Start with the decision

Many data requests fail before the query begins. “Show me the data for this product” does not explain what decision the result should support, so it is impossible to know whether sales value, unit volume, price, or market share matters most.

A better starting point is a decision-shaped question. Is this product growing consistently, or did one promotion create a temporary spike? Did a brand grow because it sold more units or because its average selling price increased? Is a category attractive because demand is expanding, or because incumbents are losing share?

Different decisions require different entities, metrics, and baselines. A data tool can shorten retrieval time, but it cannot repair an undefined question.

Fix the scope with four coordinates

Before querying, write down the entity, metric, period, and comparison. This is the smallest useful analysis brief.

Clear coordinates make a query cheaper and its result easier to audit. A prompt such as “How is this brand doing?” forces the agent to invent a scope and often returns more data than the decision needs.

  • Entity: product, brand, shop, or category? Is there an exact ID or canonical JD.com URL?
  • Metric: units, GMV, ASP, or a deliberate combination?
  • Period: one month, the last three complete months, or a year-over-year window?
  • Comparison: the entity’s previous period, a competitor, or the entire category?

Resolve the entity before calculating

Use an exact product ID or standard product URL for product-level questions. Titles change, and similar names may refer to different variants. A stable identifier prevents unrelated products from being merged.

Brand, shop, and category names should be aligned to the dataset dictionary first. Spacing, capitalization, translations, and platform aliases can otherwise split one entity into several values.

Preserve the exact filter values used in the result. A useful analysis does not only produce a number; it leaves enough context for the next query to repeat the same scope.

Begin with the minimum useful query

Do not request every metric and dimension at once. Start with the smallest query capable of testing the idea, then drill down.

To test whether a product is growing, begin with monthly units for the last three complete months. If the trend is real, add GMV and ASP to separate volume movement from price or mix movement. Add brand, shop, or category dimensions only when the explanation requires competitive context.

This sequence avoids a common failure mode: receiving a wide table with many fields but no clear answer to the original question.

Read level and change together

Growth rates can exaggerate a tiny base, while absolute values can hide an emerging shift. A decision-ready result should answer both: how large is it now, and what changed relative to a meaningful baseline?

Use complete months for monthly data. A partial current month should not sit beside closed historical months in the same trend comparison.

  • Absolute values indicate commercial scale.
  • Month-over-month change reveals recent direction.
  • Year-over-year change reduces seasonal distortion.
  • Category share separates market growth from entity-specific growth.

Run three checks before concluding

First, check definitions. Confirm that units, GMV, and ASP match the dataset contract; do not treat displayed platform values, order counts, and item quantities as interchangeable.

Second, check scope. Keep month boundaries, category levels, aliases, and product coverage consistent across every comparison.

Third, check anomalies. Promotions, stockouts, launches, and delistings can all create sudden movement. Data identifies where change occurred, but one curve rarely proves why it occurred.

Do not infer a monetary unit from the platform name. The current JD.com contract defines the GMV calculation; currency should only be stated when the dataset contract explicitly provides it.

A reusable query brief

A strong request can follow this structure:

The value of the template is not its formatting. It forces the question to be defined before retrieval begins. Querying is the middle of analysis; a decision comes before it, and validation comes after it.

  • Entity: exact product URL, product ID, brand, shop, or category.
  • Period: the last N complete months or explicit start and end months.
  • Metrics: units, GMV, and ASP as needed.
  • Grouping: month, brand, category, shop, or product.
  • Baseline: previous month, prior year, competitor, or category total.
  • Output: trend, top movers, and a statement of the query scope.