For the first time in hundreds of years, writing is set for an evolution. Certainly, throughout the past several centuries, how people write has changed. From etching on stone and metal tablets and quill pens to movable type and the art of the swipe, the instruments enabling humans to write has seen dramatic improvements. In each of these cases, however, the driver behind the writing instrument is human. He or she has to think, plan, outline and draft whatever content they may be thinking. The manual process for creating content is slow, expensive and difficult to scale, and with content becoming an increasingly more critical part to the marketing engine, speed, costs and relevance matter more than ever. In the age of data and technology, there is a better way.
Data Never Rests
The total number of Internet users surpassed 2.9 billion in 2014. Every minute Google receives more than 2.9 million search queries, Instagram users upload 120,780 photos, Tumblr receives 113,520 posts, Facebook users share 2.46 million pieces of content, and Twitter users tweet 546,000 times. Data never rests and online sites and services are collecting a ton of it to help them better understand their users and customers. With the data, they are producing content to engage their audience and trigger some kind of transaction. Traditional static and one-size-fits-all content created based on gut-level decisions are now being replaced with real-time dynamic information flow, massive personalization and data-driven optimization.
Unfortunately, there appears to be a dramatic disconnect between what they want or need to produce to be effective compared with what they are capable of producing using manual content development processes.
Manual Writing and Economies of Scale
The average writer can produce copy at about 125 words per minute, not including research and editing, which leads to long cycle-times. Manual copywriting also costs about an average of 10 cents per word, which leads to high marginal costs. Cost and time efficiency plateaus even for rote or formulaic writing and there are no economies of scale. Some online content sites have simply tried to hire more writers, but crowdsourcing only provides incremental gains, not the disruptive level needed to increase output.
In addition, crowdsourcing reaches to the lowest common denominator as the writing pool expands to less expensive and lower quality writers. As a result, it is difficult to bring manual writing into the dataconomy. While companies may have an innovative infrastructure for capturing and analyzing data and storing, managing and distributing content, a matching infrastructure for creating content may not exist.
Economics of Algorithmic Writing
[bctt tweet=”If an Algorithm Wrote This, How Would You Even Know?”]
A March 2015 headline in the NY Times asked the question, “If an Algorithm Wrote This, How Would You Even Know?” The truth is you probably wouldn’t. Content developed with analytics, algorithms and robots is becoming much more common, particularly as the content vacuum that is the Internet demands a 24/7 content development cycle. The benefits of the algorithmic writing and natural language generation (NLG) are driven by their economics, which combine low production costs with the ability to reduce the production cycle times from hours and minutes to milliseconds. Costs associated with building and setting up training discourse models in specific domains are one-time costs, and the algorithms actually get smarter the more they are used, benefitting all users with improved quality.
No Such Thing as a Write-Once Scenario
Engaging people is one of the most fundamentally important actions of content marketing. Only with analytics can organizations really know what works when it comes to traffic and revenue. Currently, companies can analyze the performance of their content, but they can’t continue to refresh the content based on the analytics to get better results. There is no such thing as a write-once scenario, and manual writing simply can’t keep pace with accelerating content trends. Turning data into stories is one piece of the puzzle, but is it the right story for the right person at the right time? Did the user engage, learn, click, share or buy? Only with NLG can you immediately change the content to improve the performance.
Beyond Analytics and NLG
Analytics create insights and big data can multiply the number of insights, but the insights need to be turned into actions. Using the analytics to fine tune algorithms can accelerate and contextualize content development, helping organizations deliver knowledge, initiate conversations, share stories and drive performance.
Combining analytics with machine-learning algorithms is the next step to optimizing every interaction with a perspective customer, ensuring that adaptive dynamic content is timely, personalized and contextually relevant. Companies can use Natural Language Processing to understand sources and extract new information, and use machine learning and statistics to find patterns and derive new insights. Continuously optimizing content with technology that can read, reason and write will deliver recurring value and build core assets. With adaptive dynamic content, companies learn from users behavior and KPIs that allow them to continuously optimize and revise. The algorithms never stop improving, and as fashions, seasons and consumer behavior change, so will the content.
About John M. Pierre, CEO & Co-Founder of Linguastat– John M. Pierre is a seasoned software entrepreneur with a track record of developing new technologies and products at startups and larger firms in industries ranging from enterprise software to national defense. He was a member of the founding technical team at Metacode Technologies Inc. where he spearheaded the development of technologies and core IP that led to an acquisition by Interwoven Inc. At Interwoven, he managed several product releases for the Content Intelligence product line and was involved in company wide software architecture efforts. He has published several papers in automated classification and text mining. John has an MBA from the MIT Sloan School of Management, a Ph.D. in Physics from U.C. Santa Barbara, and B.S. degrees in Physics and Humanities from MIT.
Photo credit: Kyle McDonald / Modern Furniture / CC BY-NC-SA
Love, love, love this article, John. I think you’re calling this trend way ahead of time. The game theory here is only waiting on the tech, and the tech is due for a tipping point any time now. The push for semantic in recent Google updates is going to narrow the path to discoverability, and this in turn might very well cause there to be no human-detectable difference between methods of composition. Great article. I would love you or your team’s feedback on our API, which correlates themes in unstructured information streams. Thank you!