Machine learning capabilities are increasingly accessible to enterprises and even SMBs. And with marketers aiming to take advantage of the vast volume of data now available to them, many are turning to machine learning to better utilize that data to automate processes, inform decision-making, and engage their audiences.
From personalization to product recommendations that drive up-sells and cross-sells, testing to inform ad campaigns and offers, and more, machine learning holds tremendous promise for marketers. To gain some insight into the best ways marketers can apply machine learning, we reached out to a panel of marketing pros and asked them to answer this question:
“What’s the #1 way marketers can apply machine learning to improve their digital strategies?”
Meet Our Panel of Marketing Professionals:
Read on to learn what our experts had to say about how you can leverage machine learning to improve your digital strategies.
Zack West is the Director of Marketing at Novomotus, a digital marketing agency specializing in health and wellness related projects.
“Leverage systems that are being built by…”
Companies with much larger budgets than their own. One of the key requirements of machine learning is data — lots of data! Most companies aren’t going to be generating enough data on their own to have a lot of novel application for machine learning. However, their customers, expenses, and strategies all generate data points that can be optimized by leveraging larger datasets of corporations with vast resources and immense amounts of data. Probably the most recognizable example I can think of is Facebook’s advertising platform. They are able to take data generated from a websites users that generate conversions and find similar uses from within the Facebook community to advertise against. That’s to say: Facebook can apply conversion from a website to identify millions of other people that are likely to convert in the same manner.
Ran Craycraft is a Managing Partner at Wildebeest.
“The #1 way marketers can leverage machine learning to improve digital strategies is…”
One size does not fit all, and machine learning gives marketers the power to serve the most effective content to the right user at the right time. A great place to start is to measure your brand’s awareness and sentiment across the web to help identify and categorize your social audience into actionable groups.
For most marketers, machine learning might as well mean computer voodoo. They hear about it, read about it, and see awards given for it, and so they’ve also gotta have it. There are a lot of misconceptions about machine learning and AI that are typically rooted in underestimating the massive amounts of clean data required. If a marketer wants to dive into ML, plan for half of your time to be spent accumulating and sanitizing the data.
Another big no-no for marketers is to understand that the academic research they’re reading about will take many iterations in order to evolve into production-ready campaigns and consumer products. Many times, we read about a proof-of-concept from a university’s researchers that isn’t yet ready for prime time; however, the headline of applying this to our brand’s offering is irresistible. Remember that applying ML to your product takes time and experimentation – today, it is just as much art as it is science.
Andrew Becks is a passionate digital marketer. He is the co-founder and COO of 301 Digital Media, a Nashville-based marketing agency that helps to build and grow brands of all sizes. He has spent his career as a digital marketer for restaurant chains, cable television networks, media companies, and beyond.
“The #1 way marketers can apply machine learning to improve their digital strategies is to…”
Apply the technology to multivariate creative & audience testing and budget allocation. Think of an old fashioned A/B test, with two versions of ad creative, each painstakingly designed and flighted, tagged with manual analytics tracking. After the A/B test concludes, budgets are shifted around by a campaign planner or analyst in an effort to deliver the highest possible return on ad spend (ROAS).
Enter machine learning, which allows savvy marketers to not only create a nearly unlimited number of creative variations and combinations, factoring image, copy, headline, size, font, color, and beyond – but it also allows marketers to sit back and let machines allocate media budgets to the best performing creative and audience segments/cohorts in a way that optimizes performance for ROAS at a rate that 10x-100x+ more efficient than human optimization alone.
And that is truly just the tip of the iceberg when it comes to the power of machine learning and its application in marketing.
George Hartley is the CEO and Co-Founder of SmartrMail.
“The #1 way marketers can apply machine learning to improve their digital strategies is in…”
Their product recommendations. Amazon has said that over 35% of their annual revenue is generated through machine learning-powered recommendations both on-site and in marketing campaigns. While solutions for SMBs didn’t really exist 3 years ago, a handful of affordable solutions are now available in the market. While our company, SmartrMail, focuses on making machine learning accessible for email marketing specifically, there are other solutions in every marketing channel from social to display advertising.
