Digiday+ Research: How marketers are using AI to target ads, recommend products and provide customer service

With e-commerce sales soaring in recent years, thanks in part to pandemic shutdowns, and the impending death of the third-party cookie driving a need for new data collection capabilities, more marketers are turning to natural language processing (NLP) and data-driven personalization to automate customer service and gather data for ad targeting. 

Over the last five years marketers have increased use of NLP, mainly in the form of chatbots, while keeping use of data-driven personalization steady, according to Digiday+ Research. Although both technologies fall under the larger banner of artificial intelligence, each offers marketers distinct benefits. 

Data-driven personalization gives marketers the ability to tailor product recommendations and target ads using consumer data. Experiences can be contextual for general recommendations or hyper-personalized with data from an individual consumer. But marketers that offer hyper-personalized experiences run the risk of violating data privacy laws, something they must keep in mind as regulations increase.

“When [AI] is done right, it is extremely helpful,” said Fred Gerantabee, chief experience officer at FGX International, an eyewear company owned by EssilorLuxottica. “There is no single magic bullet, but AI allows us to look through mountains of data and make sense of it in a way that a human being couldn’t.”

Marketers mainly use NLP in the form of chatbots to streamline customer service responses and increase e-commerce sales. The pandemic accelerated marketer use of chatbots when stores were closed and e-commerce sales hit all-time highs. Other forms of NLP, like social listening, allow marketers to track brand mentions to make business and product decisions.

With marketers increasing their use of chatbots and maintaining a focus on data-driven personalization, marketer adoption of AI is likely to continue at a consistent pace, as long as the technology continues to evolve. Additionally, the market value of AI in worldwide marketing is expected to reach more than $107.5 billion by 2028, according to Statista.

For this report Digiday+ Research surveyed 388 industry professionals at organizations including agencies, brands, retailers and publishers to uncover how they’re currently using data-driven personalization and natural language processing — and how they plan to incorporate the technologies in the future.

Digiday also conducted interviews with executives from the following companies and agencies:

• FGX International
• General Mills
• Huge  
• IPG Media Lab
• Levi Strauss & Co.
• Syte

• Data-driven personalization has remained a priority for marketers over the last five years, with 68% of respondents saying it is a priority in 2022 compared to 67% in 2017. 
• Marketers primarily focus on targeting an individual user for a hyper-personalized experience (64%) versus creating a customized experience based on contextual information for a more widely appealing experience (36%).
• The percentage of marketers who use some form of NLP, primarily chatbots, has increased from 31% in 2017 to 44% in 2022.  
• Chatbots are the primary type of NLP marketers use because of their practical applications for customer service. Social media listening and text sentiment analysis are second and third. 
• Marketers who do not currently use data-driven personalization are evenly split on whether to invest in it in the future, with 48% of marketer respondents saying they will invest in the technology and 48% saying it is not relevant to their business.
• Marketers see less potential for expanding their use of NLP. Fifty-five percent of marketers who do not currently invest in NLP said that it is not relevant to their business and only 36% plan on investing in the technology in the future.

Before marketers can personalize ads and product recommendations, they first must establish the building blocks of personalization by collecting or acquiring data and constructing personalization tools. When it comes to both of these things, marketers predominantly use third-party vendors. 

Digiday’s survey found that 45% of marketers use third-party partners, such as Adobe Analytics and Google Analytics, to gather data and 53% of marketers use them to build applications for data-driven personalization. Using a combination of in-house offerings and third-party solutions was marketers’ second most common means to both acquire data and build personalization tools, with about one-third of respondents selecting that mix for each category. Marketers were least likely to exclusively use in-house options for data collection and application construction.

FGX International’s Gerantabee said collaborating with external vendors, in addition to conducting its own market research, gives his company access to more robust data pools. “When you’re working with established retailers, like Target or Walmart, their dot-coms serve as great test-and-learn places, great ways to get data,” he said. 

“Most of these franchises have done a really good job of connecting the buy online and pick up in-store [experience]; and the buy in-store and manage customer service requests/online reviews [aspect],” Gerantabee added. “As things become really unified from the channel perspective. It makes the insights a lot more concise because then you really understand who that consumer is, regardless of whether they walked into the store.”

In fact, after Google announced the deprecation of the third-party cookie, more retailers like Target and Walmart began pitching themselves as data providers to brands. According to a pitch deck Target sent to clients in 2019, Target had profiles of more than 147 million customers at the time, including data on where its shoppers made purchases, how they paid and how often they shopped.

Steve Croll, general vp of technology at Huge, a digital agency focused on creative growth acceleration, said with numerous third-party data gathering options available to marketers, as well as their own in-house capabilities, many companies are still refining data gathering best practices. 

