Crystal Taggart: Using AI to Optimize SMB Digital Marketing – Episode 40

Podcast: Play in new window | Download

Subscribe: Apple Podcasts | Google Podcasts | Stitcher | TuneIn | RSS

Crystal Taggart helps small- to medium-sized businesses improve their business results with artificial intelligence. AI is rapidly emerging as a viable technology that even small businesses can use to optimize their marketing efforts.

Crystal and I talked about:

• her company, Atlas AI, and its mission to empower SMBs to use artificial intelligence to better promote their businesses
• how AI lets you personalize your communication with your customers – way better than the algorithmic marketing we’re doing now
• how personalized content experiences may affect SEO and other internet marketing practices
• how small businesses can use AI – first, they need to become more data-driven
• how most companies can benefit from stepping up their use of data – to identify customer problems and create content that helps the customer solve them
• how shockingly limited Google Analytics data is – and how to fill the gap – to tie that limited information to specific customers – to learn what they care about and how to personalize your messages to them
• how to optimize your content creation based on actual customer behavior on your website
• how to integrate affordable information resources like Pipl, which charges 5-15 cents for a customer profile – “If you’re spending a dollar [with AdWords] to get your customer through the door, and then you don’t spend 5 cents to learn more about who they are, you’re not doing yourself any favors.”
• how to use ZIP code information, which is free, to start to build a demographic profile of your customer that can help you predict their likelihood of becoming a customer
• the hazards of gender and racial profiling
• how chatbots can be a great way to engage customers – but most of the current ones are “pretty dumb,” just picking up keywords
• the importance of monitoring how people use your chatbot to avoid unintended consequences (like the infamous Microsoft chatbot that became a racist troll in just 24 hours)
• how to create your own chatbot
• how tools like flow.io can help non-programmers build their own chatbots – but you need to invest some time to use them properly
• how training data sets and AI are used to predict chatbot answers
• how AI can help SMBs with churn predictions – the likelihood of a prospect becoming a customer and then becoming a loyal repeat customer
• how to combine the power of customer segmentation with A/B testing and other metrics to drive real business decisions
• how AI can link specific customers to the drivers that actually engage them
• her theory of how Google will handle the emergence of AI: that they’ll reward AI-based dynamic solutions – you’re giving the searcher a better answer if you do that, and Google appreciates that and will reward companies that do it
• how marketers – who are creative, interesting storytelling folks – have been forced to become bean-counters in the current analytics-driven online marketing environment
• how AI can help marketers return to their storytelling roots
• how her goal is to help people measure the effectiveness of “every post, page and tweet”
• how to link personas to actual individual customers and then customize your messages to them
• how imminent much of this AI marketing technology is – “months away, not years away”
• how AI discoveries can focus your marketing spending
• how AI is a journey – first, you measure things to make sure things are working; second, you personalize the customer experience; and third, you prepare for technologies like personalized video content
• how to align emerging AI technologies with your content strategy
• how now is the best time to make mistakes

Crystal’s Bio

Crystal Taggart is an entrepreneur and data scientist. She is the founder of Atlas AI, an Open Source marketing analytics platform that enables SMBs to leverage AI and analytics to improve inbound marketing results, improve conversions, improve segmentation, and improve customer experience.

Video

Here’s the video version of our conversation:

Transcript

Larry:
Hi, everyone. Welcome to episode number 40 of the Content Strategy Insights podcast. I’m really happy today to have with us Crystal Taggart. Crystal’s an entrepreneur and a data scientist, and also, more recently, the founder of Atlas AI, a company that focuses on AI. Well, I’ll let Crystal tell you a little bit more about what she does there. Welcome, Crystal.

Crystal:
Thank you for having me, Larry. So, I am super excited about what we’re working on. So, our mission is to enable small business to take advantage of AI technologies. So, we think that SMBs have to make the most of every marketing dollar, and we provide the science to maximize their time and budget with easy and affordable AI solutions with our open source platform.

Larry:
Cool. I saw Crystal speak at WordCamp Phoenix a few weeks ago, and I really ran up after her talk because I’m like, “Wow. Are we really at the point where AI is no longer just a buzzword and a thing that only enterprises get to do? Is it actually well understood enough and practicable enough that even us little businesses can work with it?” So, it sounds like that’s the case. Well, let me start just about the highest level about content and content strategy . . . much about just at that level, the implications for content creation planning, management, all that stuff with the AI?

