Skip to main content

Overview

DataMilk provides an API to allow companies to utilize AI algorithms in order to optimize their ecommerce websites to improve user experience and get effective results such as higher Average Order Value (AOV), Conversion Rate, and Revenue.

The DataMilk Attention Data Script automatically collects user behavior and website content which feeds into the DataMilk AI Platform to train machine learning models that power DataMilk's privacy first AI algorithms.

The DataMilk Attention Data Script actively tracks and records the active content of your site so that integration of back office content management systems with DataMilk is NOT required.

This User Manual is subject to change as the API is being enhanced on a regular b

Privacy First​

DataMilk takes a shopper’s privacy seriously. Data is segregated from other websites and no personal identifiable information is used or stored. General usage patterns are anonymously used to train machine learning models in order to assist the shopper and the ecommerce website to achieve their goals.

Content​

Shoppers visit ecommerce websites looking for a certain product or browsing to see what might interest them and eventually purchase.

Existing Page Content​

Ecommerce websites provide content for their shoppers in the form of Product Listing Pages (PLPs), Product Detail Pages (PDPs), or Search Terms (aka Search Result Pages). These pages allow the shopper to find and purchase what they want. DataMilk provides algorithms to surface and highlight content that may interest the shopper based on the shopper’s activity, predicted intent, historical activity of similar shoppers, and potential of the content to increase revenue.

Custom Content​

In addition to content already on a website, the DataMilk platform can also serve content which is created in house to increase revenue. Such content could include button labels, website banners, and other custom content developers wish to optimize on a site.

All content can be served to the user in UX friendly and effective ways such as Navigation Bars, Related Searches, Buttons, Banners, and other UX elements.

Algorithms​

DataMilk provides various AI algorithms to choose from that can be configured to achieve a company's business goals. Algorithms in the system are designed to choose content in order to maximize different goals such as increased Average Order Value (AOV), Conversion Rate, or Expected Value (EV) (Combination of Average Order Value and Conversion Rate). Additional new algorithms are added on a regular basis.

Personalized Value Ranking (PVR)​

DataMilk’s proprietary algorithm to predict what shoppers will do next and to rank options that are most beneficial to shopper and business goals.

As stated in the name there are two components to PVR: Personalization and Value. The Personalization part takes into account the interests of the specific shopper based on their journey on the website. The Value component indicates what is the value that each page can provide to the desired goal and takes into account the number of people who visited a page, their conversion rate pattern and how much revenue a page can contribute to the desired goal. These two parts are combined to optimize both the shopper's experience and the value for the business at the same time.

AI Auto Select (AIAutoSelect)​

Often it is useful to pick one option to show a user which has the highest likelihood of achieving the optimization goal. DataMilk provides a Multi-armed Bandit algorithm that can automatically determine this choice out of a set of choices. This Bandit algorithm works with a reward. I.e. it needs to know when the desired result was achieved. This desired result can be more engagement with the content (e.g. user clicking on content), or the content resulting in a user adding a product to their cart (e.g. click ad to cart button), or any other reward that the developer wants to optimize.

Use Cases

  • Product page - Which picture should be the first to be displayed based on measurement with a reward defined as “Add To Cart button was clicked”.
  • Button - Which label, or which color to use measured with a reward of “button was clicked”. Hero Banner - Which banner to show measured with a reward indicating if banner was clicked.

Note: The developer can specify the custom content in any way they wish by encoding it as JSON, HTML, or other text format.

Engagement Data Counters​

DataMilk provides engagement counters which can be used to provide signals to users and motivate action to achieve business goals. These counters include views,and conversion events. Other event types will be possible in the future depending on the site such as likes, add to cart events, or checkouts.

These counters can be used to generate UX elements that show social activity on any page in order to increase action by the user. For example product pages can indicate how many shoppers are looking at the product or how many have added the product to their carts. This information can also be used in combination of PVR pages to display a navigation bar that shows Product Listing Pages (PLP) and display the number of shoppers visiting those PLPs to show the shopper how popular they are.

Predicted Top Sellers​

The DataMilk platform can provide predictions of top selling products using machine learning methods that do not require integration with an inventory system. This information can be communicated to users to nudge a desired outcome. Examples of nudges include “Selling Fast”, “Limited Supply”, and “Popular Purchase”.

Data Cleaner​

Commerce websites can suffer from tainted data which results in inaccurate performance analysis and subsequent actions taken. Tainted data originates from different sources such as bots, fraud, repeat orders, automated end to end tests, and outlier transactions that are not typical to a website. DataMilk provides a data cleaner api that allows the developer to identify these sessions or purchase transactions in order to exclude them from analytics for more accurate analysis and actionability.