“We still see a lot of our clients struggling with the fundamentals of gathering and organizing their data,” Croll said. “Organizations are struggling to organize their data into a single store that they are able to then apply to personalization. That’s where we’re seeing the majority of our work. It’s the nuts and bolts of gathering and obtaining the data that is already rich within your environment, and putting it into a repository that’s actionable.”

Although survey results showed that data gathering is primarily done by third-party vendors, the type of data marketers are most often collecting is first-party data. More than 80% of respondents said they’re collecting data directly from customers. 

First-party data is typically collected in-house when consumers register their email addresses and phone numbers on a website, when they fill out forms and surveys, and when their browsing and purchase activity is tracked across websites and apps, often through the help of loyalty programs. That likely accounts for more than one-third of marketers saying they use both in-house and third-party teams to gather data.

Carter Jensen, global e-commerce lead of direct to consumer at General Mills, said his company has established an in-house data team for first-party data collection, which allows the company to provide more effective ad targeting and product recommendations. “It’s a great way to collect data about [consumers] and to understand what they need and how we can serve them better,” he said.

“First-party data from a manufacturer standpoint sometimes gets a bad rap,” Jensen added. “Many people would say, ‘Why do you need that? You’re going through retailers already. How would that work?’ What we’ve learned is, as we begin to understand the consumer better, we’re able to provide better or friction-free conduits for purchase for that user.”

Digiday’s survey results found that, indeed, marketers primarily use the data they’ve gathered — be it through in-house or third-party vendors — to personalize ad experiences (72% said they do this), indicating an emphasis on increasing product awareness and purchase conversions. Marketers also use data-driven personalization to prioritize content and product placement (66%) and to recommend products (46%). These second and third uses can have similar marketer applications for increasing product sales. Brands may also use them to enhance customer retention within loyalty programs and improve in-app experiences by tailoring shopping recommendations and updating user interfaces.

In fact, brands may escalate their use of data-driven personalization for loyalty programs as deprecation of the third-party cookie becomes a reality. For a mass brand like Maybelline, for example, which sells through retailers like CVS and Walmart but does not have its own DTC e-commerce play, the elimination of third-party cookies is a pressing issue. As a result, in November 2020, the company launched a rewards program called Maybelline Express to both capture more first-party data and better understand its consumers.

“We wanted to own the connection with our consumers and personalize experiences, and become less reliant on third-party media companies to provide consumer data for us,” said Marnie Levan, Maybelline’s vp of integrated consumer communications. “In order to maximize our competitive advantage, which is our ad dollars, we needed to build out our customer databases.”

Key takeaways:

• 45% of marketers use third-party partners to gather data and 53% of marketers use them to build applications for data-driven personalization.
• Marketers were least likely to exclusively use in-house options for data collection and application construction.

Marketers have kept their prioritization of data-driven personalization almost exactly the same in the last five years, largely due to its usefulness for creating consumer profiles with which they can target ads and recommend products. Sixty-eight percent of marketer respondents said data-driven personalization is a top priority in 2022 and 67% said it was in 2017.

When it comes to ad targeting, the agencies marketers rely on to generate creative content and buy and plan media placements are becoming savvier at using data-driven personalization. And industry experts say use of AI has grown as demand for ad targeting using first-party data increases. AI can help automate some ad agency processes, from creative to segmentation, and potentially help agencies optimize their ad spend by lessening monotonous work.

One of the ways agencies are testing AI that’s most relevant to this report is in the creative process, using data points to drive content. For example, Tyson and Mindshare recently partnered with intelligence startup socialcontext.ai to create a tool called Impact Index. Using this index, the organization is measuring the social impact of its editorial content in the Black community. With these types of insights, AI generators are able to create relevant content in minutes or even seconds. Agencies are also using AI tools to produce writing, music and other visuals.

TBWA Worldwide uses AI generators to produce content for some of its clients. Ben Williams, TBWA’s global chief creative experience officer, described AI as part of a current “creative revolution” that can inspire teams and clients to think differently. “I believe the true power of these technologies and tools is not to replace creatives, but rather make us more efficient,” Williams said. 

Marketers themselves are using data-driven personalization beyond ad content to make the shopping experience easier for customers by recommending specific products they believe will meet consumer needs. And it goes further than just higher top-line conversion rates: If consumers are happy with items they’ve purchased, they are less likely to make returns, which helps retailers improve their bottom lines.

Big-box store Walmart, for example, has continued to put data-driven personalization at the forefront of its recent wave of tech upgrades. In September, Walmart said it was updating its app and website by adding the ability to curate and share gift registry lists and adding a filter to view EBT-eligible products. Walmart already offered a styling feature that suggested items to complete an outfit, like bags and shoes. 

“By making it personalized, it helps narrow that decision-making for [consumers] because we know who they are, we know what they’re shopping for,” said Brock McKeel, svp of site experience at Walmart eCommerce. “We get to know them.”

