At this time, Moloco Commerce Media requires the following data inputs.

Catalog feed

The catalog feed is a set of metadata that represents all the items on your marketplace (i.e., What items are the marketplace or retailer selling?). Moloco uses the item metadata to create the campaign manager and train our machine-learning algorithms. Integration of catalog item metadata is delivered via a catalog feed, which synchronizes seller and item information. For detailed information, please see: catalog specifications.

User events

User events are the user/buyer activities that take place on a marketplace or retailer, such as browsing a page, adding an item to the cart, buying an item, and more. Moloco’s ML engine uses this first party of data to deliver relevant ads to the right user at the right time. This is accomplished by sending user event data to the Moloco Commerce Media Event service API. See this guide for detailed information on user event data specifications.

Inventory configurations

An inventory is a slot or a place to display ads. It can be placed on the home page, search result page, product detail page, and so on. Every page has its own characteristics, so the inventory should fit within these pages. Reach out to your Moloco representative for more information about inventory requirements. Once the inventory slots are decided, your Moloco representative will help you to register it to the Moloco Commerce Media platform.

Campaign settings

The campaign is a logical set of advertisement configurations nested in the ad account. Within the Campaign Manager tool, campaigns consist of the content to advertise, start and end date, daily budget, and max bid price for your bidding strategy. Typically retailers using the marketplace platform would create and manage campaigns within their own ad account(s), though Campaign Manager does provide other "roles" (e.g. Agency Role) to help to manage campaigns on the retailer's behalf.

Ad serving configurations

Ad serving configurations determine who will see the ads in a given ad inventory.

Moloco Commerce Media uses the inputs for the following purposes: to train the ML models for your marketplace, and to predict the ad items or banners. The table below shows what data it uses for ad prediction or ML training.

ML training

The model is initially trained from two types of data: items and their specs (i.e., catalog feed) and what user sequences (i.e., user events) are seen in your marketplace now. We require approximately one week to train the initial model. Model training is continuous when you send the required event stream data to Moloco Commerce Media. In addition to the constant training of the existing live model using event data, Moloco may introduce new model versions with different architectures, features, and training methodology - when new model versions are introduced in this way, Moloco ensures that these models pass validation and confidence criteria and will notify you before slowly rolling out to the existing system.

Ad Prediction

To return the right response for each ad request, Moloco Commerce Media needs three types of information: inventory configurations (where on your site or app the ads will appear), campaign settings (which campaigns are active and how they are configured), and ad serving configurations (which users you want to reach and in which contexts). Using these inputs, Moloco Commerce Media can select the appropriate sponsored items or banners, decide which campaigns to run, and determine which users see which ads in each section of your marketplace.

After our models have been fully trained on your marketplace data, you must call the Decision API to generate a list of ad items or banners with respective tracker URLs. These ad items or banners are displayed in the marketplace, and the tracker URLs record user clicks and impression data.

Tracker URLs

The trackers measure an ad campaign's effectiveness. Moloco Commerce Media has two types of trackers: an impression tracker and a click tracker. These are provided as URLs in the response to the ad decision request (Decision API call), and must be called when users have either seen (impression tracker) and/or clicked (click tracker) on the ad. See Ad Serving and Tracking for more information.