Documentation

  All Documents

Demystifying Tech Stack & How FuelX Built Theirs
Technical

A Tech Stack is a series of components that make up the different parts of a technology program and enables it to spread out into wider capabilities. There are many benefits to having an internal tech stack, but it can be capital and labor intensive to both build and maintain. In order for a company to be considered a digital platform, it must have a multifaceted approach to solving an issue and aggregating different tech components. It must also serve multiple clients with the same technology as opposed to being custom built for each user. In the case of performance advertising, it is imperative to control all facets from data, quality control, and ad distribution.

FUELX TECH STACK

FuelX has built their tech stack to become one of the few performance advertising platforms that is an end to end solution. The technology offered includes all components of the programmatic landscape with some evolved aspects such as: » Data Management Platform (DMP) » Demand Side Platform (DSP) » Ad Server » Bidder » Machine Learning (ML) Other Components of a Tech Stack » Supply Side Platform (SSP)

Data Management Platform

The Data Management Platform, referred simply as a DMP, is a repository of data that collects, stores, and indexes anonymized proprietary user data. The DMP is the backbone of advertising within the digital and programmatic world. This platform allows marketers to collect data such as behavioral interests, buying affinities, demographic information, device data, and much more. With the use of this information an advertiser benefits from real time data collection and distribution to customize an end user’s advertising experience. The process of data collection begins with ingesting, storing, and analyzing both first and third party data from consumers to create user profiles for ideal targeted campaigns. Lastly, sends it to the DSP for buying information.

5 ASPECTS THAT A PROMINENT DMP CAN DO

Data Collection Collects data from both first and third parties and places a tag on on the site which then helps back the information that is being sent to the DSP. Data Classification Once the data is collected, the DMP categorizes and organizes the information. The DMP organizes the data depending on the business itself. This data classification helps create better and more efficient consumer profiles. Data Analysis The DMP then analyzes the data it has collected within the relevant categories. It analyzes the user’s previous clicks, purchases, individual preferences and likeability to click or react positively to a particular ad. Data Transfer Once the DMP has collected, sorted, and analyzed the data, it sends it out to other parts of the digital advertising world, such as DSPs, p3ortals, and trading desks. Data Scalability The DMP should store millions of data points into one area. Some examples of the data points it should be able to store are: First party and Third party data, Customer Relationships Management (CMR) data, and Audience data.

Demand Side Platform

The Demand Side Platform, abbreviated as DSP, is an essential component that is used to purchase advertising/impressions from ad exchanges via Real Time Bidding (RTB)1 with the data provided by the DMP. The DSP buys, tracks, and serves ads all in one place and purchases the impressions all in the time it takes a web page to load (less than 3 milliseconds). DSPs do not purchase or sell impressions directly from publishers, but are connected to publishers through ad exchanges or Supply Side Platforms (SSP). The DSP is also the program that businesses use to set their buying parameters, such as demographic or psychographic preferences. When understanding the FuelX DSP, it is important to know that it is an evolved distribution platform because it aggregates six different exchanges and networks that include Google AdX, AdMob, TripleLift, Oath, Facebook, and Instagram. FuelX has taken this consortium of networks one step further by mapping each user and advertising IDs to its own mapping function. This has unified all of their 6 inventory sources and their variant ad formats (including video, social, and static) under one data driven access point.

Ad Server

An Ad server is a web based tool that is used by both advertisers and publishers that aids, serves, and tracks both ad management and campaign management. The Ad server sells ads and reports on optimal performance through an Ad tag, HTML or Javascript code that is placed on the Web page’s source code to help the ad server record efficient ad campaigns. Before an ad is served, the Ad Server goes through a number of steps to make sure the consumer will be receiving the right impression at the right time. The ad serving technique depends on certain elements such as, demographic and behavioral targeting.

AD SERVER PROCESS

A consumer visits a publisher’s website, which sends back a code/Ad tag, telling the browser which server to get the content from and what format the content should be in. As the page loads, in real time, the ad slots send requests to the ad server. Before serving the ad, the Ad server uses the collected data to determine which ad would best be served for the highest chance of conversion.

Machine Learning

Machine learning is a field in computer science that is a subset of AI that uses statistical analysis to give computers the ability to learn heuristically using structured and unstructured data without being completely programmatic or automated. This means that is allowed variance in order to get a better understanding of the outcomes. Machine learning is an evolution of pattern recognition and learning through a series of computations that not only structure algorithms but create new ones from learnings. Because it uses computational statistics it focuses on being a way to optimize and predict outcomes. One of the most important factors of machine learning is to understand your weighting and being accurate with your predictive analysis.

SELF LEARNING MODELS

At FuelX, Machine Learning is used to deliver unparalleled performance that focus on driving incremental revenue, higher conversion rates, and increasing average order values. » Audience trends » Targeted cluster nuances » Publisher to user evaluations to black and white list sites » Bid magnifiers on high intent users » Viewability and fraud detection » Post ad serve we look at engagement with the brand

Supply Side Platform

The Supply Side Platform, abbreviated as SSP, is a software that allows publishers to connect their inventory with a multitude of networks, ad exchanges, and DSPs all at once. Since the SSP has the ability to connect with many networks, ad exchanges, and DSPs, a publisher has a higher chance of maximizing the price of their inventory. Once the SSP is included in FuelX’s tech stack, it will not only assist the DSP, but also help control inventory costs. The inclusion of the SSP will only further strengthen FuelX’s components and optimize the performance of our system.

Bidder

The bidder is an extension of the Ad server that finds placements to communicate between the server and DMP for the business/advertiser, using a protocol called RTB. RTB is the process of buying and selling impressions within milliseconds which depends on the buying protocol (see example below) to place the winning bidder at the right place at the right time. Before a bid is placed, FuelX uses its indexed and analyzed data to locate the users that display the highest intent to convert and then places the bid in real time. FuelX has mapped users using cross-device in conjunction with behavioral understanding. The data is used to determine if a bid is up to date and is refreshed every 24 hours to ensure a higher chance of conversion.

THREE KEY COMPONENTS FOR RTB

DSP Automates online purchases on behalf of the advertisers then uses the information from the DSP to set the parameters while also monitoring the performance of companies. Publisher Provides inventory and sends it to the Ad Exchange. Ad exchange Connects advertisers and publishers through auctions for an optimum and efficient advertising exchange.

RTB PROTOCOLS (1.1)

First price The winning bidder pays exactly what they bid. This bid maximizes Revenue Potential. Second price The winning bidder pays $0.001 above the second highest bidder. In this auction it is better for the bidder to pay the highest they are willing to bid. Header bidding Also known as Advanced bidding or Pre-bidding, is a protocol where a code is placed on the web page header, allowing a limited number of advertisers to bid on reserved inventory outside of the primary ad server. This protocol creates an auction before the final auction in a publisher’s ad server. Private Marketplace Invite only auctions between a publisher and a buyer on specific inventory that runs programmatically. This Builds relationships between buyer and publisher allowing trust and transparency while optimizing programmatic sales. First Look Inventory Allows preferred access to buyers who meet the floor price and gives them the opportunity to buy an impression ahead of a publisher’s reserved inventory. Gives the publisher more opportunity to sell to a wider audience and the bidders to bid on impressions earlier.

BIDDER PROCESS

Publisher provides the inventory to the ad exchange (depending on the parameters) who is in charge of holding an online auction. The DSPs place a bid on each impression presented within the auction. The highest bid wins and the ad is served on the page. All of this takes place within the millisecond it takes for the consumer’s webpage to load.

Previous Next