Portfolio optimization for programmatic advertising
The term “portfolio optimization” originates from the world of investing and asset management. The simplified explanation for portfolio optimization is to extract the highest possible return while minimizing risks across various assets.
The same concept and approach are directly applicable to marketing. At any given time, there’s a variety of activities running across multiple channels, split into campaigns consisting of numerous audiences or tactics, each with a dedicated budget, bid, and creatives.
One doesn’t need to be a rocket scientist to see that this is a handful to manage, let alone manually.
Enter portfolio optimization
The benefit of portfolio optimization is closely related to managing budgets and ensuring that the best performing activities are given as much weight as possible. For portfolio optimization to yield optimal results, the granularity of the campaign setup plays a crucial role.
Well, how does one go about applying portfolio optimization to (digital) marketing, then? The answer is relatively simple. However, it all boils down to structure, categorization, and distinguishing the variables contributing to the bottom line.
What are the significant elements in your digital marketing campaign that affect the outcome from a broader perspective? For the sake of simplicity, we’ve outlined the most popular ones:
- Objectives & KPI’s
- Product categories and products
- Audiences & targeting
- Environments, devices & placements
- Creatives & messaging
Going through these variables, mapping their priorities, and determining whether these variables could be split into independent and isolated activities will start to form the foundation for portfolio optimization.
To provide a tangible and actionable example, let’s look at this from the perspective of programmatic advertising and how you could build a “portfolio” within a programmatic campaign.
Objectives & KPI’s
Perhaps the most straightforward and highest-level approach for categorizing and structuring a campaign would be to separate insertion orders and line items based on their primary objective and KPI.
A campaign’s overall goal is to increase sales, but there may be activities targeting prospective customers in different stages of the purchase process.
Measuring all of these activities based on direct conversions or purchases might not be feasible. However, separating the campaign based on proxy metrics or attribution modeling could provide better results.
For example, a campaign could be structured as follows:
- Activities focused on generating awareness and consideration, optimized towards viewable impressions
- Activities concentrated on engaging prospects and driving traffic, optimized toward clicks or driving quality traffic
- Activities focused on (re)-activating users further in the purchase process, optimized towards conversions or revenue.
- Activities aiming at cross-or upselling additional or adjacent products and services
Each category and the activities within would form an individual “sub-portfolio” that should be optimized towards a relevant KPI.
Products & product categories
Retailers and e-commerce-advertisers often have a large inventory, consisting of multiple categories, brands, and applications, among others. To mitigate risks and ensure optimal performance, aligning campaign setups with your inventory is one of the most powerful approaches.
To account for seasonality, changing trends, fluctuating prices, and competitors’ actions, building a robust line item portfolio allows you to tailor tactics and focus optimization to smaller chunks, rather than spending copious amounts of time to analyze which variable is affecting performance.
To give a concrete example, a furniture retailer could structure their campaign’s line items as follows:
- Kitchen & appliances
- Living room & dining
- Utility room
- Garden & balcony
Audiences & targeting
A campaign usually consists of several targeting methods. Targeting could be based on context, geography, demography, interests, intent, or a combination of multiple elements.
A universal recommendation would be to refrain from using multiple AND-clauses, which complicates analysis and distinguishes which variable is most dominant and drives the most results.
Mixing signals from different stages of the purchase process would also not be advisable. Instead, we recommend separating interest and intent signals to make sure you’re aware of which targeting influences performance the most.
For example, separating line items by a larger group such as geography or demography and layering other fundamental targeting rules is generally reasonable.
Environments, devices & placements
Campaign performance often varies significantly between mobile apps and websites, on different devices, and between different inventories and placements. To determine which variable has the most effect on performance, splitting by device, ad environment, or even by publisher could provide valuable insights and performance gains.
If you have a well-defined target audience and a good set of targeting rules, expanding your portfolio by separating line items based on the categories mentioned above sheds more light on how and where your campaigns are delivering at peak performance.
Creatives & messaging
As the final and arguably the most critical part of your campaign are the creatives. You might have your audience strategy fully on lock, your bids and inventories carefully researched, and the final step is to deliver your message to your prospective customers effectively.
From a portfolio optimization point-of-view, creatives provide an excellent foundation and present an amazing opportunity. The simplest categorizations are naturally the format of your creatives. Splitting display ads, videos, native ads, or audio to independent line items is a practice no one can argue against.
It is possible and, on some occasions, favorable to split line items based on the creatives’ size or length as well for more accurate bidding and improved performance. However, several DSPs offer the opportunity to control these variables using bid multipliers within the line item.
Combined with rotators or dynamic creatives, your DSP and ad server can automatically serve the best-performing versions of your creatives.
Portfolio optimization is a potent approach and tactic for programmatic advertising, which will simultaneously distribute the risk of underperforming to a broader set of variables, allowing you, the buyer, to quickly distinguish where potential problems lie and which activities contribute most towards your objectives.
Admittedly, building and maintaining a highly robust and granular campaign does take a bit more of your time. The improved structure and clarity will allow you to perform better and notice potential problems more quickly, especially with always-on campaigns.