A few months back, we hosted our NYC Optimize conference. Optimize is geared toward the folks who put the tech in ad tech – engineers, data scientists, product managers, and the like. It’s a great opportunity for technologists from our industry’s leading companies to meet and share knowledge with one another.
But if you’re not a tech whiz yourself, don’t worry – Optimize has something for you too: our Break Down the Breakouts video series.
A key part of Optimize was the AppNexian-led breakout discussions centered on the cutting edge ad tech products and strategies helping our clients succeed. In order to spread the knowledge beyond just the day’s attendees, we asked those AppNexians to turn their breakout sessions into short, digestible whiteboard videos that make these complex ad tech concepts accessible to all. Below, you can watch three of our Break Down the Breakout videos, covering topics like augmented line items, click prediction, log-level data, and more. Enjoy!
The Algorithms that Power Augmented Line Items
One of the biggest value-adds our Console for Buyers offers advertisers is campaign optimization that “just works.” By that, we mean that we give buyers the ability to input two simple figures – a campaign’s budget and its KPIs – and let our system run the campaign’s bidding strategy from there. It’s the rare tool that gives you both quality and quantity, in that it allows advertisers to run more campaigns more effectively with less work.
It sounds too good to be true, so how do we do it? At a high level, our system looks at the campaign budget and KPIs, then translates them into a strategy comprised of three parts:
- Bidding: how often the buyer bids
- Valuation: the amount the buyer bids
- Allocation: what inventory the buyer bids on
Console for Buyers determines the first iteration of bidding strategy based on its initial assumptions, implements it with a small piece of the campaign budget, then optimizes as the campaign continues. Watch the video to learn the nitty gritty details from Senior Engineer Brian Wu and find out exactly which systems and applications are involved!
Using Genie to Customize Click Predictions
Cost per click (CPC) is one of the most important metrics advertisers try to optimize their campaigns around. To help them do that, AppNexus Programmable Bidder (APB) v8 Genie – the engine that powers our custom optimization API – generates an Expected Value for each impression to determine how much they should pay. That Expected Value is made up of two components: Expected Value of a Click multiplied by the Probability of a Click.
The buyer provides their own Expected Value of a Click depending on their internal data around pricing and conversions, while Genie-powered models determine the probability of a click. AppNexus is able to do this by tapping into all the buy and sell-side transaction data our platform collects, then running it through a logistic regression model. Don’t fret if that last part has you scratching your head – you can watch the full video to see Senior Data Scientist Yana Volkovich diagram how a logistic regression model is built. You’ll also see two of our engineers – Noah Stebbins and Ankit Shah – talk about how we refine our in-house click prediction models.
Advanced Data Analytics and Log-level Data
Our Client Insight Analytics team works with our biggest enterprise clients to help them make better use of their data. And lately, those clients have been more and more interested in log-level data. Log-level data is data gathered in real time from the moment a programmatic advertisement appears on a user’s screen. It’s incredibly granular, and when used properly can help marketers tie ROI back to every single impression in a given campaign.
Many advertisers are a bit intimidated by log-level data and not entirely sure how to get started using it. But as Client Insight Analytics Director Travis Stowe explains in this video, log-level data is actually pretty easy for analysts to use once they get it in the right place.
Stowe also lists out a few golden rules for advertisers who want to use their data more effectively, so be sure to watch and learn some valuable, quick-hitting analytics tips!
Subscribe to our YouTube channel to get notified whenever we release educational videos!