Christopher Penn is the Co-Founder of Brain+Trust Insights, Inc. He is an authority on digital marketing and marketing technology. A recognized thought leader, author, and speaker, he has shaped four key fields in the marketing industry: Google Analytics adoption, data-driven marketing and PR, modern email marketing, and artificial intelligence/machine learning in marketing.
“The #1 way marketers should be applying machine learning to improve their digital strategies is…”
Through unstructured data mining. So much data sits in digital desk drawers and filing cabinets, never to see the light of day – so-called dark data. Marketers never analyze it because until now, it was cost- and resource-prohibitive to read it all. What used to take months or years now takes seconds with machine learning to extract basic insights and find valuable information.
BrainTrust Insights recently mined over 50,000 customer complaints from a financial services database and found that identifying emotions of complaints – both type (anger, fear, joy, etc.) and strength – accurately predicted customer outcomes. Most of the time, complaints like these just sit in a CRM, ignored for all time. By mining similar data, organizations could identify problems and improve products, services, and brand equity much faster – and less expensively.
Once customer data has been mined, tagged, and analyzed, use data enrichment services to provide additional context for analysis. What customer characteristics also lead to desired outcomes (or prevented outcomes)? In the example above, we did not have data like household income or social graph data, yet these additional dimensions might have shed more light on the customer complaints. With platforms like DMPs and CDPs, machine learning tools are necessary to fully understand the customer and their data, structured and unstructured alike.
Mike Khorev is a Growth Lead at Nine Peaks Media, a digital marketing company that helps tech and SaaS businesses generate more leads and grow revenue online. They offer expert advice on marketing your company the right way through performance-based SEO, digital marketing, web design, social media, search engine marketing, and many other online practices.
“The #1 way to apply machine learning in marketing that is currently available is to…”
Install a chatbot or AI-driven live chat platform on their site where you can set up multiple playbooks/sequences according to your needs and goals.
Currently, the most advanced chatbot platform offering machine-learning features is Drift. It is quite expensive compared to other options in the market but packs some unique features not offered by others, including the ability to set up playbooks mentioned above and lead the conversation without marketer or sales to be invoiced.
For example, we can set up a sequence to allow the chatbot (Drift, in my case) to ask pre-qualification questions before a live salesperson jumps in. For example, the chatbot can ask preliminary questions like their name, email address, company, phone number, the specific time to call them, and so on. We can also set up the chatbot to ask their specific needs related to our product, their current problem, their budget, and other questions according to your product/service. After they answered these questions, a live salesperson can jump in the chat in real-time.
This way, the salesperson can immediately start the sales closing process, avoiding the preliminary questions asked above. Sometimes, asking their email address and other details can be a back-and-forth process ruining the prospect’s experience. By making this process more seamless, we can improve their overall experience, and so increasing the chance of conversion visitor to a lead.
Andrei Vasilescu is a renowned Digital Marketing expert and CEO of a money-saving platform in the name of DontPayFull. He has been providing cutting-edge digital marketing service to various international companies and different online coupons of various brands for years.
“In today’s marketing industry, content is the prime ruling factor, and therefore marketers are commencing their best efforts to…”
Create and publish content that really produces successful conversions. Machine learning has empowered marketers to produce the exact type of content that really works for their businesses. Therefore, the best way to improve your digital marketing strategies is to implement machine learning tools for your business content.
There are several ML tools which help in several ways to construct the most useful content which are actually capable of converting your target audience into payable customers. ML tools accurately analyze the engagement levels through predictive analysis which can guide you to reconstruct your content to improve audience engagement. Some ML tools can indicate that certain parts of your content are not hitting the mark to create certain level of engagement.
In addition to that, ML tools can schedule the time of publishing your promotional content right at the time when it can produce maximum engagement among your social media audience. By analyzing the customers’ online behavior and habits, ML tools are capable of sending the right promotional content to give consumers the right information at the right time – right when they’re looking for it. Machine learning can improve your email marketing strategy by sending the right content to instantly engage your target audience and make them purchase from your business. So, immediately introduce machine learning tools in your digital marketing process for much better ROI.
Jeff leverages his thirteen year record of success scaling PPC accounts and building strong client relationships to provide his team with the tools and resources to successfully manage Hanapin’s clients. Jeff shares his industry knowledge by writing for Hanapin’s blog, PPC Hero & Search Engine Land, and has spoken at industry events such as Hanapin’s Hero Conf and SMXL Milan.
