What Intelligence answers

The short definitions

If you only need the one-line version of each term:

• MixShift Intelligence is an analytical service that explains why a business metric moved on an Amazon account, decomposing the change into contributions that reconcile exactly to the measured total and attaching the evidence behind each one.
• Attribution is the Intelligence service available today. It answers five requests, listed below.
• An Operational Bridge is a price, volume and mix decomposition of a retail metric between two periods: it splits the total change into the parts contributed by traffic, conversion, price and the mix of what sold, with the parts summing exactly to the whole.
• An Advertising Bridge is the same decomposition applied to advertising performance, at campaign, ad group, keyword or target, search term, or ASIN grain.
• An Item Group is the grouping level between a single ASIN and the whole account: the set of products a team manages together, such as a line or a pack-size family. It is the level most explanations actually live at.
• HCAM, the Hierarchical Causal Attribution Model, is the method underneath Attribution. It walks the account hierarchy from the total down to Item Groups and products, decomposes each metric, retrieves the supporting context, and limits how strongly a conclusion may be stated based on the evidence available. The full method is in the whitepaper, RAG-Enabled Hierarchical Causal Attribution Model.
• Weeks of cover is how long current sellable inventory would last at an expected rate of demand. See Limits and freshness for how the window is set on each data source.

One distinction worth stating plainly, because the words are overloaded. In marketing measurement, "causal attribution" usually means incrementality: media mix models and geo experiments estimating what would have happened without the spend. HCAM answers a different question. It decomposes what actually happened, exactly, and states only what the evidence supports. The two are complementary. Use an experiment to size incremental lift; use Attribution to find out what moved, which products carried it, and what the evidence says about why.

Which metrics Intelligence decomposes

Revenue is the entry point, not the whole of it. Each metric below gets its own decomposition, with the pieces summing to the measured change for that metric.

Operational metrics, decomposed by the Operational Bridge: sales, units, sessions and glance views (traffic), conversion, buy box, ASP and price, sellable inventory, weeks of cover, and lost sales.

Advertising metrics, decomposed by the Advertising Bridge: spend, sales, ACOS, ROAS, impressions, clicks, CTR, orders, CPA, AOV, CPC, and the same-SKU, other-SKU and view-through splits.

Together these are what the Intelligence page means when it says Intelligence investigates changes in sales, advertising, and operations. The Combined Read joins the two sides for the same period; it does not add a third set of metrics.

Returns and availability are read as inputs and as supporting context; they are not decomposed metrics of their own today.

The five requests

All five requests below are part of the Attribution service, and every one of them needs the same two basics: a merchant (a seller ID and marketplace, or a brand) and a period to look at. Most also compare that period against another one, and most let you choose how the result is grouped. The differences are in what each one is answering.

Operational Bridge

Why did ordered sales move, and what else moved with them

This request bridges your total ordered sales between two periods down to the item level: how much of the change came from traffic, from conversion, from price, and from the mix of what sold, with every piece summing exactly to the total change. It runs on retail data, from Seller Central or Vendor Central. It needs a merchant, a period with something to compare it to, and optionally a grouping (by item group, ASIN, or sub-brand) if you want to see which items drove the move.

Which metrics it decomposes

The operational set: sales, units, sessions and glance views (traffic), conversion, buy box, ASP and price, sellable inventory, weeks of cover, and lost sales. This is what the Intelligence page means by "the supported operational metrics" on the Operational Bridge card. Advertising metrics are decomposed by the Advertising Bridge below, not by this request.

Advertising Bridge

Why did advertising results move

This request does the same kind of bridge for advertising: it decomposes a change in ad performance at whichever grain you care about, campaign, ad group, keyword or target, search term, or ASIN. It needs a merchant, a period comparison, and the grain you want the breakdown at.

Which metrics it decomposes

Spend, ad sales, ACOS, ROAS, impressions, clicks, CTR, ad orders, CPA, ad AOV, CPC, ad conversion rate, and the same-SKU, other-SKU and view-through splits of both sales and orders.

Combined Read

This request puts the retail bridge and the advertising bridge for the same period side by side and joins them: TACOS, attributed share, paid pressure, and ad AOV against ASP. It needs just a merchant and a period; there is no separate grain to choose since it works from both bridges at once.

Lost Sales

This request estimates sales lost to stock-outs, showing sellable inventory and weeks of cover for the merchant and period you ask about. It ships with evidence attachments on by default, so you get the supporting detail without asking for it separately. It needs a merchant and a period.

Monthly Read

This request bundles month-over-month and year-over-year retail and advertising performance into one result, and it is the substrate behind the Monthly Report Max plugin skill. It needs only a merchant; left alone it defaults to the last complete month, though you can point it at a specific one.

Related

• Get access
• Use the Intelligence Workspace
• Run an insight from the command line