Nowadays, computer algorithms can be programmed to learn from the data and the inputs that it is fed. Machine learning is set to impact the content marketing industry in exciting ways, in fact, 84% of marketing organizations are now implementing or expanding to AI and machine learning. However, the key question remains – how exactly can machine learning affect the content game?
We set to find out the answer to that in our latest #SEMrushchat with special guest, Geoff Colon who is the Communications Designer at Microsoft, and a compelling voice at the intersection of technology and marketing. His insights, along with our other chat participants, on machine learning and its impact on content cannot be missed. Here is what they had to say:
Q1. In your opinion, will machine learning be the death of content A/B testing? Why or why not?
The AI algorithms that use machine learning can process large amounts of data in milliseconds. In fact, it combines both the speed and the accuracy that goes into the regular, human-based, A/B testing process. Machine Learning has opened up the possibilities to test not just two variables, but multiple variables a lot quicker and simultaneously.
A1: So A/B testing actually becomes A to Z to the nth power testing. Because you can test millions of things at once (if you so desire). #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
A1. Machine learning should actually mean more A/B testing of content because feedback can be faster and more accurate (and the tech to do it is becoming cheaper) #semrushchat https://t.co/fJHPJ1QAGx
— Stephen Kenwright (@stekenwright) May 23, 2018
However, there are some roadblocks before it can completely replace A/B testing in mainstream marketing.
Availability of Data for Machines to Learn
For machine learning to give accurate results, the initial data should be in a format that the machines are able to understand. Therefore, using structured schema markup is important.
A1: ok there's a lot to unpack here:
- for ML to really work, you need to have data that can be consumed by the machine This is where structured data/ markup comes into play. #SEMrushchat— JP Sherman (@jpsherman) May 23, 2018
So, companies that are late to adapt to this kind of format may not be able to churn out that data with enough accuracy to actually facilitate machine learning.
A1: another issue is greater access for late adopting companies (or companies w/out a data scientist) to be able to churn that data. #semrushchat
— JP Sherman (@jpsherman) May 23, 2018
The Size of the Company
Using machine learning to carry out the tests can be adopted by a lot of companies, provided they have the infrastructure to support it. Bigger companies, which have access to data that is usable by machines, will have the ability to switch to machine learning first and in turn, can replace regular A/B testing.
A1: I think it depends on your business size. For big companies the answer is yes. For small companies still the answer is no. #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
Others, which do not have the necessary infrastructure to support scaling, may suffer, unless they have the right tools developed to execute it.
A1: too many companies right now are using ML as some sort of marketing panacea that will solve everything - and without access, consumable data to measure & some sort of affordability at scale, ML will struggle at scale #SEMrushchat
— JP Sherman (@jpsherman) May 23, 2018
A1 we are just SEO's not fortune tellers!
If someone builds tools to replace A/B testing that use machine learning then yes. If not then no. #SEMrushChat— Simon Cox (@simoncox) May 23, 2018
Knowing What Exactly There Is to Analyze
Another reason that machine learning is not expected to take over completely is that although it can speed up the actual process of testing, in the end, it is still a test. It learns from past data that it is fed and the results will vary vastly from real-world situations.
A1: That being said, a machine learning test is still a test. And people act different in real environments so we may still need to build two or more landing pages to figure out how people are behaving when they land on them. #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
Therefore, unless the algorithm gets human help on what exactly it should be testing, it may not give accurate results (despite its efficiency).
A1: I think anything that can be automated that removes human subjectivity, will be. A/B testing is something machines can do thousands of times faster and more efficiently than humans. That said, they still need to know WHAT to test :) #semrushchat
— Jeff Baker (@baker_rithms) May 23, 2018
Ultimately, machine learning still has a long way to go before it becomes completely foolproof and can only happen with data that is learned from the regular A/B testing that we do today.
A1. The basic principles of A/B testing still will exist for a good few years to come - users will still devise strategy and identify elements to test. Without that base, #MachineLearning can't actually learn what's working better #semrushchat
— AccuraCast (@AccuraCast) May 23, 2018
On the other hand, it can help to smooth out a lot of challenges that we see in A/B testing. For example, it can shorten and narrow the length of testing and can probably give suggestions on how to approach a problem (i.e., it can tell you how to search a query, instead of just merely giving you a choice between two questions).