Crystal:
Yeah. So, I think that we are really in a changing world, right? So, to me, the best use case of AI is personalization, right? So, if you rewind the clock 25 years, the sales and marketing process was about being personal with your customers, right? With automation and digitization, it’s actually gotten away from that, right? Everything’s a commodity. Everything is a plug and play. Everything is kind of a “you fit into this box,” right? So, we’re focused on eyeball marketing today, and so, what AI gives us the capability to do again is to actually create personalized experiences based off of what a person needs, not what is going to grab the most attention based off of a numbers game. So, to me, that’s super exciting about how the world is changing, and it’s going to be interesting the things that kind of happen as a result of that. We’re around the corner from personalized websites and personalized content experiences. So, what’s going to happen with Google if they’re trying to index content that changes based off of whoever is coming to the site, right? So, it’s kind of an exciting time to be here, and for marketers, if they’re not already adopting these types of techniques or researching into them, they’re going to see rapid changes very, very quickly.

Larry:
Yeah. You’ve already scared me, and I’m scared on behalf of all my small business and friends who run small agencies in the sense that one of the biggest challenges is from the WordPress world is all those agency folks, they struggle a lot with content and how to get it. Now, it’s like, “Oh, yeah. Now I have to create infinitely larger amounts so you can personalize it,” and there are huge implications you just kind of hinted in there, like SEO and other established marketing things. But I gather from what you’re saying that we’re at a point where … Well, tell me, I guess specifically, how small businesses can take advantage of AI.

Crystal:
Yeah. So, to me, AI is a process, right? So, it’s not a plug and play. It’s actually kind of a mindset of how you operate as a business, right? And I would argue that small businesses don’t invest the time to become a data-driven organization, right? So, if you take a step back and look through kind of what that means, it’s really about creating some research upfront, which is what you do, right? So, researching upfront what your customers want, what they’re responding to, what your competitors are doing, and then, creating a strategy around that, and most companies don’t do that. They just kind of throw stuff on the wall and see what sticks, and they come up with hypothesis. So, to me, if you take a step back, and you say, “Okay, let’s go analyze the data and go see what’s actually working,” right? And we can go build on that.

Crystal:
That’s kind of the first level of AI because what you’re doing in that scenario is you’re segmenting your customers, who’s good, who’s bad, taking that data. What are they interested in? What are they looking at one your website? Which, we all have that data. It’s hard to get to. We’re actually working on solving that problem, making that easy to get to, but if you invest the time in doing that analysis, then you can see what types of problems your customer is looking to solve, and then, you can create more content around those specific problems, really focusing on the best customers for your business.

Larry:
Right. That’s interesting because I don’t think I’m alone in having had moments or episodes where I’ve been frustrated with Google analytics and other tools to just go in there and like, “Great, what do I do with this information?” And so, I assume that’s just maybe one data set that you would operate from to come up with these sort of research tools to say, “Oh, and if you look at this data clearly, these people are interested in that.” Because there’s a lot in there you can segment by demographics and other factors. Can you give me an example, I guess, of how you’re doing that now?

Crystal:
Yeah. So, there’s a couple different things. So, we actually looked at integrating to Google analytics to get the data out, right? Almost every single website has Google analytics, and the limited amount of information they give you is shocking, right? So, it’s kind of broad strokes of the truth, which is [crosstalk 00:06:09] 1000 people or 5000 people hit your homepage. They went to these other two screens and left. They came in from these sources, and those are fine, right? But the thing that we’re missing and the thing that we’re working on is being able to tie that information to an actual customer, right? So, being able to say, “This customer came in and looked at these specific pages.” So, you know what they actually kind of care about, and you can do that if you do a lot of coding with Google Tag Manager and things like that, which it’s fine. The problem is that there’s browsers, such as Duck Duck Go, that are coming along that are blocking those Google injections, right?