Key takeaways:

• Marketers have kept prioritization of data-driven personalization steady in the last five years, with 68% of respondents saying it is a top priority in 2022 versus 67% in 2017.
• Marketers use the data they collect to create consumer profiles with which they can target ads and recommend products.

Customized user site and ad experiences can stem from different paradigms: When it comes to tailoring creative and product assortment for consumers, marketers primarily focus on targeting an individual user with a hyper-personalized experience based on individual data (64%) versus targeting page context based on tags and other semantic data (36%).

Hyper-personalized experiences are highly customizable and can conform dynamically to customer traits and preferences. Feed-based content, like that offered by Instagram and TikTok, represents the high water mark of hyper-personalization. Instagram shows users posts based on their activity, including their connections and what posts they have liked, saved or commented on, among other things. Users can also tell Instagram if they don’t want to see a suggested post and the post will be removed from future feed suggestions, further increasing the algorithm’s intimate knowledge of individuals and their preferences.

The most straightforward way marketers can hyper-personalize content is to directly ask customers what products interest them. Online personal styling service Stitch Fix asks customers to fill out a comprehensive survey about how closely a series of outfits matches their personal style before signing up for the service. Stitch Fix then tracks which items customers keep or return to better learn about a specific consumer’s preferences.

Stitch Fix also developed a mobile game called Style Shuffle that asks players to give a thumbs up or down to a series of accessories, clothes and footwear. The company then created an algorithm to predict other items a customer might be interested in based on what players with similar tastes like. 

“We use the data today to essentially build products in-house and also share with our brand partners,” said Stitch Fix’s CMO Loretta Choy. “The reason why we do that is because we want to help our partners grow their brand as we grow our brands and [to connect] their products to our clients.”

Third-party AI platform Syte provides visual AI technology to its clients, including luxury retailer Prada and diamond jeweler De Beers, in order to help them hyper-personalize clothing and accessory recommendations for customers. The technology predicts other products a consumer might like based on pictures the shopper uploads to a brand’s website or images they click on while browsing. For example, if a user uploads or clicks on a picture of a dress in a particular style, length and color, they will be shown dresses similar to it as they continue to explore the site. 

Syte CEO Vered Levy-Ron said this form of hyper-personalization helps marketers adhere to privacy standards while also potentially increasing sales with targeted product recommendations. “A user is personalized not because they’re a woman or they live in Israel [for example], but because of what they chose [through an image],” she said. “They are shown similar dresses in their size, eliminating the dead ends, which improves the user experience and boosts sales.” 

General Mill’s Jensen said investments in data-driven personalization technologies can benefit a marketer’s bottom line, but the efforts are only truly useful if they’re providing a good customer experience. “You have the financial benefits of becoming more efficient and targeted and having a greater ROI on any type of immediate spend,” Jensen said. “But you also have consumers who expect the results of that data. If it’s a bad experience, naturally they’re going to avoid it, or realize the poor experience and fall away.” 

Additionally, IPG Media Lab’s executive director, Adam Simon, said many marketers are not taking the next step and asking consumers for feedback about whether hyper-personalization efforts are effective. “Very few companies are forming a good feedback loop of consumer input into that personalization,” he said. “Once you design an algorithm to personalize an experience, you should want consumers to give you feedback on how well that’s working.” 

Across the board, data-driven personalization also needs to surmount limitations to current automation capabilities. While certain portions of the experience can be dynamically tailored, such as ads promoting products a consumer was recently browsing, other portions must still be manually updated in the backend of the website, like manually tagging products with categories. 

In 2022, almost half of respondents said they use a combination of manual and automated updates to personalize content and another 36% said content personalization is fully automated. Only 15% of marketer respondents said their user experience was purely manually updated. 

Hyper-personalization requires a more finely-tuned update system and infrastructure since personalization is done dynamically at an individual level, while contextual targeting relies more on manual updates, like product tagging. Survey results indicate more marketers skew toward using automatic updates, either in combination with manual updates or on their own, supporting findings that marketers use hyper-personalization – at varying levels – more than they use contextual personalization for targeting.

Key takeaways:

• Marketers primarily focus on targeting an individual user with a hyper-personalized experience based on individual data (64%).
• Hyper-personalized experiences are highly customizable and can conform dynamically to customer traits and preferences, but they can run the risk of violating privacy regulations.

While marketer respondents said the majority (64%) of their company’s creative or product assortment is hyper-personalized toward an individual user, one-third (36%) of respondents said more of their content is tailored based on the context of the page on which it will appear —  making contextual personalization a less popular form of personalization.

Contextual personalization broadly targets consumers by gathering information from the page a shopper is viewing, rather than collecting specific information about the user themself. Since contextual personalization does not require precise individual consumer information and far less data needs to be collected, brands often find it easier to create contextually relevant content. However, because marketers generally don’t own the web pages through which they’re advertising or selling goods, contextual personalization is less frequently available to them beyond general site and category targeting.

