Naga ‘Preethish’ Pandurangi
Performance product manager, MiQ
The programmatic industry is constantly tasked with making better decisions to improve ROI. Optimizing inventory selection, so that you are delivering your message where your target audience is most likely to convert, is an obvious way to do it but is not as simple as it sounds. The challenge is that digital ad inventory is seemingly endless and supply is heavily fragmented. Inventory selection cannot be a simple, straightforward ‘if, then’ logic. Instead, campaign strategists need to work towards segmenting supply by value to improve decisioning–similar to high, medium, and low value audience segments.
Supply Path Optimization (SPO) automates inventory selection based on predetermined or intelligent logic. However, those who optimize inventory within a standard DSP set up often find that they have to break supply features into isolated ad groups to route investment to high-performing inventory combinations, which is difficult to manage. They end up with a bloated campaign architecture that requires constant manual interventions – essentially undoing any gained benefit and making it difficult to see the features’ true potential.
Simply put, SPO for programmatic campaigns is a way to simplify decision making, and boost ROI by trimming all the clutter in the supply path. It’s one of the first tactics a trader, the person who instructs the computer how to bid in the auctions, will deploy to improve campaign performance. This is typically reviewed within the first week of a campaign’s go-live date, referred to as a ‘test-and-learn’ period. But, with a standard DSP setup, this manual process can be extremely time consuming.
At MiQ, we take a data-driven approach to evolve industry capabilities into custom opportunities for our clients. We wanted to sharpen supply path decisioning, so we looked to build a prioritization strategy that would allow us to control impression delivery using a custom logic that preferred specific inventory groupings – underlying performance metrics that brought the most value to our clients.
To gain better visibility into the digital supply chain we armed ourselves with log-level data, the details tied to each impression, using an enterprise feed from The Trade Desk, a programmatic demand side platform. We then applied our data science methodologies to analyze the contextual data (Exchange, Seller/Publisher, Site Domain, Deal ID) from every single impression to see exactly what was occurring in each transaction, in every possible feature combination. We then developed a custom scoring logic, which weighed a combination of primary and secondary goals, such as CVR with CPA, to waterfall spend across each cluster of feature combinations to create a compounded efficiency score for the contextual path. This score was then used to direct the volume of ad impressions towards the best results, essentially creating a logic for a dual KPI value-based supply segmentation (see Table A).
Table A: Feature combinations are first assigned a custom metric score (a weighted combination of the primary & secondary KPI performance), which then involves segmenting supply paths into four segments and using machine learning to assign priority (Table A). The segment with the highest average custom metric value will be assigned to Segment 2 and Segment with least average custom metric value will be assigned to Segment -1.
|| What does it mean?
Prioritize buying impressions on this vector before spending budget anywhere else.
If the budget remains and there is no more scale with my first choice, prioritize this option instead.
Any impression that does not match an Inventory Control list will be defined as “neutral” by default.
||The last resort
If the budget still remains, buy impressions here to deliver in full.
With this method we’re able to drive better outcomes against multiple custom KPIs at scale without having to manually create additional ad groups, or adjust budget allocation – at a level of constant complexity beyond what traders could manually analyze, freeing up traders’ time to go deeper on broader client challenges. We saw this feature outperform campaign performance goals and standard optimizations in nearly every campaign executed:
- MiQ’s Inventory Control SPO feature was initially tested on 11 performance campaigns with an average CPA reduction of 24.5%. This solution is now running on more than 200 active monthly omnichannel campaigns.
- When MiQ’s VC was implemented, the average CVR increased by 39%.
- Average conversion volume also increased with 14.3% more conversions after implementing.
- Conversion volumes increased and CPA decreased on 10 out of 11 beta campaigns (a 91% improvement rate).
Brands need an SPO strategy in order to make better choices and improve ROI with programmatic advertising. MiQ’S Inventory Control SPO drives more value from inventory to achieve the best results for your brand. Contact us to learn more.
To find out more about MiQ:
- Check out our open positions
- Check out our previous post on measuring cross-channel reach across YouTube and linear TV
Preethish is a product manager for performance at MiQ. Based in our Bangalore office, he enjoys spending time outside of work reading up on all things historical.