Help

Use DataCrunch Pro in Claude

Ask Claude about your store in plain words, like "Why did sales drop last week?", and get answers from your own DataCrunch Pro numbers. This page shows how to set it up, what Claude can and cannot do, and how to turn it off.

What this is

A connector is a safe link between Claude and another app. The DataCrunch Pro connector lets Claude look up your store and ad numbers in DataCrunch Pro while you chat, so you don't have to copy numbers back and forth.

Under the hood it uses MCP (Model Context Protocol), an open standard for connecting AI assistants to apps. You don't need to know anything about it to use the connector.

What you need

A DataCrunch Pro account with at least one project (one project is one store). Start a free trial if you don't have one.
Your store and ad accounts connected in DataCrunch Pro, so there are numbers to ask about.
A free trial that is still running, or a paid plan. When a trial ends, Claude gets a message telling you where to pick a plan.
A Claude account. On the free Claude plan you can add one custom connector.

Connect in three steps

In Claude, open Customize → Connectors. Find DataCrunch Pro in the list and click Connect.
Not in the list yet? Click Add custom connector and paste this address: https://datacrunchpro.com/mcp
A DataCrunch Pro page opens. Sign in with Google, the same way you sign in to the app.
Tick the projects Claude may use, then click Allow. Untick any store you want to keep out of Claude.

That's it. Start a new chat and ask a question. If Claude doesn't use DataCrunch Pro, click the + button in the chat, open Connectors and check that DataCrunch Pro is switched on.

Using Claude Code (Claude in the terminal)? Create a personal token under Settings → Claude / MCP access and run:
claude mcp add --transport http datacrunch https://datacrunchpro.com/mcp --header "Authorization: Bearer <your token>"

Things you can ask

Ask the way you'd ask a colleague. You can write in English, Norwegian or Swedish.

“How did my store do last week compared with the week before?”
“Why did ROAS drop in October?” (ROAS means return on ad spend: sales from ads divided by what the ads cost.)
“Which products sell the most, and which ones get clicks but no sales?”
“Which Google search terms cost money without bringing orders?”
“What should I fix first today?”
“How many customers came back for a second order this year?”
“Which products are missing a price or a link for the Swedish market?”

Not sure what a word means? See the metric glossary (needs sign-in), or just ask Claude to explain it.

What Claude can see and change

Claude can only reach the projects you ticked, and only what your role on each project allows. Your role is set by the project owner on the Team page.

Your role	What Claude can do for you
Viewer	Look at project settings, products, markets, the to-do list and recent activity.
Analyst or Ads editor	All of the above, plus answer questions about sales, ads, customers, products and search terms.
Editor, Admin or Owner	All of the above, plus make changes: edit product titles and descriptions, add or edit markets, mark to-do items done, start a data refresh, and send product listings to Google Merchant Center.

Every tool is labelled as either look-up only or a change, so Claude knows to ask for your OK before it changes something (unless you told Claude to always allow it). Every tool Claude uses is written to the project's activity log, so you can see what happened.

What Claude can never do

Change your Google Ads or Meta ads, budgets or bids.
See passwords, API keys or other stored login details.
Move money, change your plan or see card details.
Delete a project, invite people or change anyone's role.
Reach projects you didn't tick, or anything your role doesn't allow.

Full list of tools

A tool is one thing Claude can ask DataCrunch Pro to do, like "get top products". This list is read straight from the connector, so it always matches what Claude sees.