For marketers who do use contextual personalization, the most common examples of contextually curated content are “you might also like” or “complete the look” recommendations on an e-commerce page. These sections display products tied to a shopper’s assumed interests based on items they’re currently viewing and are intended to entice users to click on additional products, ideally resulting in increased purchases. 

A few years ago, influencer platform and app LTK, then known as Like To Know It, began offering a feature that let customers search for products and gave them results showing posts from influencers who belonged to the app’s network. The idea was for customers to “go down the rabbit hole” searching for products, be inspired by items shown next to the original item, and continue to search and purchase. 

“We wanted it to include all this information from our influencers and what they said about the products in our search data,” said Ben Newell, then vp of product at LTK’s parent company, RewardStyle, and now senior director of product at Stitch Fix. “People could search for ‘Christmas,’ which isn’t a product but is something that people would search. … Once our consumers got the hang of searching for jeans or boots, we pressed them on what else to search for … .”

Of course, without individualized data informing it, contextually personalized content may not always resonate as well with consumers as hyper-personalized content. Brands also cannot react in real time to adjust product ads and recommendations in response to shifting customer needs and preferences. 

In general, marketers who are satisfied with sticking solely to contextual personalization run the risk of being left behind competitors. IPG Media Lab’s Simon said, for example, that Netflix — known for its contextual recommendations rather than user-specific suggested content — may struggle to keep up. 

“We used to talk about how great [Netflix’s] recommendations were, but they haven’t evolved as fast,” said Simon. “It’s one of those things that might start to become a differentiator in the streaming space — how you surface new content to users, what those interfaces look like, and how smart [streaming services] are about recommending what [users] should watch next when they finish a show or movie.”

FGX International’s Gerantabee said contextual personalization needs to advance beyond simply suggesting related products. “I would argue that good AI will ultimately be able not just to say, ‘You bought this and maybe you want this,’ but it may understand where [consumers] run into hesitation points,” he said. “It may be able to say, ‘You abandoned this product, … here’s a video or a supplemental piece of content to help you make a more concise and confident decision.’” 

“A lot of examples [to date] have been very vertical: video to deliver video; shopping to deliver things to buy,” he added. “The magic will be when they combine, and I haven’t seen too many real world examples of that just yet.”

Key takeaways:

• 36% of marketer respondents said more of their content is tailored based on the context of the page on which it will appear —  making contextual personalization a less popular form of personalization. 
• Contextual personalization can be easier to implement, but may not resonate as well with consumers.

As marketers strive to personalize product ads and recommendations, the top consumer data points they use to power this data-driven personalization are site behavior (e.g. clicks, time on page and cart abandonment rate), location and browser/search history — all data points that marketers can acquire by tracking customers’ site activities and IP addresses.

Seventy-four percent of marketer respondents said they look at site behavior as a main attribute around which to personalize ad creative and other aspects of the user experience. Sixty-nine percent of respondents use location and 52% consider browser and search history, indicating that marketers are primarily focused on how, where and what users are seeking online.

The rise of retail media has made all of these data points more readily available and more directly relevant to brands. The recently announced merger of Kroger and Albertsons grocery chains, for example, means the combined entity may be able to challenge Walmart, the second biggest player in the retail media space, by offering brands a greater breadth of data to use in ad targeting.

“The combined company will be able to reach an expanded national audience of approximately 85 million households nationwide, fueling growth in alternative profit businesses such as Retail Media, Kroger Personal Finance and Customer Insights,” the two companies said in a press statement.

Over the last five years, marketers have shown more interest in retail media because of the massive size of the ad business e-commerce giant Amazon has built. Currently, Amazon leads the retail media ad landscape and took in $31 billion in ad revenue last year. Consumers also shifted many purchases from physical stores to e-commerce during the pandemic.

During the same time period, marketer respondents have increased their use of most of the consumer attribute categories Digiday first asked them about in 2017, with the exception of age, hobbies and occupation. Financial/purchase data – accessible at higher volumes than ever thanks to retail media networks – and browser history had the biggest increases of more than 10 percentage points each.

And the application of this data doesn’t end at checkout. FGX International’s Gerantabee said he finds the most value from analyzing after-purchase data to understand if his company’s personalization efforts have been effective. “If we sell 100,000 things and we meant to sell 100,000 things … it’s only half the picture,” he said. “Really understanding if we were successful [happens] post purchase.”

“It’s less looking at sentiment about the product reviews,” Gerantabee added. “It’s looking at returns and all of the different data points to really understand whether people liked the product. Because, realistically, if you sell 100,000 of something and 60% of it was returned, that’s not a success. You should not be making decisions to repeat that. There are a ton of data points we use to truly understand how consumers are gravitating toward [a product].”

