How do you optimize RTB data for cross-channel attribution?

I'd say that, from an RTB buyer's perspective, Bid Price is not the part of the RTB data (input) but is an output. It is a "Y“ and not one of the “X“s. I.e., the buyers' algorithms look at the data that is available to them, such as user location, user audience segment(s), user browser, user device, publisher site / app, contextual category, etc. Then, based on that data, the algorithms decide whether to bid and at what price. Secondly, I belive that saying that RTB data such as listed above, can be used for measuring the effectiveness of ad campaigns is a big stretch. To measure campaign effectiveness, one needs identifiers, which tie the ad impression or click (the thing you pay for) to the desired results, such as a conversion.

Data integration is important because different sources use different technologies to measure the same User behaviour , which is conversion . It can be a View ,Install , etc. With so many indepandant technologies, deriving information are limitless , a DMP can help establish Geo data , Contextualizers can categorise , Pixels can track Upper funnel , API can help track deep funnel cross platform etc therefore combining all of this can help understand the lifecycle of an impression . Different details can be valuable to different Advertisers depending on their spending goals , therefore integration here plays a key role to help maximise the output by bringing all the necessary information in front of the right audience in the right format.

From the Programmatic Buyer's perspective, Data integration usually means transferring the data from its sources (Data Providers, Data Exchanges, the buyer's 1st party data that typically resides in the buyer's DMP, etc. to the places where imortsnt decisions are made, such as the buyer's DSP. Sellers can also use data to make decisions on how to price (floor) their inventory in real-time. That is another important decision to be made. One rarely speaks about "combining data from various SSPs, because most buyers use only one DSP to make their buying decisions.

Common identifiers do not necessarily improve audience data quality. One could have the most accurate identity solution (for example, when everyone logger in to the site), but also have poor quality audience data. For example, you can perfectly identify a particular group of users, but the users might not share thir attributes, such as income or age, correctly with you. The lack of data quality is a big conundrum for the industry. The data providers get paid in proportion to the volume (number of impressions that can be targeted with the data). So they have strong motivation to increase the volume while sacrificing thr quality. The hen is guarding the chicken coup. There is an elegant Machine Learning solution to check data quality.