In a recent blog post on practical applications of artificial intelligence for email marketers, we tried to define what AI means for email marketing. Although some AI claims are more hype than reality, we believe it offers some great practical applications for the future.
We work on some of the most complex and advanced email programs in the world. We’re always looking for ways to innovate and improve our customers’ programs. After working with OneSpot for years, we believe the personalization technology company is changing the game with machine learning and its practical application for email programs.
I recently sat down with several members of the OneSpot team to get their perspective on the AI landscape and what their technology offers the modern marketer.
1. It seems as if the email industry has talked about delivering a 1:1 experience for over 10 years. What is OneSpot’s unique differentiator?
“Personalization” is used to describe different things in the email space, from adding someone’s name to the subject line and swapping out imagery based on location, to standard segmentation. OneSpot defines personalization as a series of decisions made at an individual, 1:1 level. We refer to it as individualization.
Three things make our approach unique:
- Our content decisioning engine is powered through machine learning. We focus on selecting the most relevant content, not on how to deliver the content.
- We built a solution that can be used in the email channel, not an engine for the email channel. Our system makes decisions independent of the delivery channel. So, a request for individualized content from a web page is treated almost identically to a request for individualized content from an ESP.
- We use machine learning concepts on both the input and output, employing natural language processing to ingest and analyze content into our system. This makes onboarding and integration lightweight and also removes the potential for human bias or error in categorizing content.
Many personalization solutions claim to be machine learning based but still rely on humans to categorize content. Machine learning allows for a true understanding of an article’s substance, not just the keywords the author associated with it.
One important note: OneSpot leverages billions of pieces of data to make decisions about which article, video, recipe, infographic or other content piece is right for a specific person in that specific environment. Because a user’s interests change, we also update our users’ profiles in real time so that the content they receive is always tailored to them.
2. In Part One of our post, Trendline defined machine learning and its three methods for how machines learn. What method of machine learning does OneSpot use to deliver personalized email content?
We use several machine learning methods to individualize content, from natural language processing and image tagging to collaborative filtering. All of these are typically classified as unsupervised approaches, where the system looks for patterns but does not ruthlessly optimize toward an outcome.
Supervised approaches for content recommendation tend to do well for maximizing outcomes like clicks but fall short in individualizing for each user. This is why you see so many click-baity recommendations on the web.
3. OneSpot is changing the way the email industry looks at personalization and content marketing, which some would say is an amazing innovation. However, innovation takes many forms and sometimes gives only false positives in the form of short-term results. What long-term results have you seen with clients?
As you would expect, we certainly see a lift in core engagement metrics like click-to-open rate. But, what has been most rewarding is that we are fundamentally changing the way customers execute email programs. We’ve simplified operations while improving results and have enabled our customers to provide a much better experience for their subscribers.
Instead of slaving over five-layer-deep “If/Then” statements to deliver relevant content, our customers can type something like “Treating soreness after running or exercising” into our tool and generate a content-rich email on that theme.
What’s more, subscribers will receive specific content within that general theme based on their behavioral history.
For years, the last thing anyone in the email industry wanted to be called was a “batch-and-blaster.” We’ve not only redeemed “batch and blast,” we’ve also in many cases turned it into the right way to send.
We call it “Batch and Individualize:” one campaign, one segment, millions of versions.
4. Companies have long promised personalization across channels but made it difficult and confusing for marketers. What makes OneSpot different?
We’ve simplified the approach overall, specifically when it comes to implementation.
We’re using existing content on the existing site so that the customer doesn’t have to do anything different when publishing new content to have it added to the mix.
We also make decisions based on observed behavior. So, we record every interaction the individual has with the content on your website and then use that data to fuel our algorithms.
This means marketers aren’t forced to make subjective decisions about how a particular workflow should look or what content readers might want to see based on the segment where they’re assigned.
5. What is the toughest hurdle for marketers when they first speak with you? What fears do organizations have with your solutions?
Although our data shows our algorithms can choose content more effectively than people can, marketers are most worried about relinquishing control. OneSpot has tools that allow editorial teams to retain as much editorial control as they desire. Still, we see almost universal apprehension over not knowing exactly what content is going to be served to each individual.
To get the most benefit from individualization, editors must be OK with sending the content that their subscribers are interested in, even if it’s not what the editors want. It sounds easy, but it can represent a significant shift from operating norms.
6. Content is and will always be a key component of email marketing. How does OneSpot envision staying out in front of the proliferation of content in email? What is your vision for the future?
From OneSpot’s inception, our focus has been on working with long-form editorial content. You can find a product recommendation engine on every corner, but we see a surprising void for tools that were purpose-built for content. Content plays such a key role in email marketing that we saw that void as a huge opportunity.
We see content reaching beyond the newsletter and being incorporated into all forms of email communications. Automation and machine learning allow email marketers to add individualized content to stale welcome series and triggered messages. Transactional messages with content related to purchases can now add more value than just serving as receipts. Subscribers can expect a better overall experience.
7. Where do you see the greatest opportunity for machine learning and email in the next 2-4 years?
We are just seeing the tip of the iceberg with applying machine learning to email. We believe we’ll ultimately see it applied in layers. Right now, we are interpreting existing content and selection of content for individuals. Other companies are addressing when and to whom to send.
The real opportunity is in combining both layers, where the result is knowing when a particular person is most likely to be hungry for a certain kind of content and then to deliver that content at the right moment.
The best part is that every incremental application of machine learning to an email program should make that program more efficient. Email marketers can benefit from both increased engagement and reduced cost — simultaneously.
8. We have been led to believe that data needs to be in one place for AI to work. But, you get customers up and running in weeks. How does that happen?
Ease of implementation has been a pillar of our product strategy since Day One. Over the years, I have seen countless companies suffer from data paralysis. Some say that you need all the data before you can act. The reality is that there’s tremendous opportunity to act on just some of the data.
OneSpot’s approach is to keep the data where it belongs and then to combine it as necessary. We collect behavioral data through our script tag on the customer’s site. Even with moderate site traffic, it takes only weeks — not months — to get enough data to be able to detect inherent relationships and patterns between users and content.
We can combine our observed data with our customer’s profile-level data as part of the request for content. So, we never have to combine all the data in one system. If you tell us that a specific user is a female and has checked a box saying she is a vegan, then we can combine that with her behavioral history to select a highly relevant content set.
Start a conversation with Trendline?
Interested in learning how Trendline can take your email marketing program to the next level? Start a Conversation Today.