Testing, testing
Last year Umbraco acquired uMarketing Suite, allowing web content managers to analyze first-party data to understand which content works best for different users and make the necessary tweaks to optimise engagement and boost conversion rates. The resulting Umbraco Engage product was hailed as finally enabling personalization at scale.
Finding the holy grail
Enterprises have had personalization capabilities for almost two decades. However, the additional copywriting required to cater to different user preferences made it too costly to scale.
Generative AI services like ChatGPT have cracked that nut and made tailored marketing campaigns and personalized content at scale financially feasible.
Some marketing industry experts argue that generative AI is already making next-generation personalization possible, whereby enterprises can create individualized content for users.
How AI makes personalization possible
It’s often said that good salespeople are the best listeners, because they pick up on cues to understand what customers really need. In the AI realm, machine learning acts as that star salesperson, picking up on user queries, previous purchases, and interactions to understand intent and deliver better product recommendations. This is the secret sauce that makes personalization possible.
Recommendation engines process vast amounts of data and apply item-to-item collaborative filtering, so that each customer is offered a personalized selection of information or products that are most likely to meet their requirements, based on their previous engagement and purchase history.
Tailoring content to users
As a CMS platform provider, we are conscious that 69% of a potential customer’s content journey has already taken place before they reach your website. They’ve been funnelled towards your business by media, advertising, social media, and personal recommendations. Once they’ve arrived at your site, it’s vital to make that content relevant.
AI provides an opportunity to make that final leg of the content journey a valuable use of your customer’s time. AI can be used to validate content, transform it to a customer’s preferred format, and provide helpful suggestions that relate content and relevant media such as product images, instructional videos, or related articles.
Reading the room – adapting to emotions
One of the more recent and fascinating aspects of AI is its ability to empower marketing teams and web designers to adapt content to suit customers’ moods. Sentiment analysis allows web designers, marketing teams and content managers to adapt the way information is presented according to the way customers are feeling.
Last year, one of our digital agency partners, true Digital, undertook a study to understand the impact of website design on users’ emotional states.
Using Odaptos, an AI tool that analyses facial expressions to detect people’s emotional state, true ran a series of tests to see how adapting website layout and content affected test subjects’ emotional reactions.
A really interesting finding was that Odaptos registered higher happiness levels in test subjects visiting a website selling beer. However, when the same people visited a website selling wine, the AI detected higher levels of fear and anger. The research team at true inferred that, because people associate wine knowledge with higher social status, test subjects were worried about buying the ‘wrong’ wine and appearing uneducated, or lower class.
The web design team at true found that by adjusting how web content is presented, so that wines are grouped into white, red, and rosés, and are as clearly labelled as they would be in a supermarket, these negative emotions could be reduced. By using AI to analyse web user sentiment, true found that web developers can proactively help to allay these common fears and create a happier online experience for people which encourages more online purchases.
The team’s findings are shared in a whitepaper, ‘The value of emotion in ecommerce.’ Page 13 states, “When interpreting others’ emotions, verbal communication accounts for just 7% of our understanding, compared to 55% through facial expressions and 38% through tone of voice. Self-reported feedback only gives us part of the picture. New AI technologies make it easier for us to measure participants’ non-verbal emotional responses and expressions, allowing researchers to gain quantifiable data and insight on the behavioural and physical responses of users.
All you need is a webcam and a platform trained to interpret facial expression, gesture analysis or eye tracking, to capture and understand users’ emotional states, their attention, interest, and engagement. We worked with Odaptos, whose proprietary video conference tool uses artificial intelligence (AI) and Natural Language Processing to analyse the seven universal emotions – happiness, sadness, surprise, fear, anger, disgust and neutral.”
The research from true revealed that users bring ‘emotional baggage’ to certain types of websites. As an example, test subjects arriving at financial websites were already exhibiting facial expressions that indicated fear, or anger. The challenge for web designers and content managers is to adjust for those negative emotions and create a personalized and positive experience.
What’s next?
A plethora of AI tools are being explored by marketing teams and web content managers in the quest for individualization. That’s why it’s important for us to remain flexible by providing a platform that facilitates the most appropriate AI tool for each use case.
_____
Matt Sutherland, head of Technology at true summarises this position perfectly when he comments, “It would be impossible to not acknowledge AI and its role in the future of content marketing, and content management. With the influx of generative AI tools, the open-source CMS platforms that adopt, adapt and use AI in the most secure, strategically sound and responsible ways will no doubt succeed in this huge evolution for content generation. Advantageously, open-source CMSs should, by nature, be able to allow for choice as to how you use AI, what you rely on for your AI tooling and the models that your AI tooling is learning from, and I believe that level of control will make the difference between successful, or unsuccessful adoption of AI for all CMS platforms.”