Key takeaways:

• The top consumer data points marketers use to power data-driven personalization are site behavior, location and browser/search history.
• This indicates that marketers are primarily focused on how, where and what users are seeking online — data points that the rise of retail media has made more available.

Consumer data collection is top of mind for most marketers as third-party cookie deprecation moves toward reality, however slowly (Google delayed depreciation of the third-party cookie again until 2023). Marketers are weighing how best to collect customer data for personalized ads and product recommendations while respecting customer privacy. They’re also grappling with the advent of more stringent data privacy laws – and their enforcement.

In May 2018, the European Union enacted the General Data Protection Regulation (GDPR) law, which sets guidelines for collecting and using an individual’s personal information — companies must tell consumers how they’re using their data, and if and when it is breached. Similarly, the California Consumer Privacy Act (CCPA), also enacted in 2018, gives consumers the right to know what personal information a business collects about them, and how it is used and shared. It also gives consumers the right to delete personal information collected from them and to opt out of the sale of their personal information.

Some marketers, like DTC brand Doe Lashes, are finding ways to work around data tracking regulations by gamifying data acquisition. Similar to the Style Shuffle game Stitch Fix offers, Doe Lashes’ “Lash Quiz” asks consumers to answer a series of questions about their eyelash and makeup use. At the end, they receive a lash recommendation. The quiz has had a 4.6% average conversion rate, according to a case study by software company Octane AI and e-commerce platform Klaviyo. 

Octane AI president and co-founder Ben Parr said quizzes allow brands that haven’t yet fully prepared for data tracking regulation changes to collect data without violating privacy rules. 

“Consumers don’t want to be tracked without their consent, but they do want personalized experiences,” Parr said. “Through quizzes and surveys, consumers know they’re volunteering that information. ​​It’s a fundamentally different process.” 

Other companies have their own in-house data teams that focus not only on consumer data collection, but also on privacy best practices. General Mills’ Jensen said his company’s in-house data team helps the company adhere to privacy regulations by giving consumers the option to opt out of data collection. 

“Regardless of how big or what brand, [consumers are] one click away from controlling the data that General Mills has,” Jensen said. “You’re one click away from being able to manage it and one unsubscribe away. If you don’t want to hear from General Mills again … there’s never any confusion or convolution.”

Clothing company Levi Strauss & Co. uses browsing history and clickstream data to personalize search results and make product recommendations for shoppers. However, its chief AI officer Katia Walsh said it also makes opt-out options clear to consumers. “At the end of the day, the data is that of the consumers,” Walsh said. “They have the right to dictate what happens to it.”

“As long as companies responsibly and clearly explain to consumers what they’re doing with that data, in a language that’s efficient, fast, and simple to understand, the consumer is basically in the driver’s seat,” she added.

Walsh said she believes that adherence to privacy regulations and data-driven personalization can work hand-in-hand. “Personalization and privacy do not have to be at odds,” she said. “However, it is an utmost responsibility, and indeed an imperative for companies to ensure they provide that information to consumers and give them the opportunity to exercise the right to have control over their data.”

While many brands have strong stances on data privacy, the reality of what consumers experience when they visit some websites can be considerably different. In a 2021 study, Digiday visited websites of the media properties (mainly owned by publishers) listed in the Comscore 50 — a ranking of top U.S. digital media properties — to measure dark pattern behavior. Dark patterns are tactics used to get people to take an action they may not want to take or realize they are taking. In data privacy, it is often in the form of data collection notices not having a clear opt-out process or highlighting opt-ins with more prominent buttons while obscuring opt-out options.  

Only 11% of sites Digiday measured showed users a privacy notice, according to study results. Of that 11%, 62% of sites had policies of implied consent, meaning that site users who did not configure their privacy settings or explicitly opt out were automatically giving the sites permission to continue gathering data on them. 

While the dark pattern behavior Digiday found on many of the media property sites is not explicitly illegal, the overall strategy does not reflect the transparent, customer-first policies that most brands expressed.

Key takeaways:

• Marketers are weighing how best to collect customer data for personalized ads and product recommendations in light of increasing privacy regulations like the CCPA. 
• Some marketers have found workarounds by gamifying data collection through quizzes, and many offer the option to opt-out of data collection.

While marketers have kept their use of data-driven personalization steady in the last five years, their use of NLP applications — primarily in the form of chatbots — has increased in the same time period, up from 31% in 2017 to 44% in 2022. Many brands view customer service as an extension of marketing, with chatbots playing a role in that as well. 

Unlike in 2017, the broader NLP category considered in our 2022 survey includes multiple technologies, including social media listening and text sentiment analysis. But chatbots are still the most prevalent form of NLP used by marketers, with the majority of brands using NLP saying they use chatbots in 2022, as seen in the chart discussing social listening and sentiment analysis farther down.

