Real-Time Real Talk

Real-Time Real Talk: Programmable Splits

When we built the AppNexus Programmable Platform (APP), we had a simple goal in mind: We wanted to build a DSP that could free up traders’ time, let them focus on the strategic rather than the tactical, and help them achieve consistently higher campaign performance.

Blog Post
The AppNexus Team

The secret ingredient to make that possible? Machine learning. We built APP around powerful machine learning algorithms that can adjust campaign spend in flight, analyze inventory performance in real time, and automatically achieve traders’ goals around KPIs like CPC, CPM, vCPM, and the like.

But traders have made it clear to us that alone isn’t enough. After all, digital advertising isn’t just about serving X ads to Y users at Z price. Just as important are the harder-to-measure “soft goals” brands have – they want to hit users in specific locations, or who fit specific demographics, and test different creative messaging on different audiences. 

In other words, traders need more from their DSP than just simplicity and performance. They need customizability as well.

Previously, any kind of campaign customization has required traders to build more line items, which they would then have to monitor closely to ensure none compromised overall campaign performance. With APP, we’re pioneering a simpler way to accomplish soft goals: Programmable Splits.


What are Programmable Splits?

In a nutshell, Programmable Splits are a straightforward way for traders to ask a line item to “do more” than just hit a certain KPI goal or budget number. So, let’s say a trader has set up a line item as they normally would, with a CPC goal and a set budget. With Splits, they would have the opportunity to further modify the line item’s behavior from a simple UI. Check out the screenshot below:

AppNexus Programmable Splits screenshot

In this case, the trader has decided that they want to make sure the line item spends some of its budget on users aged 18-34, some on inventory, and some on users browsing in Boston. Ordinarily, this kind of targeting would mean building more line items – or possibly even building custom algorithms, depending on the complexity of the campaign. But with Programmable Splits, it’s as simple as listing off a few criteria within the original line item. The trader then chooses what percentage of the line item’s total budget should go to each Split under the “Allocation” column.

But what if a trader chooses a Split that compromises the line item’s ability to meet its delivery goal? Take a look at the column titled “Cap.” That’s where traders can tell APP how it should balance delivery vs. Split allocation. If they mark a Split as “Uncapped,” then APP will spend elsewhere to hit the delivery goal, even if it means ignoring the Split. If the trader marks the Split as “Capped,” then APP will spend the amount allocated to the Split, even if it means the line item as a whole under delivers.

Similarly, what if a trader chooses a Split that hurts the line item’s ability to hit its KPI goal? In that scenario, APP would spend on higher-performing inventory to meet the overall KPI goal rather than the Split’s allocation goal, as KPI goals are almost always higher priority. However, this would still be a valuable learning experience for the trader, as they would now know that the audience or inventory segment they chose to target doesn’t work for this campaign. But if they want to, traders can opt to favor their Split’s allocation goal by turning off APP’s automatic optimization setting.

Splits aren’t just limited to telling the line item how to allocate spend. You can also use Splits to determine bids on specific pockets of inventory, allowing you to granularly control how you optimize each Split. You can also select different creative assets to deliver to each Split, allowing you test which messaging most resonates with the key segments of each brand’s audience.


What makes Programmable Splits so valuable?

One of our driving goals with APP was to free traders from the repetitive tasks associated with programmatic campaigns – line item setup, manual optimization, budget pacing adjustments – and instead let them focus on providing their clients with higher-level strategic insights. Programmable Splits make the second half of that sentence possible. While our built-in optimization algorithms handle the tactical nuts-and-bolts of a campaign, Splits are where traders get to apply their expertise and experiment with different messaging, audience segments, and inventory choices.  

We believe that Programmable Splits are why APP is better aligned with the true goals of an advertising campaign than any other DSP. Programmable advertising becomes less productive when traders are too narrowly focused on KPI goals. Advertising is more about getting to know an audience and figuring out what messaging will bring them closer to a brand. Programmable Splits make that process simple and intuitive for traders. 

Keep an eye on our blog for more posts on Splits, including guides to different use cases like prospecting and retargeting. If you’d like to learn more about what Programmable Splits can do for your next campaign, please contact us here.