A1: No, it isn't the death of A/B content testing. Does it help to narrow down the length of testing? In my opinion, yes. #semrushchat
— Netvantage Marketing (@netvantage) May 23, 2018
A1 We see places such as Google learning how to ask the right question, rather than just choosing between two questions: https://t.co/XIz1fVIvRx #SEMRushChat
— Bill Slawski ⚓ (@bill_slawski) May 23, 2018
As long as users are doing the searches and the machine can understand the data, machine learning can be used to predict user behavior, but will not entirely replace A/B testing...yet.
A1: Absolutely not. Anyone had to rewrite an email to sound less angry, when you weren't - same thing applies... I A/B test my content on myself every day - then hit send! #semrushchat
— James Scroggie (@seoscroggie) May 23, 2018
A1: As long as users are doing the searches, I don't machine learning will. I do think it will help make better informed decisions and provide predictive data #semrushchat
— Danny Ray Lima (@dannyraylima) May 23, 2018
Q2. What are the benefits of machine learning when it comes to your content strategy?
Machine learning can take over the repetitive day-to-day grunt work from marketers and give them the opportunity to focus more on strategy and ideation. This enables the entire workflow to be more sophisticated and more intelligent to bring in better results.
A2. Over the past few years, our marketing processes have become a lot more sophisticated, intelligent, and efficient thanks to #MachineLearning. We can now focus on #strategy and #ideation, while letting the algorithms do the mathematical and repetitive tasks#semrushchat pic.twitter.com/GQBHQ9Ix1n
— AccuraCast (@AccuraCast) May 23, 2018
A2: Increases in efficiency, process, and focus. Aid in data analysis, and segmentation. Content recommendation. #semrushchat
— Jeff Löquist (@jmloquist) May 23, 2018
Here are a few ways machine learning can add value to your ongoing content marketing process and strategy:
Better Efficiency
Content creation is not easy; it involves a lot of thought, a lot of labor, and a lot of time to create content. On top of all of that, you also have to test to make sure that what you are doing actually resonates with your target audience — this is where machine learning can have a huge impact. It can make it a lot more efficient by taking over testing, finding out which creative works well, and, in turn, get you quicker results.
A2 Micro changes can be tested en-masse, changes applied automatically and recycle the process with only occasional human intervention to check that the site makes sense. Low performing content can be rewritten whenever needed and measured.
Death of the copy writer#SEMrushChat— Simon Cox (@simoncox) May 23, 2018
A2. In simpler words, we spend our time thinking of better content, while the machines do the grunt work of testing options and sorting out what's working vs what isn't #semrushchat #MachineLearning pic.twitter.com/pct5OUwUQO
— AccuraCast (@AccuraCast) May 23, 2018
A2: Machine learning can help you understand what content is working and what isn't quicker. It allows you to edit and update content that isn't connecting with your audience sooner. #semrushchat
— Netvantage Marketing (@netvantage) May 23, 2018
A2 Pt.1: Right now you have a team that decides to figure out what content to create, creates it, amplifies it, analyzes the results. It's painstakingly time consuming and requires lots of thought labor. If you could use ML to test what creative works that helps #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
Quicker Feedback Loop
With AI algorithms, the learning is instantaneous and can automatically make adjustments to the content or campaign, based on what it has just learned to provide optimal results. For example, when machine learning finds out that “Creative A” works better than “Creative B,” it will instantly adjust and show your audience more of “Creative B,” without any manual intervention. This saves you the effort of having to go back to the drawing board to analyze the data and make corrections.
A2. Marginal gains in engagement - big improvements in forecasting and attribution when we can see how large numbers of metrics we're tracking are changing on the fly #semrushchat https://t.co/gBjpPbiBg2
— Stephen Kenwright (@stekenwright) May 23, 2018
A2 Pt. 2: THe other area where ML could help is instant feedback so you can turn off waste areas. For example, when testing creative, you have to analyze the data. What happens if ML adjusts that like a plane in flight that adjusts based on turbulence and trajectory? #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
Easier Keyword Research
Keyword research and content ideation are the keys to a successful content strategy, and both can be simplified by using machine learning. Once the algorithm identifies a popular and relevant keyword for your business, it can conduct in-depth keyword research and provide content inspiration for your content pipeline.
