What Does Better Look-Alike Look Like?
The scarcest resource in the marketing technology ecosystem is quality data. Data is truly the ‘oil’ or ‘gold’ of the programmatic revolution. Brands have small amounts of first party data, which is rarely scalable, ample amounts of brokered third party data which is too scalable, and almost no ‘second party data’, the grassroots data version of a book exchange – take a book, leave a book. However, despite a scarcity in data, ad tech shows no signs of tempering the programmatic sales strategy. As each subsequent prediction ‘ups’ the growth of programmatic to full steam, the pressure only mounts for providers to fulfill big promises of big data. Look-alike modeling is often the solution. It promises to analyze, mimic and replicate the characteristics a true data set presents. As look-alike modeling is required to fulfill more and more data requirements, the process will need further optimization and alternative data sources to provide long-term ROI. Location-based strategies offer a possible solution and should be strongly considered as an addition to look-alike modeling.
We are beginning to see more location-derived modeling strategies developed for the ‘probabilistic’ world. With this approach (often in device matching), technology providers seek to identify a cell phone for example, to match a laptop. As the cell phone ‘checks in’ more often than not with the same laptop, the technology deems the devices ‘matched’ and enables advertising deployment cross device. A good percentage of mobile ad targeting is beginning to take place in this fashion.
Look-alike modeling requires more sophistication than ‘cross-device’, but can take a page from the simple success cross-device matching has seen with geo data. Look-alike modeling requires ‘matching’ across multiple dynamic characteristics of randomized audiences, requiring algorithms to sift through immeasurable amounts of data to find hopeful patterns. The general imperative is to build scalable models from finite amounts of data. As you can imagine, the success in this approach is mixed, but Locomizer offers a solution which changes the odds for successful ‘look-alike’ application. Locomizer can identify GPS coordinates in real-time and work with historic GPS coordinates to better understand an audience’s relationship to time and place. With Biological Intelligence, Locomizer can then better define the characteristics of the users being matched for look-alike purposes, by the users around them and the attributes of the places they frequent. BI presents a natural model to provide incredible insights based simply on the movement of audiences through space and time. This provides an opportunity to significantly optimize traditional look-alike modeling by adding a geo-spatial dimension to the archaic, linear approaches most common today.
If all data were first party data, marketing tech wouldn’t need much of an algorithmic approach. However, demands for data will always outpace the supply. In essence, it’s no different than oil – as in our analogy, data is the ‘oil’ of our industry. While some oil is easy to come by, the only way to obtain it cost effectively as demand increases is to improve the technology. And breakthroughs such as horizontal drilling have been game-changers for the energy industry. As demand continues to outpace the cheap supply of data for programmatic use, look-alike practices will become increasingly important to mine actionable data from large quantities of the unusable. Locomizer is that game-changing technology that will help take the process to the next level and make better data more available for all. By adding the locational dimension augmented by Biological Intelligence, Locomizer vastly improves the old model. We are all a product of where we are and who we are with. And with this component increasing the probability of accurate targeting, geo-profiling data will help ensure look-alike data looks great across all programmatic channels, as its demands continue to scale.