Advertising glossary: ACoS, TACoS, ROAS, retail media, ETL and MCP

Ad spend divided by the sales attributed to those ads, shown as a percentage. If you spend 20 and the ads are credited with 100 in sales, ACoS is 20%. Amazon Ads reports it, and a lower number means each unit of ad spend returns more sales. It is the inverse of ROAS.

Amazon Ads management → Guide: ACoS vs TACoS →

Ad spend divided by total sales, advertising and organic together, shown as a percentage. It shows how much of the whole business the ad spend represents, where ACoS only looks at sales credited to ads.

Guide: ACoS vs TACoS →

Revenue attributed to ads divided by ad spend, shown as a ratio. Spending 20 and being credited with 100 in revenue is a ROAS of 5. With the same attribution, ROAS is one divided by ACoS.

Guide: ACoS vs TACoS →

Advertising sold on a retailer's own websites and apps, targeted with the retailer's own shopper data. Sponsored product ads in a retailer's search results are the most common form. Amazon Ads and Walmart Connect are examples.

Walmart Connect advertising → Guide: What is retail media advertising? →

Amazon's demand-side platform: a tool for buying display, video and audio ads programmatically, on Amazon properties and on other sites and apps. It is separate from the sponsored ads bought in Amazon search results.

Amazon Ads management → Guide: Amazon DSP explained →

Moving data from its sources into a warehouse in three steps: extract it from the source, transform it into clean, consistent tables, and load the result. Marketing ETL does this for advertising and commerce data, so reports from different sources can be queried together.

Advertising data pipelines (ETL) → Guide: Marketing ETL explained →

The same three steps in a different order: data is loaded into the warehouse raw and transformed there. ELT keeps the original data available for re-processing; ETL cleans it first. Many pipelines combine both, landing raw files and loading cleaned tables.

Advertising data pipelines (ETL) → Guide: Marketing ETL explained →

The query language of the Google Ads API. It reads like SQL: you select fields such as clicks, cost and conversions from a resource such as campaign, ad group or search term, and filter by date. It is how reporting data is pulled from Google Ads into a warehouse like BigQuery.

Google Ads AI and automation → Guide: Google Ads to BigQuery options →

A spreadsheet in a defined layout that an ad platform accepts for creating or editing many campaigns, ad groups, keywords or ads at once, instead of one by one in its interface. Amazon Ads calls its version bulk operations sheets. Bulk campaign upload is the practice; the sheet is the file.

Campaign builder →

A keyword you add to a campaign so that your ads do not show for searches that contain it. It stops spend on searches that are irrelevant or that do not convert.

An open protocol, introduced by Anthropic in 2024, for connecting AI applications to external tools and data. An MCP server describes the tools, data and prompts it offers; an AI client connects, discovers them and calls them.

MCP server development → Guide: What is an MCP server? →

Software that uses an AI model to pursue a goal: it breaks the goal into steps, calls tools, checks the results and continues, instead of only answering a single prompt. How much it may do on its own depends on the tools and permissions it is given.

AI agents for advertising →