Marketers use chatbots more than other types of NLP because of chatbots’ practical applications for customer service and data collection. Chatbots can be used not only to answer straightforward pre-programmed questions, but also to update an order status, offer product recommendations, collect email addresses and ask consumers questions about product preferences to be used later in targeted marketing campaigns or tailored experiences.

Marketers’ need for chatbots to handle increasingly complex requests and chatbots’ overall role within marketer-consumer interactions has evolved and expanded with time. In 2017, 52% of marketer respondents said they offered chatbots mainly as an information resource to provide product education for customers. 

In 2022, that role has shifted strongly toward customer support capabilities, with 65% of respondents saying they use chatbots to provide consumer help, and only 32% of marketers using the technology as a consumer information resource.

The shift toward using chatbots for customer service can be attributed partially to when in-person shopping paused at the height of the Covid-19 pandemic. Brands turned to hosting chatbots on their sites to replicate some of the in-store experience provided by sales clerks — such as recommending products to customers — while at the same time gathering customer data and boosting e-commerce sales.

Toronto-based menswear retailer Harry Rosen deployed its chatbot in November 2020 to handle an increase in shopper inquiries corresponding with a rise in e-commerce sales. The chatbot initially focused on answering simple, pre-loaded questions such as inquiries about store hours and promotions. By March 2022, it also provided other customer service offerings like live updates for order statuses and cancellations, exchanges and garment alterations.

“When we first decided on AI chat, there was some concern about it taking over the customer service reps’ jobs,” said Manuel Maciel, executive vp at Harry Rosen. “But we wanted to integrate it to reduce repetitive tickets for our associates. … At one point, we had a seven-day backlog of tickets and morale among associates was awful.” The chatbot also provides customer service agents with context about an inquiry before they interact with the customer, Maciel added.

As the pandemic presumably winds down and consumers and marketers alike have grown more accustomed to using chatbots for complex questions and tasks, marketers will likely continue their use of chatbots for customer service interactions. FGX International’s Gerantabee said he’s seen the value of using chatbots in seasonal and other high-traffic shopping cycles as well.

“Sometimes there are not enough people to throw at a problem, particularly when it comes to customer service,” Gerantabee said. “So holiday seasons, anytime where bandwidth is high, there’s nothing more frustrating than waiting for a person to answer a simple question. I’ve advocated a lot for using chatbots as an upfront way to provide customers the easiest method of self-serve, beyond the standard FAQ or digging through a site for answers.”

“Very simple tasks, such as looking for where an order is, answering questions about warranties, these are things that probably don’t need human attention,” he added. “If it does get to that point, at least [businesses] have mitigated some of that upfront. There is both a convenience aspect for the customer and for the business.”

Key takeaways:

• Marketer use of NLP applications — primarily in the form of chatbots — has increased over the last five years, up from 31% in 2017 to 44% in 2022.
• Marketer use of chatbots primarily for customer service, instead of as a consumer information resource, increased when in-person shopping paused at the height of the Covid-19 pandemic.

Similar to data-driven personalization applications, which marketers primarily rely on third-party partners to construct, NLP applications are largely built by external vendors. Fifty-three percent of marketer respondents said they use a third-party vendor to build their NLP tools. Less than one-third of marketers use a combination of in-house and third-party solutions to assemble NLP tools, and only 18% build applications exclusively in-house.

In 2020, hair-care brand Amika used an external partner, chatbot marketing company Automat, to build its NLP tool, a chatbot named “Ace.” Amika previously had a live chat function on its website, which gave customers the ability to make simple inquiries about an order. However, the company used the Ace chatbot to personalize the shopping experience and offer product recommendations by asking customers questions about hair type, products they were looking for and their hair goals. After customers completed the questionnaire, Ace made product recommendations. The chatbot also offered the opportunity to sign up for Amika’s newsletter.

Robbi Webb, then Amika’s senior director of e-commerce and now its vp of e-commerce, said at the time that the questionnaire had a 57% completion rate and had given the company valuable insight into current and new customer product needs. Amika used that information for ad targeting and email capture, paving the way for personalized email marketing. 

Similarly, in 2020, David’s Bridal hired third-party vendor Apple Business Chat to build its AI-powered concierge chatbot dubbed “Zoey,” which the company hoped would attract more customers to its stores post-bankruptcy. David’s Bridal used the chatbot to collect questions and insights from customers within a Zoey conversation and then pass them along to an in-store stylist. When a client arrived for an on-site appointment, which could also be booked through Zoey, the stylist had already been briefed on what the customer was looking for.