Look-up tools (41): only read, never change anything

Compare periods: Compare two time periods for key metrics. Returns revenue KPIs and ecom data for both periods side by side. Use when the user asks to compare months, weeks, or any two date ranges.
Diagnose commerce trend: Multi-source causal diagnosis for changes in conversion rate, revenue, ROAS/MER, AOV, or traffic quality. Compares current vs previous periods across traffic mix, campaigns, search-term intent, discounts, products, orders, geography, and returns.
Diagnose conversion change: Deep conversion diagnosis: compares conversion rates by channel, product performance, and traffic vs transaction ratios across two periods. Use when conversion rate is changing or when asked about checkout/funnel issues.
Diagnose discount dependency: Analyzes discount dependency: what share of revenue relies on discounts, how discount vs full-price orders compare on AOV and volume, and whether discount usage is growing. Use when asked about margin health, discount strategy, or promotion dependency.
Diagnose return rate change: Deep return/refund diagnosis: compares return rates, top returned products, return reasons, and refund amounts across two periods. Use when return rate is changing or for product quality analysis.
Diagnose revenue change: Deep revenue diagnosis: compares KPIs, channels, top products, and order metrics across two periods to identify revenue drivers and declines. Use when revenue is changing or for overall business health diagnosis.
Diagnose ROAS change: Deep ROAS diagnosis: compares KPIs, campaigns, and channels across two periods, identifies top declining and improving campaigns, and computes deltas. Use when ROAS is dropping/changing or when the user asks about ad efficiency.
Get activity: Recent system log entries and timeline notes for the project (syncs, pushes, errors).
Get background task status: State and result of a background task id returned by a trigger_*/push_* tool.
Get catalog readiness: Per-market catalogue readiness: price/stock/URL coverage, Shopify handle coverage, markets without a storefront URL.
Get product: Full product detail (copy, prices per market, Merchant status, storefront link). Give product_id or offer_id.
Get project settings: Project settings summary: integrations configured, sync toggles, platform, Shopify status, editable fields. Never includes secrets.
Get source coverage: Which data sources are connected and which recent days have full/partial/no data.
List actions: Action Engine recommendations for the project, optionally filtered by status.
List markets: Selling markets (ProjectMarket rows) with effective storefront URL pattern and product_url_source (manual | shopify_auto).
List products: Paged product catalogue with revenue metrics. Filters: q (text), status, country, optimization; sort revenue|catalog.
List projects: List the DCP projects available on this connection with your role on each. Call this first.
Query action items: List prioritized Actions from the project's Actions board (same data as the dashboard). Use when the user asks about their action list, recommendations, or open tasks.
Query campaign performance: Get ad campaign performance: spend, revenue, ROAS, clicks, impressions, CPC per campaign. Covers both Google Ads and Meta Ads.
Query channel breakdown: Get traffic channel breakdown from GA4: organic search, paid search, social, direct, email, referral, AI/LLM traffic. Shows sessions, revenue, transactions, conversion rate per channel.
Query commercial targets: Read month-by-month revenue targets and planned ad spend entered in DataCrunch (Commercial targets). Use when the user asks about revenue goals, targets, budgets, or MER hints for planning.
Query customer analytics: Get customer analytics from real order data: repeat rate, LTV, new vs returning customers, cohorts, top products by customer. Use when asked about customer loyalty, retention, lifetime value, or repeat purchases.
Query customer repeat analysis: Analyze repeat customers from real order data with data-quality coverage. Use for questions like customers who bought more than N times, returning customers, new vs returning customers, repeat purchase rate, and whether customer counts are reliable.
Query daily trend: Get daily time series data for revenue, ad spend, sessions, transactions, ROAS, and MER. Use when the user asks about trends, patterns over time, or day-by-day performance.
Query DCP ads scope: DataCrunch Pro (DCP) Google Ads scope: autopilot program settings and the list of campaigns linked to DCP (created in the wizard or explicitly 'tracked' for automation). Same scope the nightly optimizer uses. Meta campaigns appear with meta_* ids; Google with campaign resource names.
Query discount performance: Get discount analytics: discount rate, outlet vs checkout discounts, top discount codes, product discount breakdown. Use when asked about discounts, promotions, or coupon performance.
Query ecommerce overview: Get real e-commerce order data from the OMS: actual revenue, orders, AOV, basket size, customers. Compares OMS data vs GA4 tracking. Use when asked about real/actual order numbers or tracking accuracy.
Query geographic orders: Get order distribution by city and country. Use when the user asks where customers are, geographic performance, or shipping destinations.
Query keyword performance detail: Analyze Google Ads keyword rows: spend without conversions, winners, low quality score, CPC/CPA/ROAS, and pause/review candidates. Use for keyword-level questions.
Query live source snapshot: Fresh read-only API snapshot from connected sources when the user asks about now/fresh/live data or synced data is stale. Supports GA4, GSC, Google Ads keyword/search-term/campaign context, Meta totals/top ads, Merchant status/performance, and Omnium today/last-60m.
Query order behavior: Advanced order/customer behaviour: AOV, repeat/new mix, geography, discounts, payment, shipping, product-line sales, returns/refunds. Use for commerce behaviour and trends.
Query payment methods: Get payment method breakdown: orders, AOV, refund rate per method. Use when asked about how customers pay or payment performance.
Query product intelligence: Find and analyze one specific product by name/model/SKU/offer id/GTIN, even if it is not a top seller. Returns catalog/PIM/Merchant fields, sync status, analytics product metrics, order-line sales, market/country rows, issues, stock/availability, and optional live Merchant lookup. Use whenever the user names a product, model, SKU, or asks why Crunch cannot find product info.
Query product opportunities: Find product and Merchant/catalog opportunities: products with traffic but no sales, no visibility, high return risk, margin/cost hints, stock/status/Merchant issue problems, and market breakdowns.
Query project settings summary: Non-secret project configuration: integrations connected (which platforms), display currency/timezone, sync toggles, ecommerce platform name. No API keys or tokens. Use when the user asks what is connected or how the project is set up.
Query returns refunds: Get return and refund analytics: return rate, refund rate, refund amounts, top returned products. Use when the user asks about returns, refunds, or product quality issues.
Query revenue KPIs: Get revenue KPIs: total revenue, ad spend, ROAS, MER, AOV, CPA, sessions, transactions, net profit, margin. Use when the user asks about overall performance, revenue, spend, profitability, or efficiency.
Query search terms: Analyze Google Ads search terms from synced data: wasted spend, converting terms, brand/non-brand and informational/transactional intent, campaign/ad group drilldown, and negative keyword candidates. Use for 'vilka söktermer?', irrelevant searches, search term waste, or intent quality.
Query top products: Get top products by revenue, views, or purchases from GA4/analytics data. Use when the user asks about best sellers, product performance, or views.
Research keyword ideas: Read-only Google Ads keyword research using Keyword Plan Ideas. Use for keyword research, market expansion, new SEO/SEM ideas, and search-volume/CPC estimates.
Run read-only SQL query: Execute a read-only SQL SELECT query against the project database. Use this only as a fallback for custom analysis not covered by deterministic tools. You MUST include WHERE project_id = :pid in every query. Allowed tables include ecom_orders, ecom_order_lines, metric_traffic_daily, metric_product_daily, keyword_metrics_daily, search_term_metrics_daily, gsc_metrics_daily, products, product_costs, project_markets, ads_managed_campaigns, action_items.