“The #1 way marketers can apply machine learning to improve digital strategies is the…”
Ability for machine learning to drive deep insights quickly. For instance, Google has recently announced that ‘responsive search ads’ has been launched into beta. This functionality allows advertisers to submit up to 15 headlines and 4 descriptions to the Adwords system. Adwords will then take those 15 headlines and 4 description lines and arrange them in a highly relevant, random order to create dozens of ad variations that contain up to 3 headlines and 2, 90-character description lines.
Why does this improve overall digital marketing strategy? Simply put, the machine can create, test, and analyze numerous messaging variations simultaneously, and faster/more accurately than a human could. As a result, marketers will learn much faster, what variations of message are resonating and which ones aren’t. That information can then be analyzed by the marketer to determine if the messages being tested through the use of paid search can be applied to other areas of the digital marketing program.
Cristian Rennella is the CEO & CoFounder of MejorTrato.com.mx.
“With marketing automation, thanks to artificial intelligence, we were able to improve our digital strategies…”
But surprisingly, 61% of SMEs do NOT use any form of marketing automation.
Artificial Intelligence (AI) is becoming more intelligent and intuitive, so, yes, in the future it is best to trust in technology and rely on the machine.
Through Deep Learning, using Google’s platform TensorFlow, you will be able to create models that will predict what time of day, segment, subject, media, design, text, landing page, monitoring conversions, etc. is the best for your digital advertising campaigns, plus even what the content should you use (based on the most engaging topics on the web at the moment).
We started using AI 5 months ago for our advertising campaigns, and the conversions increased by 21.3% (yes, AMAZING!).
Carmine Mastropierro is the owner of a digital marketing agency, several blogs, and is a self-published author.
“The #1 way marketers can use machine learning to improve their digital strategies is…”
Through the use of customer service tools. Chatbots, for example, can collect information on leads, provide extra resources, and guide users through a sales funnel. With the advancements in machine learning, chatbots will eventually be able to change how they interact with users based on individual segments. Similarly, machine learning in customer service can help discover what triggers customer loyalty, churning, and valuable metrics.
Maksym Podsolonko is the CEO of Eazyplan, an online tool that automates planning of weddings, corporate and social events for professionals and their clients.
“Artificial intelligence is the only way to effectively analyze your execution at scale…”
Only Machine Learning is capable of sifting through all data and finding the patterns that the human eye simply misses.
And this analysis is a must for effective improvements on digital strategies. At this stage, the majority of organizations are not using this opportunity, leaving analysis of data to different tools, mainly manual or slightly automated with SQL. Elasticsearch and other current offerings are still too complicated and expensive for widespread use. Within the next couple of years we will see a significant surge of easily-accessible tools automating data analytics with ML.
Walter R. Paczkowski, Ph.D.
Walter R. Paczkowski is the founder of Data Analytics Corp. which offers services in Deep Data Analytics. He is a seasoned speaker, trainer, and consultant on data analysis which includes machine learning. He is also the author of two books on data analysis with a primary focus on market and pricing data.
“Since almost everything in our modern high-tech age is digital…”
So should marketing and, hence, marketing strategies be digitally oriented. This includes not only the implementation stage of the strategies (market delivery, promotions vehicles, price setting and communications, and so forth), but also the development stage of the strategies. This latter stage is where the background work must be done before the implementation.
Machine learning enters in both stages because it’s used to extract information from the vast amounts of data that almost all businesses are now collecting and storing. The data per se that are collected and stored are just stuff – meaningless things that take up space. Buried inside that stuff is information that has to be extracted to be actionable, insightful, and useful; the data per se are none of these. The analysis of that data stuff is where machine learning becomes important.
This is the #1 way marketers can apply machine learning – to extract information for their marketing decisions. Machine learning is the intersection of three disciplines: statistics, artificial intelligence, and computer science. It is NOT the extraction of information from data automatically done by computers (i.e., machines) but by humans thoughtfully and scientifically applying methods from all three disciplines.
For marketing, it greatly helps marketers gain insight into consumer problems, their willingness to pay for product attributes and features (i.e., pricing), their characteristics and the segments they belong to, and so forth. It also greatly helps the new product development process. Once a product is in the market, machine learning methods aid sales tracking, promotional campaign assessment, problem/issue identification, and competitive monitoring.