Holly Carroll, then vp of customer service and contact center operations for David’s Bridal, said the chatbot helped alleviate some consumer tension around dress shopping. “We know that the customer is telling us that this is a really stressful situation for her, whether it’s planning a wedding or picking the perfect prom dress,” she said. “By quieting the chaos and having a seamless handoff between digital and brick-and-mortar, we can really connect the conversation.”

Despite mainly having external vendors build their NLP tools, marketers primarily host chatbots —- their most-used form of NLP —- on owned-and-operated platforms. Forty-seven percent of marketers said they host chatbots on their own websites and apps in 2022.

Many brands began increasing their overall website commerce capabilities, including hosting chatbots on their sites, during the pandemic when consumer use of e-commerce options intensified. The aforementioned chatbots in service of David’s Bridal and Amika, while built by third-party vendors, are hosted on the companies’ own websites.

While owned-and-operated platforms have grown in popularity among marketers as chatbot host platforms, Facebook Messenger, once the top choice for chatbot deployment, has decreased in use since 2017. Only 42% of marketer respondents who used chatbots said they use Messenger to deploy them in 2022 versus 88% of all survey respondents in 2017. 

Messenger’s chatbot application launched in 2016 and drew a lot of initial interest from brands that used it to provide customer service, commerce options, product discovery and entertainment. However, interest declined shortly thereafter when Facebook revealed that bots failed to respond to about 70% of requests.

To combat decreasing use of its Messenger chatbot application, Facebook added new features to the service in 2019 that were intended to increase lead generation and commerce. Capabilities included allowing businesses to send users to Messenger by swiping up on Stories ads or clicking on ads that go straight to Messenger. Burger King, for example, used the service to give customers the ability to plan a meal and pay for it through Messenger. Facebook’s efforts seem to have worked to some degree, as almost half of publishers employing chatbots said they are still using Messenger to deploy them in 2022.  

Key takeaways:

• 53% percent of marketer respondents use a third-party vendor to build their NLP tools, but 47% percent of marketers host chatbots on their own websites and apps. 
• Facebook Messenger, once the top choice for chatbot deployment, has decreased in use since 2017. Only 42% of respondents use Messenger to deploy chatbots in 2022.

While chatbots are still the most prevalent form of NLP used by marketers — outpacing other forms by more than 20 percentage points — the more diverse NLP applications of social media listening and text sentiment analysis are closing in, coming in second and third place, respectively. Fifty-six percent of marketers said they are using NLP for social media listening and 46% are using it for text sentiment analysis in 2022.

Social media listening gives marketers the ability to track consumer conversations and mentions of products and companies on social media platforms, and analyze them for insights about customer traits, behaviors and brand perceptions. Those insights can then be used for everything from developing and tweaking products to improving consumer targeting.

Earlier this year, cosmetics maker Huda Beauty used feedback from social media to reformulate and relaunch its concealer, previously known as Overachiever Concealer. Brand founder Huda Kattan said her company decided to adjust the product’s ingredients and release it under its new name, the #FauxFilter Luminous Matte Buildable Coverage Crease Proof Concealer, after taking users’ online comments into consideration. 

“When we did a lot of social listening, there were definitely callouts [around] areas to improve,” Kattan said. “With all of our makeup, I’m always going to be looking for areas of improvement. There are new technologies and new ingredients. You always have to be working with manufacturers to find out, ‘If I were to tweak this and make it 1% better, what would I do?’” Kattan said the brand ultimately adjusted ingredients to meet consumers’ suggestions for improvements.

In addition to listening on social platforms like Instagram, TikTok and Facebook, many brands are turning to Reddit to understand customer sentiments and target ads. The social news aggregation, content rating and discussion website has more than 52 million daily active users and offers a prime opportunity for marketers to tailor ads to specific user groups, according to Will Cady, then Reddit’s head of creative strategy and now its global director of in-house agency KarmaLab.

“Reddit has a rich array of over 100,000 different communities each with their own culture,” Cady said. “For brands to find a way into that can feel complicated at first, but it’s actually remarkably easy when you identify the culture that your brand can be at home in.”

“The best practice for brands entering Reddit is to lead with listening first,” he added. “[Reddit] provides this incredible opportunity to look at people’s passions and how they are expressing them in these contextual environments.” Marketers can then place ads they know will resonate with users within those same communities. In February 2019, pizza chain Little Caesars brought back its “Pretzel Crust Pizza” campaign that culminated in crowning a Redditor “Chief Pretzel Officer.” 

According to Little Caesars’ CMO Jeff Klein, while the campaign was spread across social media, Reddit was a core aspect of its success. “Our engagement on Reddit, while just one aspect of this multimedia campaign, played a huge role in our ability to directly communicate with customers in a real-time authentic way,” Klein said. 

While social media listening gives marketers the ability to monitor brand mentions and discussions, text sentiment analysis — marketers’ third most used form of NLP — helps them understand the feelings and attitudes behind consumer comments and conversations.