Change tools (11): need Editor role or higher

Create market: Create a selling market.
Delete market (removes data): Delete a selling market (products are detached, not deleted).
Push merchant listings: Queue a Merchant Center push. mode=missing pushes only market variants that do not exist yet; mode=all re-pushes every listing (URLs, copy, prices, availability).
Refresh merchant catalog: Queue a pull of product statuses/issues from Google Merchant Center.
Regenerate actions: Run the Action Engine now (LLM budget applies) and persist new recommendations.
Trigger catalog refresh: Queue a catalogue import from the store platform (Shopify or Omnium) and Merchant.
Trigger sync: Queue a full data sync + analysis (same pipeline as nightly). Optional lookback_days.
Update action: Change an action's status; dismiss_reason is stored when dismissing.
Update market: Update fields of a selling market.
Update product copy: Set user-edited title / description / Google product category / lock on a product (project-wide copy).
Update project settings: Update non-secret project settings. Editable: name, brand_name, display_currency, display_timezone, pim_default_locale, merchant_sync_enabled, meta_catalog_sync_enabled, ecommerce_cutover_date.

Your data

Claude gets only the answer to each question, such as a table of last week's sales. It doesn't get a copy of your whole account.
DataCrunch Pro does not receive or store your Claude chats. For each tool Claude uses, we log who used it, which project, when, whether it worked, and the short settings it was given (for example a date range). We keep this so you and we can see what happened, and to fix problems.
Customer names and emails are hidden from Claude.

Read the full privacy policy.

Turn it off

For one chat: click + in the chat, open Connectors and switch DataCrunch Pro off.
In Claude: go to Customize → Connectors and remove DataCrunch Pro.
In DataCrunch Pro: go to Settings → Claude / MCP access and click Disconnect. This stops access at once, even if Claude still lists the connector.

To change which projects Claude can use, disconnect and connect again, then tick a different set.

If something goes wrong

Claude says it can't find any projects: You may have unticked every project, or you are signed in to a different Google account than the one that owns the store. Disconnect, connect again and tick the right projects.
Claude says I don't have permission: Your role on that project doesn't allow it. See the table above, and ask the project owner to change your role on the Team page.
Claude says the free trial has ended: Pick a plan on the pricing page. The connector works again right away.
The numbers look old or empty: DataCrunch Pro updates every night. Ask Claude "Which data sources are connected?" to see what's missing, or reconnect the account in the app under Project settings.
Claude keeps asking me to sign in: Remove the connector in Claude and add it again. If it still happens, contact us.
Claude doesn't use DataCrunch Pro at all: Check that the connector is switched on for the chat (the + button, then Connectors), and mention your store by name in the question.

Get help

Email hello@f4.no. We answer in English, Norwegian and Swedish. Found a security problem? Email the same address with "Security" in the subject.

DataCrunch Pro is made by F4 Labs AS, Oslo, Norway (Org. no. 936714315).