Artificial intelligence (AI) has long been a part of science fiction, from books, to TV, to the big screen—but now, it’s part of reality, and at the center of the marketing discussion. At our latest Speaker Series, "AI: The Future of Customer Acquisition," our panel of Red Door subject matter experts shared insights on the impact of AI, and how it affects the way brands reach their customers.
If you missed the event or just want a refresher, don’t worry, we’ve got you covered! Read on for some of the panel’s insights on AI’s influence in areas such as e-commerce, content, search, programmatic advertising, personalization, and more. A full livestream of the event is also available for viewing on our Facebook page.
What is "Artificial Intelligence"? How is it being used today, in a broad sense?
Artificial intelligence refers to machine intelligence vs. human intelligence, or the science and engineering of making machines intelligent. In today’s landscape, AI is present in just about every field there is, most prominently in the military and medical fields. Not only do we have self-driving drones using AI, but AI can help doctors more easily understand which solutions to apply for a specific cancer treatment.
Are there different categories of "AI"? Where does "Machine Learning" fit into all this?
Most of AI can be categorized as “narrow AI” (also referred to as “applied AI”). Narrow AI applies machine learning to a specific field, and is able to make suggestions and decisions for us. Then there is “general AI,” which involves having a human level of intelligence with the distinct ability to reason. General AI is what we are most familiar with in the context of pop culture: these machines can think, react, and make the right choice.
One of the areas where we've seen machine learning for some time is in programmatic media. What exactly is programmatic advertising and how is it related to AI?
Essentially, programmatic media buying is using machines to buy ads instead of using humans to go through a more manual process of setting up ads and negotiating each buy for each individual site. Programmatic leverages AI technologies to bid across inventories, display, social channels, and video, and allows advertisers to reach their target audiences more effectively. Rather than manually placing ads on a site and hoping the target audience visits that site, it allows brands to reach their target audience regardless of the site they’re on. This works by using algorithms to analyze a person’s behavior and likeliness to convert, based on thousands of factors, and then adjusting bids.
Only five years ago, we would have to spend time pulling reports with information such as time of day, sites, and noted changes. Algorithms are now so advanced that marketers don’t need to do that anymore. AI will find what is performing best, while learning more every day.
With all that said, programmatic still needs human oversight to ensure it’s set up properly. If you’re setting it to seek out clicks, it would likely go after sites with click bots, so you’ll need to make sure it’s measuring real outcomes.
What are some of the latest developments in programmatic? How can more advanced marketers continue to evolve their media efforts?
One of the most exciting things in programmatic is using it for reaching people on television, podcasts, and out-of-home billboards. Programmatic billboards are especially interesting with the unique ability to analyze traffic patterns and then serve the ad that’s most relevant to people in that area. Brands can then use mobile retargeting to serve this ad to people who have previously seen it.
People tend to think that traditional forms of media—such as out-of-home—are declining, but they’re actually expanding, and overall spend is expected to increase in the next couple of years with so many new features. Dynamic creative is also interesting in this context. For example, if a brand chooses to show a specific ad within the conditions that it’s snowing and a specific sports team has won, dynamic creative can create that experience. This isn’t specific to out-of-home, but this is where things are headed.
What about Google and Facebook? What are the more interesting things these platforms are doing now with AI?
Facebook has obviously been under fire for featuring unwanted content such as propaganda and fake news. With this in mind, it’s not surprising that Facebook has chosen to focus AI engineering on facial recognition, analyzing text, and enhancing their environment to create a safer place for both users and advertisers. As an example, Facebook is using AI to scan users’ posts to analyze whether that user is showing signs of depression. The AI will then alert a human that can offer real help.
In the case of Google, “Google lens” is a unique tool that uses phone cameras to analyze surroundings and serve consumers contextually relevant content. This has the potential to have a tremendous influence on e-commerce, meaning brands will need to have a search strategy set in place.
How has Google's ranking algorithm evolved to leverage AI, and how are SEOs keeping pace with these changes?
Artificial intelligence is part of SEO whether we like it or not, and it all started in 2014, when Google acquired a company called “DeepMind” out of the UK, and unleashed the technology on the world under the name of RankBrain.
RankBrain serves a number of purposes for Google’s web search platform, but one of the biggest issues it solves for web search is the elimination of a “one-size-fits-all” algorithm.
Google crawls the web and gathers information about every page, including word count, title, links to the page, etc., and the algorithm is a formula to decide how much each of those metrics is weighted when ranking search results. The problem Google was having was that each time they made an algorithm update, it would make search results within one area better, but often at the expense of another vertical.
RankBrain allows AI to adjust the weighting of these ranking factors and, in essence, create an infinite number of unique algorithms, each tailored to a specific vertical or topic. This allows Google to use different ranking factors for “open a bank account” than it does for “fantasy football sleepers,” and everything in between.
