5 Solar Energy Breakthroughs with Machine Learning
Intro: Machine learning in the solar energy industry The high availability of data in the energy sector makes it a...
Client: Fortune 500 Bank
Vertical: Marketing Optimization
Model: Upper Confidence Bound/Epsilon-Greedy
A Fortune 500 Bank needed to improve its conversion attribution modeling scheme. Namely, an outdated qualitative evaluation method needed to be replaced with a solution that estimates the marketing attribution values at the keyword level.
The bank had a third-party ‘black box’ ad bidding system and they wanted to verify efficacy while improving price-per-click (PPC) bids on search terms in Google Ads Marketplace. The company lacked an accurate way to estimate their marketing attribution per keyword. They relied on qualitative tracking to assess the third-party bidding tool.
By using a qualitative tracking method, the bank could not make effective choices in online advertising. Results were not measured with accuracy. Therefore, it was difficult to know which marketing efforts delivered and which failed.
To solve this problem, we developed a quantitative model to optimize the bank’s marketing. The solution involved several components.
1 - Estimate Keyword Value: Assembling An Advertising Dataset of existing Online Ads
Knowing which exact keywords drive conversions is crucial for tuning the search campaigns to run at their most efficient ROI. Since we had a production predictive model, our focus was to identify and improve future results based on existing results. We began by aggregating historical data to estimate the value of each keyword as a benchmark. The bank had stored historical bid data on previous campaigns going back several years and determined a correlation keyword attribution.
2 - Choosing The Right Quantitative Model
The model selection process was aided by testing several strategies simultaneously, known as “Multi-Armed Bandit.” This model is helpful in solving resource allocation problems. If you have option A and option B in online advertising, allocating a limited budget between those options doesn’t have to be guesswork. The model will identify winning campaigns that would benefit from receiving additional budget and de-rank failures that need to be turned off.
3 - Testing New Solutions
A few approaches were applied to optimize different campaigns. Other models were tested but these resulted in the most immediate improvement.
Upper Confidence Bound: This strategy is based on the Optimism in the Face of the Uncertainty Principle, and assumes that the unknown mean payoffs of each arm will be as high as possible, based on observable data.
Epsilon-Greedy: A randomly chosen campaign was selected a fraction ε of the time. The other times, the arm with the highest known payout is pulled and these were measured in comparison then reinforced.
4 - The Result
Blue Orange Digital improved price-per-click bids on search terms in Google Ads Marketplace. We tackled the problem by means of a custom reinforcement learning algorithm. By using historical data of past campaigns as a baseline, the model was trained to estimate keyword values and thus find the best allocation among keywords. Parallel campaigns were used for further training of the model, which provided real-world reinforcement of its parameters.
Such a solution provides the marketing team with a solid foundation for immediate and continued optimization efforts.
Big data still has massive potential to drive market growth, as trends across industries show. Businesses that wish to leverage the power of big data and gain valuable insights need to invest in infrastructures and solutions that can efficiently process and analyze this data. Marketing and sales are no exception to this rule. In the past years, martech leaders have been investing more and more into machine learning-based marketing optimization tools.
Tackling marketing optimization should not be a daunting task!
At Blue Orange Digital, our mission is to help companies get a grasp of their data and make it do the work for them. We have vast experience setting up data-centric architectures and implementing modern analytics solutions.
Expert knowledge can assist your team in every step of the process. Tell us about your project and get started today.