Data analysis
Analyze business data with AI
Ask business questions in Chat and investigate sales, Affiliate, ads, finance, and inventory.
#You do not need a ready-made dashboard
Data analysis is an independent AI capability, not just the charts already available on a page. Ask business questions in Desktop Chat, the mobile App, or a connected messaging channel. AI can query authorized data, compare metrics, investigate details, and prepare findings or files.
For example, “Why did sales fall this month?” may require product sales, Affiliate contribution, advertising, and returns rather than one dashboard.
There are currently 29 analysis views across nine business data families: orders, Affiliate, ads, finance, reviews, returns/refunds, Customer Service, inventory, and Shop SPS. These are capabilities, not menu names or technical identifiers you need to memorize. See Available business data for coverage.
#Start with a specific question
A useful instruction states scope, period, comparison, output, and restrictions.
Analyze sales changes for Brightloom US across the last two complete calendar weeks, using Shop-local payment dates. Start with the Shop, then identify the largest product and SKU declines. Separate GMV, orders, and units by currency. Show findings, supporting figures, and next checks; do not change settings.
Compare current US warehouse stock with SKU sales over the last 30 days. Flag potential shortages and excess stock, explain mapping or data gaps, and do not create replenishment or replacement orders.
For Brightloom US Sample Applications created this month, compare review, fulfillment, and conversion by BD and product. Distinguish current status from application date and avoid double-counting linked orders.
You do not need to select a data view first. AI selects available data around the question and should clarify ambiguous scope or explain missing access.
#What an analysis involves
- Define scope: Shops, advertisers, or warehouses; dates, time zones, currencies, and comparison periods.
- Check availability: authorization and historical coverage; current observations versus historical summaries.
- Start broad, then investigate: products, SKUs, Creators, creatives, aftersales reasons, or warehouses.
- Make results verifiable: figures, definitions, scope, and limitations, with Excel or CSV if requested.
- Let you decide the next action: advice is not a saved setting, a sent message, or a replacement order.
Follow up with “expand the third product,” “compare without canceled orders,” “use settled income only,” or “export these results.” Keep related analysis in the same conversation.
#Common tasks
| Goal | Example request |
|---|---|
| Explain sales changes | Compare two weeks at Shop, product, and SKU level; separate order-volume and average-order-value changes. |
| Evaluate sample investment | Compare sample approval, fulfillment, and subsequent sales by application month and BD; list Creators needing follow-up. |
| Investigate GMV Max | Explain changes in spend, orders, revenue, and ROI by campaign, product, and creative. |
| Reconcile finance | Organize this month's statements, fees, and payout links; distinguish buyer payments from seller payouts. |
| Identify product issues | Query all 1–5-star reviews from the last 30 days and compare them with return/refund reasons. |
| Evaluate support | Summarize response and escalation trends; separately report unpaid-order outreach attribution and mature holdout results. |
| Plan stock | Combine current available stock and recent demand, identifying warehouses shared by multiple Shops. |
| Investigate SPS | Start with current US Shop SPS, then drill into abnormal metrics and relevant products or providers. |
You can combine business areas, but the same order can appear in sales, Affiliate attribution, and ads attribution. Do not add those totals together. See Interpreting results.
#Prerequisites
- Shop data: connect and authorize the Shop, with the required analysis permissions and service access;
- Ads: connect the advertiser and confirm access to the relevant advertising stores;
- Inventory: connect WMS, synchronize goods, and map warehouses;
- Mobile or other channels: keep the handling Desktop online with a working model connection.
Authorization does not guarantee all historical platform data is available. Ask AI to state actual coverage. Missing data is not zero.
#Reports and recurring follow-up
For a downloadable result, specify format, scope, and detail level. See Reports and exports.
Validate a daily or weekly report in Chat before turning it into a Scheduled Task. Recurring work needs an explicit schedule, business scope, and delivery destination; a one-off analysis does not authorize ongoing delivery.