A2
Analysis of target keyword is key to content strategy.
Our tool @SocialAnimal_io utilizes NLP and Machine Learning techniques to peer into what makes good content tick.
I use the tool to find top shared articles, keyword insights and content inspiration. #SEMrushchat
— Krishna Rg🚀 (@krishnarg22) May 23, 2018
More Personalization
Machine learning can understand what kind of content is appreciated by a specific target audience and then curate the content tailored to them. This kind of personalization can enhance your pay-per-click ROI and improve engagement.
a2 With Machine Learning or AI is forecasted to personalize & enhance advertising to the targeted audience with more precise content - specific to that consumer.
Wild but True! #semrushchat
— Debi Norton (@BRAVOMedia1) May 23, 2018
A2: I think ML will help automate content optimization. It's going to help us how users are engaging in real time, and how it's being pushed out across channels. It's also going to be a big asset in content development and asset creation. #SEMrushchat
— Danny Ray Lima (@dannyraylima) May 23, 2018
A2) More analysis and personalisation could lead to lead to better engagement and better sales#semrushchat
— David Rosam (@davidrosam) May 23, 2018
Predicting Trends
Better yet, machine learning can be predictive and give marketers a first-hand look at which trends to expect. This will help marketers to create a content pipeline that resonates with the audience better.
A2 If ML is writing content though, I am going to be using ML to find and catalogue content that interests me - so your ML generated content will be read by my ML content reader. If a tree falls in the forest...#SEMrushChat
— Simon Cox (@simoncox) May 23, 2018
A2: i think the major benefit of ML will be the ability to predict trends of interests of the consumer to which a content pipeline can be created #semrushchat
— JP Sherman (@jpsherman) May 23, 2018
Despite all of the obvious advantages, the algorithm is still controlled by a machine, and like A/B testing, it may need the occasional human help. More importantly, it could be manipulated, which will need to be re-checked with a human eye to see if the conclusions drawn are indeed accurate.
A2 p2 but the machines can be gamed - until the platforms adapt, and the cycle continues so your content strat has to include a bit of futurecasting #SEMrushchat pic.twitter.com/Kf26m675jx
— Thomas Broadus (@TbroOnline) May 23, 2018
A2: One of our twitter profiles uses a news agitator, but it's easy to push fake news through, the intelligent filter sets help with a lot of the spam so it cuts operating time, but is also not the person then when making decisions its just a bot. #semrushchat @djgeoffe @semrush
— Alexis Huddart (@Flexoid) May 23, 2018
Q3. What are some concerns marketers may have about automating content marketing or using machine learning to analyze and gather content?
Depending solely on machine learning to curate and analyze content comes with a variety of advantages, as we have just seen. However, as with any technology, it also comes with its own set of disadvantages.
Here are some of the drawbacks:
Lacks Human Emotion
To resonate with a target audience, the content you are putting out has to connect to them on some kind of emotional level. This is still not possible with machine learning as its understanding and capacity for compassion is severely limited. When trying to connect with your audience on an emotional level, human-created content will most likely work better than machine learning generated content.
A3: Reaching your readers on an emotional level is still exceedingly important. Machines are not yet able to interact with the necessary levels of understanding, compassion and motivation to make content analysis fully automated. #semrushchat
— Sarah Weissberger (@skweissberger) May 23, 2018
No Human Intelligence
Creativity, sarcasm, context, and semantics are all indications of human intelligence that is still missing with machine learning. The algorithm can easily struggle when trying to make a distinction between a sarcastic remark and an exclamatory statement and thus, can diminish the value of the content created. Similarly, the content created may not be creative enough to attract attention.
