AdBeacon

List the Google Analytics 4 reports and the exact vocabulary each one accepts. Call this BEFORE the first get-ga4-data call in a conversation. It returns, per report (traffic, landing_page, product, path): its breakdowns, the breakdowns that read it by default, its selectable columns, the fields each breakdown may filter on, and the column rows are ordered by. Also the filter operators, paging limits and the rules a request must follow.

list-ga4-reports

Show how each order's credit was split ACROSS every channel, one row per order. This is the cross-channel view: for each order you get the date, customer email, order total, and the fraction of that order credited to meta, tiktok, pinterest, snapchat, googleAds, bingAds, chatgptAds, direct, googleSearch, bingSearch, klaviyo_email and attentive_email — plus `others` for everything unnamed (including any other email platform, e.g. Mailchimp). Use it to answer "which channels contributed to this order?" or "how much of last week's revenue did Meta actually drive per order?". Choosing between the order tools: - `get-order-attribution` (this one) — every channel's share of each order. NOT channel-scoped. - `get-order-list` — orders attributed to ONE channel. Use when the user names a channel. - `get-order-journey` — the individual touchpoints behind ONE order. Notes: - `start_date` and `end_date` must be `YYYY-MM-DD`. - For the fractional models the channel shares plus `others` sum to 1 per order. For `full_impact` each channel counts whole orders instead, so they do NOT sum to 1 and `others` is null. - `lighthouse` is NOT available here, unlike the other order tools. - `attribution_window` does NOT accept 0/lifetime. - Results are paginated; read `pagination.has_next_page` before assuming you have them all. Every page costs the same scan, so prefer a large `limit` over many pages. - Data comes from the hourly attribution summary, so orders placed in the last hour may not appear yet. This matches the Optimize screens exactly. - An account with no connected store is refused with an explicit error rather than an empty result, so an empty `orders` array always means "no orders matched", never "this account has no store connected". FOR ANY TOTAL, READ `totals`. NEVER ADD UP `orders` YOURSELF. The response carries a `totals` object holding each channel's credit, the residual and the revenue summed across the WHOLE filtered set — every matching order, not just this page. Those figures are computed in the database and are exact. Summing the rows by hand is not: a week can hold hundreds of orders, and adding them up produces confident, wrong answers. `totals` is the whole result set regardless of `page` or `limit`, so one call answers "how did my channels do" without paging at all. Page through `orders` only when you need the individual orders themselves. DO NOT VOLUNTEER A SUMMARY OF A SUBSET. When you list some orders, present the rows and stop. Do not append a count or a value total for "these N orders" — nothing in the response provides it, so you would be adding the rows up yourself, and that is exactly the arithmetic that produces confident wrong answers. A real example: asked for 25 orders under last-click, a summary was appended whose channel counts totalled 27 and whose value was $100 out. If a total for a NARROWER slice is wanted, do not compute it — re-query with filters that describe that slice (a tighter date range, a different customerType) and read `totals` again. If that is not possible, say the figure is your own count of the rows shown rather than presenting it as data.

get-order-attribution

List the individual orders attributed to one advertising channel over a date range. Returns one row per order — order id, order date, customer email and name, and the order total both formatted and as a raw number for arithmetic. Use it to answer questions about specific purchases; use get-summary-data instead for channel totals. Notes: - `start_date` and `end_date` must be `YYYY-MM-DD`. - `attribution_window` is in days, and `0` means lifetime (no look-back limit). - By default this lists EVERY attributed order for the channel, which on a busy account is a very large result. When the user is asking about a particular campaign, ad set or ad, pass `breakdown` + `entity_id` to scope it — that is both faster and what they meant. - `search` matches the start of an order id, and also the customer email and name — but for accounts that mask customer details it silently narrows to the order id alone, so a name or email search will legitimately return nothing on those accounts. - Results are paginated; read `pagination.has_next_page` before assuming you have them all.

get-order-list

List the products sold in the orders attributed to one advertising channel over a date range. Returns one row per product variant — name, product id, SKU, units sold, order count, distinct customers, average units per order, and revenue both formatted and as a raw number for arithmetic. Use it to answer "what did this channel actually sell"; use get-order-list for the individual orders behind those totals. Notes: - `start_date` and `end_date` must be `YYYY-MM-DD`. - `attribution_window` is in days, and `0` means lifetime (no look-back limit). - Revenue is the raw product basket for the attributed orders, not a re-weighted attribution share, so it will not tie out to the channel's attributed revenue. - There is deliberately no search parameter — filter the returned rows instead. - By default this covers EVERY attributed order for the channel. When the user is asking about a particular campaign, ad set or ad, pass `breakdown` + `entity_id` to scope it. - Results are paginated; read `pagination.has_next_page` before assuming you have them all.

get-product-list

Get store profit-and-loss figures — sales, orders, refunds, cost of goods, profit, ad spend and new versus returning customers — added together across every connected store (Shopify, WooCommerce, BigCommerce, Custom Store; not Amazon). These are the figures of a Store P&L tile on the Custom Dashboard. `breakdown`: - `none`: one row for the whole period (a scorecard) - `day`, `week` or `month`: one row per period, newest first; send sort_by "date" with sort_direction "asc" for a trend, oldest first Every response also carries `totals` for the whole period, whatever page you read. A `week` or `month` row is labelled with the start of its period, so the first label can fall before `start_date` — but only days between `start_date` and `end_date` are counted, and the rows add up exactly to `totals`. Notes: - These are store figures, not attributed to any ad channel. Spend is the connected Meta, Google, TikTok, Snapchat, Pinterest and Bing accounts; Amazon Ads and organic are excluded. - `customer_type` filters orders and refunds but never spend, so CAC, NCAC, ROAS, MER and POAS then compare all spend with part of the orders. - COGS is the product cost on the order date, 0 when no cost is set, and always 0 for WooCommerce, BigCommerce and Custom Store. Transaction cost is Shopify only. - `start_date` and `end_date` must be `YYYY-MM-DD`, at most 425 days apart. - Money is in the account's reporting currency; a division by zero is null.

get-store-data

Rank the top creatives for one channel by a single metric — the Creative Dashboard's bar chart. Returns the top N creative assets over the whole date range (not a time series), ordered by `metric` descending, each carrying its rank. Use it for "which creatives are our best/worst by X"; use `get-creative-data` when you need many metrics per row, or more rows than a chart would show. Notes: - `metric` is restricted to the metrics the chart can actually render, and the set differs by channel — Pinterest has no `thumbstop`. Check `list-creative-channels`. - The `google` channel does not implement this tool; its dashboard has no chart. Use `get-creative-data` with sort_by there instead. - `limit` is the bar count, defaulting to 8 to match the dashboard. - The rows carry `start_date` and `end_date` because a bar has no other period marker.

get-top-creatives