However, these evolutions of technology have complicated things on the keyword optimization side, and SEOs need to adapt their tactics quickly to keep pace with the changes. For us at Red Door, we’ve developed a process that leverages AI to reverse-engineer this process.
Everything Google looks at for rankings is all publicly available. Our tools can see the same links that Google sees, measure keywords and content in the same way, and dive into every other aspect of a public webpage that is used as a ranking factor. So, if we look at a target keyword opportunity, we can assemble all these stats for every URL across the first few pages of Google, both for the client’s site and competitors’. We also know the current ranking order of all of those websites. With these two pieces of data, we can start to crunch the numbers until we have a combination of ranking factor weightings that explain the order of results we are seeing. This allows us to identify the gaps between our client’s site and the sites that are currently ranking better, giving us visibility into what keywords to optimize and what levers to pull.
Is there a relationship between SEO and the voice technology movement? Where does voice search fit into all this?
AI has a big role in helping determine the intent behind a query. Right now, Google reports that about 15% (500 million/day) of web searches use queries that have never been searched before, and it needs to figure out what results to show. New queries can be driven by all kinds of things: daily news events, individual searcher preferences, evolving pop culture trends, etc. Google is using AI to process the intent of the meaning behind unknown queries and finding content to match the topic that searchers are seeking. To process these query types, Google uses both implicit and explicit signals:
-
Explicit signals: Is the whole query unknown, or just parts of it? Is its known entities in the search string? How do the unknown parts modify the known parts?
-
Implicit signals: Time of day, device, location, search history, age of searcher.
In many ways, this is all culminating in a move toward voice search, or “keyboardless search,” which is a broader category that voice search fits into.
Google is already using AI to understand content. When keywords include an inferred or explicit question, Google has been showing a “featured snippet” from one of the ranking pages, which pulls out some of the page content and aims to answer the question behind the query. This is a SERP feature that has been popping up for more and more keywords over the past year, and Google is actively collecting user feedback on whether a particular answer is helpful or not (both directly and in the form of user behavior).
This content is already being used to feed Google Assistant, which is the software being used with Google Home and other voice devices. As Google collects more feedback, the accuracy of results will improve, and they can start feeling more confident in providing a single answer to a question, rather than the traditional 10 links to choose from.
How is AI helping marketers better understand their customer?
The short answer: AI means we will know way too much. Facebook is a good example of this. With so much information about its users, the platform can actually make predictions about what users are likely to be interested in, and then serve them specific content based on that. However, we’ve seen Facebook get in trouble for this in the past, and now we’re moving more toward consent-based personalization and customer segmentation.
If you’re already doing customer segmentation, you might be basing it on a number of “if, then” statements, creating a rule-based segmentation environment, or maybe you’re operating on a cluster-based analysis, pooling consumers together based on attributes you know about them. The problem with this analysis is that it’s pretty static: you do it once, but then end up introducing new products, a new market, and a new generational target within the next three months.
One of the key things here is that with AI, we’re going to get more into performance-based segmentation and ongoing segmentation, incentivizing brands to start getting more data on their customers and pooling that together. We’ve been talking about big data for a long time now, but what it resulted in was gathering information, and not really doing anything with it. With AI, we can start getting more contextual data. Customer data platforms (CDP) categorize this data in real time, allowing you to see all your data in a single space and make more accurate, unbiased predictions.
Beyond website personalization, are there other ways to use machine learning to reach customers with personalized messaging?
Website personalization typically starts with developing a hypothesis, creating personalized content for that hypothesis, and then testing it against a control group.
AI flips this process on its head: we’re still setting parameters, but we can now more easily add different offers, different images, and set up different work streams, letting the machine adjust the experience in real time. When customers come to a website, it’s making a decision for you based on so many different attributes using look-a-like audiences. AI also facilitates product recommendations, which is especially useful in moving people from learning to shopping within a website.
What’s the impact on creative?
Robots can’t draw. Dynamic creative is a good option for combining images and experiences together, but AI is not skilled at choosing the right creative concepts. While narrow AI is rules-based, AI will not be able to recognize a subjective aesthetic.
Are machines coming for our jobs?
Ultimately, AI will be making our jobs easier by taking away the amount of data we need to analyze ourselves. Yes, repetitive tasks are likely to be replaced by AI, but there will also be new opportunities for trainers and explainers. These people will need to provide the data and train the models, and then explain the output to other humans. The more we probe into the future of marketing, the more we open up the possibilities.
Have additional questions about artificial intelligence, and how it plays a role in your customer acquisition strategy? Feel free to drop us a line on social, or reach out to us on our site—we’d love to chat.