I recently asked Christian Persson, developer of the platform that supports our recommendation engine, “How much code do we use for our recommendation algorithms compared to everything else?” In our academic research code, everything else is close to zero. At Infobaleen, we have developed a tool that helps marketing managers turn their transaction data into optimized campaigns. The campaign optimization tool accomplishes more than recommendations, but how much more? I wanted a number. “Let’s check,” Christian replied. I was surprised by what he found out.
But before we explore that, let’s think about what makes working days of marketing managers strenuous. Think about them as a race — a hard one. Transforming raw data into insights and valuable actions is a point-to-point off-road cycling race. You start in the valley of Raw Data, and you must journey through the checkpoints of Acquired Data, Cleaned Data, Analyzed Data, New Insights, and Decisions Made, before arriving at the finish line of Actions Taken in hopes of receiving the trophy and the prize check.
At the finish line, depleted and dizzy, you are bombarded with questions: How did it go? How do you feel? How will you do better next time?
Next time? Yes, it is a tour. Never-ending. When one race ends, the next begins. Evaluate and start over.
The race can be challenging, strenuous, and risky, but it can also be joyful, rewarding, and epic. How? Because in the data action race tour rules of competition, there are no requirements about mandatory equipment or gear. You are free to choose your support team and means of data transformation.
Two tips for marketing managers
Look for a support team of people who are passionate about their mission, your success. Passion brings a suite of benefits. Working with the attentive, skilled, and tireless takes you further, faster.
Stick to one vehicle of data transformation. A unified end-to-end AI platform will simplify your day, and the one that suits you best will increase your sales, cut your costs, and free your time.
So how would you describe the architecture of an AI platform that can change your day and your business? The deepest neural networks put into a small-footprint platform? No; imagine driving a road bike with a thirsty V Twin engine — a massive burnout, and then you run out of gas. Not cool, and it takes you nowhere.
The other extreme would be linear regression put in a petabyte data platform — a toothbrush engine in a cargo bike. No action.
The well-designed AI platform
In the well-designed AI platform, balance is key. Machine learning and platform components that are designed to work together run blissfully smoothly. You have found your solution when you forget you are interacting with a platform and instead think the algorithms are your reasoning, the recommendations your ideas, and the actions your power. Combined with your business expertise, strategic thinking, and human creativity, you form a winning team.
“I have the numbers now,” Christian told me. “We have implemented the core machine learning models in no more than one percent of all lines of code!” And it turns out that the code for training and evaluating the models represents less than 20 percent of all code. The remaining lines, more than 80 percent of the codebase, control storage and form the backend.
In practice, our AI platform — developed in iterative cycles of rethinking and improvements — is a whole lot about automation.
Do you want to pump adrenaline into your veins with the roar of an oversized engine? Look elsewhere. We’d rather want to seat you on a carbon fiber frame powered by an integrated electrical engine with auto shift included — think instant torque, accurate handling, and silent speed. Feel pure joy.
We were surprised when we checked the numbers. However, looking back, we realize how much effort we have put into simplifying the journey from raw data to actions taken. We want to help you win the entire tour.
Curious to know how we can help you? Please contact us for a free consultation.
Add your comments to the LinkedIn post.