Roman Rabinovich is the VP of Business Development at Eventige Media Group, a Digital Marketing Agency based in NYC. Eventige is a growth-focused marketing partner to mid-market and enterprise -level businesses in North America. Romans’s background is in sales management, marketing management, and business analytics.
“One of the most popular ways marketers use machine learning to improve digital strategies is clustering algorithms…”
Unsupervised clustering helps to filter new segments of customers that may not be apparent to a marketing professional. The obvious customer segments do not pose a competitive advantage, so we need the help of clustering to find segments that are hidden within the variables. Marketers can use clustering for business development as well as market development to define new products for current customers, find new customers for current products, and also make small changes to current products to appeal to a different customer segment (based on the analysis of the variables). Having access to the right data is important for clustering, and it is a strong way to develop a digital strategy with a competitive edge.
“The saying ‘work smarter, not harder’ can be said about…”
Machine learning when used to build digital marketing strategies. By building a Customer Lifetime Value (CLV) model tailored to each unique client transaction, you factor in a customer’s frequency, recency, time as a customer, and monetary value to help build smarter AdWords and Facebook lookalike, remarketing, or negative audience lists.
Through the use of data science and the CLV model, you can create a paid media strategy that leans less on the marketer and more on machine learning to maximize Return on Ad Spend (ROA).
Amy has built and implemented multi-channel digital strategies for a variety of companies spanning several industry verticals from start-ups and small businesses to Fortune 500 and global organizations. Her expertise includes e-commerce, lead generation, and localized site-to-store strategies. Amy is a marketing consultant at AmyBishopMarketing.com.
“There are so many great ways to apply machine learning to digital strategies…”
I think it is a tie between audiences and bid strategies and, truly, there’s a pretty fine line between the two at this point.
Networks, data providers, and DMPs have so much data that they’re able to effectively use machine learning to create in-market audiences, lookalike audiences, and identity maps to offer really robust targeting features and, in turn, attribution and analytics.
Couple that with the ability to bid on consumers that are most likely to make a purchase, and you now have a really powerful lead-generation machine.
There’s also something to be said for all of the ABM and prospecting technologies that are built upon propensity to buy based upon machine learning. I expect those technologies will revolutionize the way digital marketing strategies are built within the next few years.
Andrew Miller, a.k.a. @AndrewStartups, has spent the last decade building growth for tech companies from Dubai to San Francisco. After his 3rd exit as head of marketing, he started his own boutique consultancy, AndrewStartups.com, and now assists companies while being a Digital Nomad on the road!
“The #1 way marketers can apply ML to their digital strategies is…”
Using tools like AdHawk and other automatic optimization tools for their PPC campaigns that use Machine Learning to optimize their campaigns. I’m personally against trusting algos and recommend using good old-fashioned elbow grease, but for a lot of companies the time saving is equal to the cost savings of optimized campaigns.
Matt Osborn is the Senior Marketing Manager at Apruve, a B2B Credit Network revolutionizing how businesses transact with each other.
“At Apruve, we have used machine learning in our blog…”
The AI program automatically collects data (CTA clicks, conversions, pageviews) and finds out what those blogs have in common. When we create new content, it notifies us what titles, keywords, length, and readability levels offer the highest ROI, and we write/edit to those specifications. Within two months we have seen a drastic increase in our conversion rate and a large uptick in immediate views per post.
Alexandra Cote works in Digital Marketing at Paymo.
“Sadly, marketers are not yet using machine learning to its full potential…”
Even worse, they’re not using it at all. The biggest advantage of implementing a machine learning system to your digital strategy is preventing risks. A machine learning tool can help you see exactly which areas are subject to facing risks, which obstacles you’ll encounter, when, and how often.
But the entire machine learning process starts with having a database. No structured data about your business, clients, and sales means you can predict future effects and their risks. Machine learning is also often less time-consuming and cheaper when it comes to lowering customer churn. Knowing what risks you’ll face can help you prepare to handle them before they occur. This will also ensure better overall customer service. Based on these results you’ll also be able to create intervention models to use when a risk actually happens.