A3: A concern about automated content would be it using the right voice to connect with an audience. #semrushchat
— Netvantage Marketing (@netvantage) May 23, 2018
A3: Someone said it earlier, machines don't have the capability for being creative to a degree. Sticking to their strengths in analyzing and assisting would be much more useful than it trying to create the content from a human perspective. #SEMrushchat
— Geeky Fox (@TechKitsune) May 23, 2018
Limited to Learning From Data It is Exposed To
All machine learning algorithms are programmed to learn from the data it comes in contact with, which can cause two problems.
First, if the data it is exposed to is faulty or biased, then its learning can be affected. Second, there is no way it can broaden its scope or learning by exploring and discovering data on its own. This data dependency can cause severe knowledge gaps within its learning, which can only be overcome with the help of humans.
Control
Machine learning algorithms are still guided by humans and need to be controlled by them. They still need approvals from people to implement and execute decisions that are made with the data that it generates (hopefully!). One wrong move and it is not the machine that takes the blame, but the marketer who gave the system the clearance to go ahead with its faulty reasoning.
A3 Control! As with any ML / AI stuff you want final approvals but that might just be on boundaries the ML can operate in . Something goes wrong and your brand is down the pan its not the ML kit that's going to lose its job.#SEMrushchat
— Simon Cox (@simoncox) May 23, 2018
Given such limitations, machine learning can be looked upon as a highly skilled aid to any marketer. It can wield technology and make the job simpler and give them the opportunities to make more informed decisions.
A3: I think we need to be cautious about ML. We need to understand that we are in control and ML is just another tool that helps us make informed decisions. #SEMrushchat
— Danny Ray Lima (@dannyraylima) May 23, 2018
A3: I also think you still need humans with ML to diagnose things. Companies will look at this as a way to cut fixed costs but with medical technology we still have doctors. Same thing here. #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
A3: It's a great marketing aide. But—it's a machine. I find it creepy when a bot greets me in the morning and tells me I've been doing good or bad with my goals over the last week. What if the machine learning tactics I use become overwhelming too soon? #SEMRushChat
— Narmadhaa (@s_narmadhaa) May 23, 2018
Q4. How might machine learning-based content marketing technologies fit into the marketing technology stack?
Provides Customized Content
The best fit for machine learning within content marketing could be to further customize content and make it highly specific for the audience – including the channel in which the content is served.
A4: It's my opinion that the best fit for #MachineLearning in #Marketing technology will be highly targeted and customized delivery of content via email, apps and social #semrushchat pic.twitter.com/y9git4XyM2
— Mike Bryant (@MichaelRo22ss) May 23, 2018
Simplifies A/B Testing
Currently, it is incredibly similar to every day A/B testing, but a lot faster and more accurate as it runs numerous variables and large amounts of data. Therefore, it will enhance creative testing, but not completely replace it, as stated earlier.
A4: For now, it should be in a place where A/B testing lives. It should be to enhance creative testing. Not replace it, to act as an analysis/diagnostic tool. #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
A4: Suppose they would fit INSIDE them as test pieces for marketing content management systems. As i said earlier, ultimately the creative directive is up to humans. #SEMrushchat
— Geeky Fox (@TechKitsune) May 23, 2018
For instance, machine learning can present website data or content, even if it is in a different UX format for each of your site visitors. Furthermore, it can learn from the conversion metrics it gets to optimize your website even more. Testing thousands of variables and then analyzing the results take an extremely long time, but, both of these problems can easily be solved with machine learning.
A4: #ML will present data/content/ux just for that specific user, so your sites content will present in 1000's of different variables, till it has its equilibrium point, so it thinks. #semrushchat @djgeoffe @semrush
— Alexis Huddart (@Flexoid) May 23, 2018
Enables Better Reporting
Automating research and analytics through machine learning can reduce marketers’ workloads and help them to focus more on their strategies. Its speed and accuracy in dealing with big data can make it a very helpful reporting tool.
A4: Automating research and analytics through machine learning can help reduce the MarTech workload. #SEMRushChat
— Marccx Media (@marccxmedia) May 23, 2018
A4 Big Data crunching. Marketeers love data - and so do machines. Thats where we will see a a lot of inroads to begin with - the Business Information level.#semrushchat
— Simon Cox (@simoncox) May 23, 2018
A4: I tend to trust ML as an unbiased reporting tool - it finds interesting things to further explore.
