Supply Path Optimization in Video Advertising Landscape

With people moving away from traditional TV and the fact that almost two-thirds of households have a Smart TV, e-Marketers predict that over 80% of online ads will be video in the next couple of years. It’s a smart move by advertisers to slowly shift advertising budgets from display to video advertising as months pass. However, with many players such as DSP, SSP, publisher, re-seller, and heaps of third-party vendors, the auction duplication deteriorates and results in a multiplexed path. In brief, it can be said that a single impression is being sold by numerous sellers at various prices in multiple paths. On average, an impression is available through 19.8 distinct supply paths. A wise path for a buyer is the one with a low CPM (Cost per 1000 impressions) to obtain a better ROI. The question: Which route/path should the buyer choose and what should be the fair price to pay? has been staying put for years now. This can be answered with supply path optimization.

The goal here is to find the optimized path in terms of cost without compromising the auction’s quality. That’s how the term Supply Path Optimization came into the picture.

Now that we have discussed what exactly the problem statement is, let us get into the approach. Being an unsupervised dataset, choosing the best path with no response variable was quite a challenge. Multiple techniques such as clustering, classification modeling, Data Envelopment Analysis, and Scoring based on classification modeling were tested. Clustering was done on three primary KPIs of a video advertisement. Then the best cluster is picked from the 9 clusters formed with a varying performance of KPIs. This solved the which part of the problem statement but not what. Alternatively, classification modeling was implemented where probability values acted as the bid price modifier, thus solving both the problem statement’s subparts.

To use Data Envelopment Analysis (DEA), a single response variable capturing multiple KPIs significance is required. With various crucial key performance metrics, it took some effort to develop one single value using digital marketing learnings. DEA is an algorithm based on linear programming techniques to measure the efficiency of individual units referred to as DMUs. Using the rDEA library in R, we obtain the relative efficiency ranging from 0 to 1 against each supply path with specified inputs and outputs.

DEA considers only the numerical features of a DMU, however, with constrained values of key metrics (ranging between 0 to 1), DEA cannot differentiate much amongst the DMU’s leading to close efficiency values. Also, the inventory specific features in digital advertising like site domain and supply vendor hold higher significance in supply path optimization. Due to the DEA’s constraint of not handling categorical features, these features were not considered in the technique.

Original post here.

Read more data science articles on OpenDataScience.com, including tutorials and guides from beginner to advanced levels! Subscribe to our weekly newsletter here and receive the latest news every Thursday. You can also get data science training on-demand wherever you are with our Ai+ Training platform.