There’s no question that 2018 saw its share of data mishaps across the adtech industry. But the fact remains that, as machine learning and artificial intelligence continue to become an integral tool to analysing data, data leakage is inevitable.
This is precisely why we expect brands to continue to take a "build your own algorithm" approach in the coming year. It allows them to own not only the data they input but also the resulting learnings. This is especially the case when it comes to traditional, one-size-fits-all demand-side platforms that rely on one-size-fits-all algorithms; the learnings from Coca-Cola, for example, would help inform the same strategy used by Pepsi – should the two use the same DSP.
Additionally, as programmatic continues to become the primary method for transacting digital media, the key to making it a scalable revenue channel lies in tailoring strategies to the unique objectives of your unique business. Only then can you truly understand, evaluate and bid on the variables or attributes of an impression at the price that makes sense to you. This is achievable through multiple optimisation methods, but it becomes most scalable through the building of custom optimisation models designed to achieve your own business objectives.
Imagine Coke and Pepsi were both using the same DSP. They would be competing solely on price – using the same algorithm, with access to the same second- and third-party data, and access to the same inventory, they are going to come head to head on every impression, and the only way to deliver the campaign is to compete on price. They wouldn't have access to unique inventory or optimisation models fine-tuned to meet the needs of their business.
Smart advertisers that understand the impact of programmatic are turning to platforms that enable them to differentiate and up-level their media buying through the use of custom optimisation models. Instead of fighting head to head with competitors on every impression, they fuel their buying with the unique intelligence that they have on their consumers and prioritise the inventory and audiences that are most valuable to them.
Optimise on things that are important to you
Traditional DSPs enable brands to optimise to a generic performance goal (eg viewability, click-through rate or cost per acquisition). However, this rarely achieves the KPIs that are most useful to the brand. Some brands, for example, have a specific way to measure viewability or they optimise to reach people who spend a particular amount of time on their site. Whatever it is, building an algorithm allows control of the data points that are taken into account during the optimisation process – so you can build your custom funnel into your algorithm.
Reduce media costs
With the rise of first-price auctions and other innovations in bidding rules that the sell side has been introduced, buyers can sometimes find themselves paying more than they need to for media. Smart buyers, therefore, are requiring more access to log files so that they can better understand the bid landscape – ie where they are winning or losing bids and by how much. This granular level of data used to inform bidding algorithms significantly reduces the cost of media, while avoiding a reduction in the quality of the media or the audience reach.
Utilise your unique assets
If a brand has unique assets, it may struggle to activate them in a standard DSP or be reluctant to expose it within shared infrastructures. First-party data is becoming increasingly more powerful as a competitive differentiator and, when used properly in programmatic, can have a significant impact on the performance of campaigns.
While most campaign needs may be met by one-size-fits-all DSPs, it’s clear that not all media buyers are the same. As more and more buyers expand their reliance on programmatic, they are also increasingly looking for technology that they can customise to suit their own needs.
So 2019 is the year for building your own algorithm – but with data security at the top of the list.
Cadi Jones is commercial director EMEA at real-time bidder platform Beeswax