Colette Nataf is the CEO and Co-Founder at Lightning AI. Previously, Colette was a User Acquisition Manager at MileIQ. After the company was acquired by Microsoft, she managed the Demand Generation team at Intercom. Colette, is now helping marketers scale their businesses through Facebook, Google AdWords, and other channels.
“I’m always looking for ways to leverage technology to help me do my job better…”
What I love about artificial intelligence is that we can actually teach computers to do the heavy lifting parts of our jobs — and that leaves the fun parts to us humans.
Marketing is all about testing – finding the best landing page, the best advertisement to run, or the best price to charge. But humans can only run a few tests at once. What if a computer designed the tests instead? Computers can run thousands of tests, all at the same time, and decide the winners and losers without any human biases. The #1 way marketers should be using machine learning is to leverage computers to create and rapidly test.
The human way is called A/B testing, and the machine learning model we can apply here is called multi-armed bandits.
These bandits are systems of weighting different options according to the probability of their successes and failures. This means that the computer can decide which landing page to show, or which price to show, and learn with every single user.
There are almost infinite ways that you can leverage multi-armed bandit models to help your marketing team, such as:
1. Find your highest paying customers and show different customers different pricing models.
2. Send email messages at different times based upon engagement rates of groups of users.
3. Advertise to your highest value customers and don’t spend money on low-value customers.
Tyler Foxworthy is a Chief Scientist at DemandJump. He is a mathematician and computational scientist who works on a wide range of problems in machine learning and market optimization. Tyler’s previous experience includes leadership and research roles in academia, biotech, and management consulting.
“Businesses must realize that location is everything…”
As businesses started transitioning money from traditional advertising to digital channels, began to put less emphasis on where the ads were placed and focused on audience-based targeting. Marketers who understand the importance of location will be ahead of competitors. Start by utilizing audience insights and an AI-based customer acquisition platform. This will help you acquire customers to better understand the “network effect” and better spend your money. This will help maximize returns and virtually eliminate the risk for wasted spend.
Lauren Hilinski is the Digital Marketing Specialist at Shred Nations.
“Right now, the #1 way marketers can apply the effects of machine learning to their digital strategies is…”
Through their content strategy. Having your business show up in the search results is the goal, and understanding how Google is using machine learning to rank web pages will help you beat the competition.
Google is using machine learning to analyze the intent of a web query. 20% of the time, Google interprets buy searches as learn searches. To beat your competition in a world where Google bypasses your direct signals and decides for itself what the user wants and the whether or not your web page meets that need, you need to write content that meets the needs of the user on their journey.
Look at the existing SERP for your keywords and see what types of pages Google is listing (informational, transactional, navigational, etc.) to get a better understanding of how Google interprets the query intent, and write to fit that.
Google uses machine learning to better solve the users intent, and if you can give Google the answer to the user’s problem, you will beat your competition.
Eagan Heath is the Owner of Get Found Madison.
“If you track your conversions for…”
Your Facebook or Google ads and you get a good number (like a few hundred or thousand) every month, you can set your ad delivery to optimize for conversions. Google and Facebook can deliver your ads based on the people, devices, and times that are most likely to work. This is the real power of their ad algorithms.
Beth Bridges is the Owner of eBridge-Marketing.com, a digital marketing consultancy.
“The #1 way marketers can apply machine learning to improve their digital strategies is to…”
Remember that machines don’t think like people (at least not yet). They are getting closer every day, but they don’t take shortcuts, use made-up abbreviations, or assume that they’ve been explicit about their intent. Ask anyone who’s gotten some seriously weird recommendations from their Netflix (which uses machine learning to suggest shows). Or keep this in mind when you’re trying to perfectly match keyword suggestions in your SEO strategy. RankBrain is getting smarter every day, but 15% of searches are still something entirely new. Humanity is still messy, chaotic, and unpredictable, so use your own brain and experience to inform your digital strategies as well.
Lee Gallagher is a Big Data Aficionado, Author, Internationally Recognized Speaker Data, and Founder of Arts & Analytics, Inc. His passion is to help small business and non-profits have access to data driven insights that will drive top-line revenue. Previously, using predictive analytics was cost prohibited; but today it can be quite affordable and easily accessed.
