Algorithms - From Raw Data to Brand Passion

What does any brand want to achieve in the world of content marketing? If you're doing it right, then it's to provide product information and capture the consumer's attention through culturally relevant content. That should open the door to discovery, and ultimately a sale through an ecommerce platform created by editors and developers.

Sounds simple in principle.

But now add social media, voice search technology, the use of environmental Bluetooth beacon signals, and other geo-location software to a brand's content marketing strategy and you get a metadata quandary. How can you make sense of such a complex system of structural multiplicity?

The ideal solution is to build a media and content publishing platform with a market-friendly taxonomy defined by editorial guidelines set to development requirements that accommodate the messaging of the brand.

The following is how you integrate these new, exciting ways of communication into a functional and effective whole.

1. Agree on the Rules. Get the relevant stakeholders to sign off on a set of rules and specifics. These may be complex and detailed but the investment is key to the success of the platform.

2. Start Tagging. Once you have stakeholder agreement, take into account the basic, key elements (both factual and editorial) that will capture and control your metadata. You must identify the field types that best capture the essence of your product or service. These content types are usually presented as keywords, tags, genres, and subjects that are designed to help the consumer in discovery and recommendations.

3. Get Data. Most of the recommendations you will make are based on analytical tracking from purchases, viewing habits, search history and social activity. The goal for companies building their own media and content publishing platforms is to marry the algorithm with the editorial through data analysis.

4. Build a Feedback Loop. Based on the data you get, workflows and process should be built on a cyclical model that provides functionality in the editors' tools for creative input and quality control.

The way a hierarchical taxonomy works is actually pretty simple—it's a parent-to-child relationship that cascades to multiple generations (fields). But where it gets complex is when it becomes much more like a family tree (field structures)—it can build relationships horizontally (tables) to a cousin, or a brother, and splinter into multiple families based on marriages, divorces, and births: that is, additional fields and tables.

The most important part of building a good algorithm is to find the right relationships within this complex structure of classifications. If I bought a garden tool on Amazon for my wife, it doesn't mean I'm into gardening. My brother Stan loves soccer and sends me links to videos of great goals, but I'm not really into soccer. I bought the new Selena Gomez single for my daughter on iTunes, and I can't stand to listen to her music.

So how do you fix this problem?

Brands and retailers need to engage with the customer in a much deeper way. For example, ask me if I'd like to be recommended similar products, or products within this type of subject area and activity, music style, or movie genre. Gather the necessary essentials to create a profile on me as a user of your ecommerce site without being too invasive. Don't rely on my web history to do that for you.

And most importantly, don't make assumptions based on algorithms.

When successfully created and implemented, an algorithm can be like an elusive skin that can transform your skeletal data into a muscular, powerful tool that can provide a channel for analysis, accuracy and discovery. Without it, data is simply a storage bin full of discarded items, a collection of random pieces organized by factual matter; color, type, material, and other types. That said, algorithms need a quality control step in the process. The analysis must be vetted through an expert lens of human expertise, led by editorial resources with an understanding of the vertical on view.

With these algorithms you can control the path to your target, form a hypothesis on products, or communicate to a large audience through preferences and passion. Now which brand wouldn't want that?