Running effective campaigns that deliver results in a native demand side platform (nDSP) requires a sophisticated optimization strategy and a constant focus on innovation. The digital content and advertising environment can be crowded, so brand advertisers need to be confident that their media buys are relevant, precisely targeted and driving meaningful results for their business. For this reason, improving and advancing optimization algorithms and technology is a core focus at Bidtellect and has always been part of the company’s DNA, beginning with its roots at Advertising.com with Co-Founder John Ferber.
We are currently preparing to release the 4th generation of our optimization technology. This technology utilizes big data, predictive modeling and machine learning to intelligently analyze and bid on billions of impression opportunities every day.
There are a variety of objectives that campaigns are measured against, and each advertiser and each campaign define success differently. Through everything mentioned above, Bidtellect’s optimization platform is working throughout the life of a campaign to ensure it is maximizing the results of the chosen KPI. A unique advantage of Bidtellect’s technology is its ability to optimize toward multiple KPIs simultaneously.
While we offer over 10 KPIs for a buyer to optimize towards, we see tremendous value in the post-click metrics. For example, Engagement Score, bounce rate, time on site, etc. Post-click activity will become increasingly critical for marketers as they seek more meaningful ways to determine how consumers are interacting with their content and brand.
According to comScore, Bidtellect reaches nearly 75% of the entire US market on over 10 million placements via real-time bidding. At scale, optimization technology is vitally important in considering not only where to buy, but when to do so and for how much. Machine Learning is key to the estimation of the KPIs and is where predictions meet learned results to continually refine a campaign from the day it is launched. It is typical and expected to see KPIs improve steadily throughout a flight.
The below process summarizes the workflow:
Step 1: Receive Impression Request: We see a constant stream of supply of billions of opportunities each day that we are able to bid into.
Step 2: Scrub Request: Begin process of analyzing the impression opportunity. Eliminate fraudulent inventory and ensure we understand the validity of the supply.
Step 3: Filter Ads by Inventory Targeting: Match our advertisements and campaign requirements to the supply source and see if there is a match.
Step 4: Run Pacer: Determine if there is enough money in the campaign to buy that opportunity. Helps ensure proper pacing and smooth delivery of campaign.
Step 5: Predict KPIs: By employing advanced machine learning methods we are able to predict KPIs for an individual request using a large number of features of the user, page, and advertisement.
Step 6: Calculate Valuation Factors: Given the KPI predictions, determine if the impression aligns with an advertiser’s goals.
Step 7: Calculate Bid: Given the campaign’s base bid price, combine the valuation factor and base bid to obtain a final bid price expressed in effective CPM.
Step 8: Ad Selection: Process of comparing effective CPM of multiple ads that you can possibly run there and determine the logical choice.
Step 9: Bid: Place a bid on behalf of the campaign with the best predicted outcome.
Although it is an extensive decision making process, all of these steps are actually happening hundreds of thousands of times per second, which is key in the always-on real-time world that we live in. An individual transaction occurs start to end within 100 milliseconds. To compete you have to have a fast, smart and efficient optimization platform otherwise you will lose out on the opportunity.