March 1, 2017
Guest post by Andrew McLoughlin, SEO/Analytics Manager at Colibri Digital Marketing
We’ve written a lot lately about machine learning and natural language processing, and how these emerging AI technologies are reshaping the digital marketing landscape. At Colibri Digital Marketing, we try to anticipate the effects of new technologies well enough in advance to get in on the ground floor and to future-proof our digital marketing efforts. Today, we’d like to share more of those insights.
What you need to know
For those who are unfamiliar with machine learning, this list will explain some of the terms and technologies we’ll be discussing in this piece:
- Machine Learning: A type of artificial intelligence that allows a system to “learn” without being explicitly programmed, through a process of trial and error. The system incorporates feedback, altering its algorithms until it can reliably and correctly process even unfamiliar data. This often leads to emergent insight, especially about networks and relationships between nodes.
- Natural Language Processing: Usually developed through machine learning, NLP allows a system to parse written or spoken language to understand content, usually allowing for idiom, context, tone, and so on.
- Unstructured Data: As opposed to a table or reference list, unstructured data either does not have a predetermined organizational structure or has not yet been organized. It is often text-heavy, so ML and NLP are frequently used to incorporate unstructured data into a computer system or process.
- Segmentation: The sorting of data into discrete groups, either by demographic or some other (often unlikely or highly specific) factor.
- RankBrain: The AI at the core of Google’s search and ranking functionality, which currently uses ML and NLP when considering search requests.
Possible applications of machine learning for digital marketers
In each of the following hypothetical case studies, the bots or tools described would employ machine learning for digital marketing and outreach purposes. They would be “trained” (taught by repetition) to perform a function even on unfamiliar, unstructured data and to deliver clear, actionable results.
1) Discover new opportunities
Consider the law firm Bay Area Bicycle Law. Since they serve such a specific niche, exclusively representing cyclists, their marketing and outreach efforts need to be tailored to a highly specific market.
Imagine a bot that scans social media for images of bicycles, or people wearing bike helmets, and then flags tweets or posts about crashes or accidents. This AI system could find potential clients and could send a message with a clear, targeted, and relevant CTA (call to action), increasing conversions by focusing on clear potential leads.
2) Understand sentiment in customer interactions
Natural language processing could be used to determine tone, sentiment, and so on, on social media, in reviews, in emails, and more.
By segmenting that data by, say, a customer’s journey through a conversion path or by demographic, a company could discover insights, feedback, and potential conversion opportunities that might otherwise have gone undetected.
A bot could be trained to flag warning signs (like comments about price, value, location, and so on) and to pass on that data for further processing, or to chart trends (like the reactions people are having to a change in branding.)
3) Precisely tailor content
By evaluating huge scores of unconnected data, from device type, average page-view duration, previous sites (from cookies), writing level on social media, and so on, an AI system could tailor a piece of content to an individual user, in real-time.
Perhaps one user prefers long-form over bullet lists, while another user favors infographics. The AI system would learn this and customize it appropriately.
Similarly, one product could be advertised in different ways: functional, rugged, stylish, or affordable. An intelligent algorithm can determine which phrasing or qualities a particular buyer is likely to be sympathetic to and help you advertise accordingly.
What’s yet to come?
These examples are just the tip of the iceberg. When it comes to machine learning, it’s important to realize the sheer scale of the potential information that can be discovered through combining huge and diverse datasets and teasing out subtle trends, or through combing hugely complex information like natural language.
Most computer programs are simple clerks, performing a prescribed function according to clearly stated parameters. Indeed, historically, this has largely defined software in general.
Machine learning produces algorithms and systems that are more like detectives than clerks, discovering new data by recombination and exploration, and able to encounter the unfamiliar. Many of these systems are still in their infancy, but others, like Google’s RankBrain, are already out in the wild and they’re thriving.
In digital marketing, content is queen. How and to whom that content is delivered, why it takes the forms it does, and all the other questions that digital marketers grapple with are already being informed by intelligent algorithms.
The algorithms are finding answers in unlikely places and generating insights that had gone unnoticed. Machine learning systems are, quite simply, the future of digital marketing.