“Machine learning itself is a tool that continually learns from data to predict or solve a future state…”
Problem: Marketers need to align how customers are consuming their data and be sure to have the table ready when they are hungry. In other words, Marketers need to have a presence at the right time and in the right channel with the right offer for the “hungriest” buyers. Previously, marketers “hoped” they had the right strategy to reach their prospects, but today ML enables actionable insight to continually improve or adjust their strategies. Today it is called “Moment to Moment” marketing vs. yesteryear’s “Spray It and Pray It Works” strategy.
Simple example: ML provided the marketer deep analysis of their current customers and then predicted the potential channels of their interest. The marketer shifted their display networks strategy into channels that interested their buyers. In other words, moved their spend from advertising on travel sites to health conscious food sites, enabling a presence in the right channel, to the right customer, at the right time. That is Precision Marketing.
Jordan Harling is the Digital Strategist at Wooden Blinds Direct. He has been a professional marketer since 2011, working across a number of sectors including charity, digital and ecommerce.
“Machine learning is perfect for increasing the speed and efficiency of typical A/B testing…”
One example of this is in optimizing hero images on your landing pages. Usually this is quite a time-intensive task, requiring you to create a number of different options and iterate through them, then waiting for the results over a number of days before starting the process again. Machine learning takes the time and effort out of this. All you need to input is a modular image which can be configured in a number of different permutations (a combination of different backgrounds, call to actions, and content) and the criteria you wish it to be judged on. Once you’ve done that, you can let the algorithm do its thing and it will provide you with the optimal results.
Cory is a serial entrepreneur, with his first exit at 20 in the IT space. He is the Co-founder and CEO of Qoints, and has 10 years of experience in the digital marketing space, leading campaigns for Crayola, Energizer, Ford, Nestle, and more.
“Machine learning is being adopted at breakneck speed…”
With nearly every industry seeing the benefits of artificial intelligence to become more effective and efficient. The marketing world is no different, with executives and professionals applying machine learning to every facet of their business. In particular, machine learning has been useful in influencer marketing, becoming a compelling way for brands to reach new customers. On average, businesses generate $6.50 in revenue for each $1 invested in influencer marketing, and this return justifies the investment in one of the hottest trends of 2018.
A prime example of an application of machine learning regarding influencer marketing is how here at Qoints we’ve been able to collect over 150,000 past sponsored influencer posts to create a dataset of success metrics for brands to build the ideal influencer profile.
Harnessing the power of our neural network, we can identify the likeliness of an individual’s influencer success for a campaign. Computers are good at repetition and following instructions, however AI applies to neural nets as they provide the opportunity for tech to learn to recognize new patterns.The purpose of using a net, rather than a singular classifier is to tackle more complex and bountiful inputs, like that required in a businesses’ influencer requirements.
The network of classifiers determines a certain characteristic of the data that is inputted, where it cycles through layers of the network which then outputs the data with a classification attached. These classifications are modified by unique weights and biases attached, creating different unique reactions to predict values. Training examples teaches the neural net to produce accurate predictions. Through training our data set to indicate successful influencers, we are capable of checking the compatibility between company and influencer, trying to find the ideal fit.
Businesses can apply machine learning tactics to their business, boosting their marketing efforts with the help of artificial intelligence. Through allowing machines to do a majority of the repetitive or analytical work, it frees up marketers to continue to craft innovative strategies to stay ahead of the competition.
Connie started her business strategy and marketing consultancy Flywheel Associates as well as her internationally downloaded podcast, the Strategic Momentum Podcast, to help business leaders identify how to work through challenges commonly impacting organizations today.
“I’ve interviewed experts in AI and machine learning across the country and there’s one common sentiment around implementation…”
The most important thing to do actually happens before applying machine learning. Many marketers and tech leads overlook the step of creating a data strategy. Without a strategy, you can’t ensure your data is really working for you, and you’ll probably end up wasting time and money. Simply hiring a data scientist to develop machine learning algorithms won’t solve your problems and generate business. If you want machine learning to work for your digital strategy, train your marketers to speak data, so they know how to connect the machine learning outcomes to specific business objectives. Then, create a hypothesis and test and learn with small, specific use cases (i.e. responding to customer FAQs) before graduating to more involved systems. So the #1 way marketers can apply machine learning to improve their digital strategies is to start with a strategic foundation.