I tend to not trust ML yet on the creation side - despite the perceived speed benefits #SEMrushchat
— JP Sherman (@jpsherman) May 23, 2018
Improves Customer Service
Machine learning, such as chatbots, can also drastically simplify customer service by providing your customers with answers to their most frequent support queries.
a4 At present Chatbots are already doing this & helping Brands that receive lots of interaction by answering commonly asked questions/inquires. As far as "solving" consumer issues - it's not there - yet. #semrushchat
— Debi Norton (@BRAVOMedia1) May 23, 2018
A4 Query Optimization based upon Entities with Machine IDs use trends and label things in categories as part of a knowledge graph that crowdsources how it learns from the Public. #SEMRushChat
— Bill Slawski ⚓ (@bill_slawski) May 23, 2018
Takes Over Back-end Tasks
Other backend tasks such as keyword research, locating influencers, and identifying new content opportunities can easily be taken up by machine learning. It can be the perfect addition to the learning and implementation side of marketing.
A4. Machine learning-based content marketing technologies will essentially fit in learning and implementation side of the marketing technology stack. The right tests could even help identify new types of data (emotional cues, sentient cues)#semrushchat
— AccuraCast (@AccuraCast) May 23, 2018
A4 a) Presenting personalized content in the presentation layer; b) Providing insights in the content creation workflow on the three pillars of content, engagement, and conversion; c) providing automated variant A/B testing at atomic elemental levels #SEMrushchat
— Flockrush (@Flockrush) May 23, 2018
A4: If ML can develope a cloud based storage, markters can access content and data simply by adding tags, interests on data. #SEMrushchat
— Danny Ray Lima (@dannyraylima) May 23, 2018
Q5. Last, but not least, how exactly is machine learning changing the content game?
Machine learning can be applied to various aspects of content marketing to enhance it and increase its overall impact. Our chat participants mentioned a few interesting ways in which it is improving and changing the content marketing game as we know it:
Improved Data Analysis
Machine learning is definitely changing the way we test and analyze data. It allows us to make informed decisions and gives us an idea about future trends.
A5: I think ML is changing the way we automate tests and data points, it's allowing us to make better-informed decisions and allowing us to count for future trends. But, It's making us smarter marketers and keeping us accountable for the data we collect. #SEMrushchat
— Danny Ray Lima (@dannyraylima) May 23, 2018
A5. #MachineLearning is changing the content game in 4 main ways:
- Sharper: processes are more Efficient
- Smarter: less scope for human error
- Faster: instant feedback
- Safer: taking risks without losing much time / resource#semrushchat pic.twitter.com/RYYI5HzrfY— AccuraCast (@AccuraCast) May 23, 2018
It also makes us smarter marketers by helping us understand the ROI of content marketing better and by providing quicker insights into the metrics that determine a successful content marketing strategy.
A5: The other area is impact. ML could be the ROI solution that evades most marketers. Did this work? No, it was a dud. Learning that quickly is just as important as knowing what is a success. Cuts waste. #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
A5: Better & easier data mining/manipulation. Easier personalization. #semrushchat
— Jeff Löquist (@jmloquist) May 23, 2018
Better Content Analysis
Though machine learning is still not adept at content creation, it can efficiently analyze what worked and what didn’t in real-time with concrete results.
A5: I think the biggest area is content analysis. Not original creation. Humans still best with original ideas but we like to see if those original ideas will resonate and analyze that in real time. This is where ML has the upper hand. #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
A5) Content analysis. Data analysis.
Content creation and other creative tasks are over the horizon.#semrushchat
— David Rosam (@davidrosam) May 23, 2018
More importantly, it can identify the holes in your content strategy and highlight what content you should be creating more of. This will give you a significant boost in identifying new content opportunities as well.
A5: It also could analyze what content you're not creating enough of that you should be. Maybe then people at your company will listen when you note from analysis you're lacking certain topics. #semrushchat
— Geoffrey Colon 🤓 (@djgeoffe) May 23, 2018
Both of these give content marketers more of an opportunity to plan their strategies better.