Text sentiment analysis identifies the emotional tone behind user-generated text that references a product — an online product review or survey comments, for example — and classifies those references as positive, negative or neutral. Marketers use insights around product sentiment to make a range of decisions from determining long-term business strategies to modifying product formulations or selecting the most effective wording for an advertisement or product page. 

Insights gained from text sentiment analysis can also help marketers address areas of concern that may be eliciting negative emotional responses, such as ineffective product return policies, customer service glitches and prices consumers deem too high. 

Key takeaways:

• 56% of marketers use NLP in the form of social media listening and 46% use text sentiment analysis, making them the second and third most used forms of NLP. 
• Social media listening and text sentiment analysis give marketers the ability to track product mentions and the emotional tones behind them to make product decisions. 

Five years ago, in Digiday’s previous iteration of the emerging tech series, IPG Media Lab’s Simon (then its strategy director and now executive director) said, “Not enough brands are investing in personalization and AI or machine learning.” Times have clearly changed: Marketers are now prioritizing AI technologies, particularly chatbots for customer service and data-driven personalization for ad targeting and product recommendations.

Levi Strauss & Co.’s Walsh said she’s noticed more brands putting funds into data-driven personalization in particular in the last five years. “There’s no doubt that companies across industries are investing in all things technology,” Walsh said. “Consumer-facing companies are certainly making sure that investment goes into technology, including AI that has a personalization effect. Whether it’s telecommunications, financial services, retail or consumer goods, it’s really an increasing focus … with a particular emphasis on personalization.”

As mentioned earlier, nearly 70% of marketer respondents prioritize data-driven personalization, while only 23% do not. The relatively few marketer respondents who do not currently invest in data-driven personalization were equally split on whether they plan to invest in the technology in the future, with 48% saying they plan on investing in it and another 48% saying it’s not relevant to their business. In fact, when Digiday asked those marketers who don’t currently invest in data-driven personalization why they don’t invest in the technology, the majority said data-driven personalization is not relevant to their business. 

A little less than one third — 30% — said the cost of building and implementing personalization tools was the biggest barrier. In-house personalization tools, in particular, can be costly for marketers to implement since they often require funding a separate data team with specific expertise in the technology. The fundamentals of organizing data for use by applications is still a struggle many marketers are facing as well.

Syte’s Levy-Ron said personalization tools also need to bridge the gap between online and in-store shopping to create smoother customer experiences. “The next frontier in AI is driving better user experiences, both in the digital world with visual search and in the physical world with digital tools [that are used] in-store,” she said.

When it comes to NLP, despite an uptick in marketer use of chatbots over the last five years, the majority (55%) of marketer respondents who do not currently invest in the technology said it was not relevant to their business and only 36% plan on investing in it in the future. 

Marketer concerns about NLP’s usefulness was especially apparent when marketers who do not currently use NLP cited lack of relevance as their primary reason for not investing in NLP (53%). Only 23% said the cost of building and implementing the technology kept them from investing.

With consumers returning to in-store shopping and in-office work, any pandemic-induced pressure marketers may have felt to deploy chatbots as a replacement for human interactions may be dissipating — likely an element contributing to marketer respondents saying that NLP technology is not relevant to their business.

Another factor possibly contributing to apprehension around business relevance is that chatbot technology isn’t yet as sophisticated as it could be, as noted by IPG Media Lab’s Simon. “We’re finding that NLP is not advancing as fast as we want it to with things like voice assistants,” he said. “Our ability to have natural conversations with them really is happening very incrementally. It seems clear that it’s going to be probably another decade before you can just start talking to Google, Siri or Alexa the way that you would talk to a human.”

However, Huge’s Croll remains optimistic about NLP’s nearer-term potential, particularly in the form of chatbots. “If we can organize our content and our data appropriately, then chatbots could actually have the ability to learn on the fly and evolve in the background of an organization of a particular conversation,” he said. “Access to the content that is going to be ingested by a large language model and surrounding models gives it the intelligence it needs to evolve and become relevant to ongoing conversation. I think you’re going to see that commercialized very quickly.”

As marketers continue supplying their AI tools with more data, regulatory and ethical questions around data privacy remain a top concern. Marketers will have to continue juggling competing demands — adhering to ever-changing data privacy laws, finding inventive ways to collect customer data in a cookieless world and respecting consumer wishes for data transparency while serving them targeted ads and product recommendations — as they expand their use of NLP and data-driven personalization.

Key takeaways:

• Marketers who don’t currently invest in data-driven personalization were split on whether to invest in it in the future, with 48% saying they plan on investing in it and another 48% saying it’s not relevant to their business. 
• Despite an uptick in marketer use of chatbots, the majority (55%) of respondents who do not currently invest in NLP said it was not relevant to their business. Only 36% plan on investing in it in the future.