the fastest path.

During beta

DataHawk MCP is currently in beta. Free for all DataHawk customers while we polish it.

A few things to expect during this phase:

• Frequent improvements: Query routing and data coverage get better most weeks
• Occasional changes: How questions get interpreted may shift slightly between updates
• New data domains added regularly: Vendor surfaces, brand analytics, and more are landing on a rolling basis
• Rare brief downtimes: We try to schedule upgrades around US daytime hours

Report what's broken

When you find an issue, email [email protected] with:

• The exact prompt you ran
• Which AI client you used (Claude, ChatGPT, Cursor, etc.)
• The tool call output (ask the AI to show it)
• What you expected to see

Pro habits worth building

Save your best prompts

The third time you write the same prompt by hand is when you should save it. Use:

• Claude Projects for repeated workflows
• ChatGPT custom GPTs for team-shareable prompts
• Your AI client's prompt library for personal ones

See Ready-to-use prompts for ready-made templates to start with.

Set session defaults up front

For multi-turn analyses, anchor the session in your first message:

"For the rest of this chat, assume US Amazon marketplace, last 30 days, and brand 'Acme' unless I say otherwise."

Schedule a weekly DataHawk hour

Block 30–60 minutes once a week to run your standard analyses: weekly summary, inventory check, ACOS review. Most users find it pays back its time within the first month.

Send the answer somewhere useful

Pair MCP with your team's tools using n8n or Make. Even one auto-generated daily digest into Slack or email cuts a lot of manual dashboard checking. See Set up DataHawk MCP for client connection guides.

---

What's next