A5: Research and analysis. It's helping us write the right content for the right audience. #SEMRushChat
— Narmadhaa (@s_narmadhaa) May 23, 2018
A5: Analyzing stats is the biggest help from ML! It will give content marketers an inside look as well as better planning for our client's future endeavors! #SEMrushchat
— Geeky Fox (@TechKitsune) May 23, 2018
Personalized Email Marketing
Another channel where machine learning has an impact is email marketing. It can help by making it more personalized, behavior-based, and extremely targeted.
A5: It's making e-mail blasts easier to oversee, but there's certainly the potential for weirding-out recipients. As for content specifically, it can make rote creation/curation easier, but there's still the need for human editing. #SEMRushChat
— Marccx Media (@marccxmedia) May 23, 2018
Identifying New Target Audiences
With repeated testing of large numbers of variables, machine learning can identify a whole new group of leads that want to consume the content that is generated.
A5: #MachineLearning can also help find the right audience: new leads that actually want to consume your content. #semrushchat pic.twitter.com/iq56THtUp8
— Mike Bryant (@MichaelRo22ss) May 23, 2018
A5. Step by step: Greater accuracy in 'successful' content (depends definition of 'success'), forecasting and greater insight into how users interact. Although I doubt human creativity will ever be outwitted by ML... one hopes... #semrushchat
— Sarah Marks (@_ofwanderings) May 23, 2018
Understanding Context Search
With Google’s search algorithm becoming significantly sophisticated, search engine optimization is no longer restricted to a few “golden” keywords. Instead, it also includes voice search and context search, both of which are understood within machine learning.
A5: When it comes to Google search machine learning means that content optimization is not about keywords anymore. Algorithms can increasingly understand what the content is about and how valuable it is. #semrushchat https://t.co/9zDYlCtsFk
— Tadeusz Szewczyk (Tad Chef) (@onreact_com) May 23, 2018
Improving Customer Experience
Machine learning will improve user-experience, which in turn, will facilitate better engagement and more conversions.
A5: #ML will ultimately improve the user experience, which in turn will improve operational expense and revenue generation for said content presented #marketing #semrushchat @djgeoffe @semrush
— Alexis Huddart (@Flexoid) May 23, 2018
A5 Content is the voice through which a brand guides and communicates with the customers online, machine learning helps us become better with that communication. #SEMrushchat
— Flockrush (@Flockrush) May 23, 2018
A5: It's changing the quality of content, high rich image information, better understanding between the content and marketers/distributors. #machinelearning #SEMrushchat #Contentmarketing
— Veeraeswari (@VeeraeswariS) May 23, 2018
It is evident that machine learning is already bringing some significant changes to the content marketing world. However, if you are using it or planning on using it to help with more than just improving the speed or accuracy of your strategy, then the learning aspect of the algorithm needs more of your attention.
It is crucial for you to do your research on how machine learning can improve your existing marketing strategy, instead of just jumping on the bandwagon because other people are. Otherwise, you may not succeed with this ground-breaking, new technology.
A5. Step by step: Greater accuracy in 'successful' content (depends definition of 'success'), forecasting and greater insight into how users interact. Although I doubt human creativity will ever be outwitted by ML... one hopes... #semrushchat
— Sarah Marks (@_ofwanderings) May 23, 2018
Do you agree with our chat participants? How do you think you can use machine learning to tweak your content strategy? Leave us a comment and let us know!
Join us Tomorrow 11 AM EST on Twitter for #SEMrushchat
Each week our chat offers a lot of educational information, and we can't possibly fit all the tips and information into our recap. The best way to learn is to join us. Tomorrow Kate Morris and our chat participants will help us demystify SEO for e-commerce!
The biggest impact I see with ML as it relates to content marketing is its ability to find out what content our target audience is interesting in, that we may be presently ignoring and should be optimizing for (ergo Google's "Searches Related to" and Search Console > Search Traffic > Search Analytics). This will help guide content strategy, make quicker and informed data-driven decisions, and track content performance in near real time.
Overall, the goal of ML (as it pertains to content marketing) should be to aid in creating content that resonates with our audience NOT trying to replace the human aspect of making real connections with people (like automating content creation), a huge no-no in my opinion.
Thanks for the detailed round-up, Becky! :)