Crystal:
So, you start getting more anonymous data coming through. So, to me, if you have a website that people are going to, this is kind of part of the problems with kind of what’s happening with the GDPR and the European Union is that if somebody comes to your website, you have a right to know what they looked at and where they traveled through your site, right? And the GDPR seems to think that your website should be private, right? And so, what we’re doing is we’re actually tracing back that data back to a specific customer, so you could see, “Hey, this customer came in. They looked at these pages. They signed up on your contact form, or no, they didn’t,” and you can go make decisions on how to adjust your content based off of those trends, and so, with that, I think you’re able to make better content strategy decisions based off of what’s working as opposed to kind of the broad brushstrokes of what Google analytics gives you. The other piece of this is there’s so much data out there that’s so inexpensive to get to that if you’re not investing in that, it’s really kind of a shame.

Crystal:
So, if you’re doing kind of AdWords strategies or things like that, and you’re driving people to fill out your content form, there is companies like … One of the ones that I love is Pipl, P-I-P-L, and for 5 to 15 cents, you can learn more about your customer. So, if you’re spending a dollar to get your customer through the door, and then, you don’t spend the five cents to go learn more about who they are and what they’re about, you’re not doing yourself any favors. You’re just another person. There’s free resources out there where you can do amazing things if you can get somebody’s zip code. Zip code is one of the best socioeconomic demographic indicators out of any piece of data that you have, and that data is available for free.

Larry:
Right, and that’s . . . because a lot of what you just said gets into privacy issues and stuff, but people being aggregated at a zip code level, that’s a nice anonymizing device that still gives you, as a marketer, plenty of information to work with.

Crystal:
Yeah, absolutely.

Larry:
Are you balancing that, throughout all of this, kind of balancing those privacy concerns with your … because you’re a data scientist. With that hat on, you just want more and more and more, right?

Crystal:
Yeah. So, to me, I think that demographics are key for understanding your customer, right? So, I use the zip code. I have a customer now who we looked at their data, and we found that people who made over $60,000, people who had two kids, people who are interested in one product, they were more likely to be a good customer. If they targeted people who were under those demographics, they were likely to cancel, or they were likely to have a bad account. So, I think that you have an obligation. So, you’d use that data to your advantage. I think the problem is that if you start racially biasing people, that’s a problem. There’s been a number of different issues with companies like Pokemon, companies like LinkedIn, who were racially profiling or gender profiling based off of the data that they had. So, I think that you have an obligation to us it. I think that you just have to monitor it to go make sure that it’s working for you.

Larry:
Yeah, and that occurs to me, that could be a whole episode just talking about those issues, but I want to get back more of the practical applications of this. For example, one of the things you’re an expert on is chat bots, and I think that’s probably the most accessible thing to people now about AI, I think, is they interact with a chat bot. They kind of know what it is. Tell me a little bit more about the application of that in your work.

Crystal:
Yeah. So, I think that chat bots are fantastic because they’re a new way to grab the attention of your customers, right, or your potential customers. I think that they’re a new way to engage. I think the reality is that most of the technology solutions out there … The chat bots today are pretty dumb, right? So, they are really just picking up keywords and then answering a reply, and that’s roughly how 98% of the chat bots work out there today. So, I think if you’re going to deploy a chat bot, again, those technologies kind of work. They’re okay, but the thing to do is you really have to monitor the responses and the interactions. So, you have to actually invest the time in researching what is actually happening with your chat bot because if not, you can really get into some areas where you’re getting unintended consequences. Microsoft had an issue where when their chat bots went wild and-

Larry:
Oh, I remember that, and it instantly became a racist troll or something, right?

Crystal:
Within 24 hours, it became a racist, Nazi-loving, feminist-hating troll, yes.

Larry:
Yeah. Yeah. And as you talk about that, I think that it seems like we’re at a point where you can buy a chat bot platform or something, I guess. So, I think a lot of my listeners are small or medium-sized businesses who are likely to … and they’re all discerning consumers as well, but there is this thing of when a new technology comes, it’s like, “Oh, you’ve got to have this. You’ve got to have this chat bot.” Can you talk a little bit about what you would look for if you’re looking to add that kind of functionality to your web presence? How would you vet possible solutions?

Crystal:
So, when I’m looking at solutions, I’m a huge fan of open source, right? So, I always look to see if there’s an open source solution out there. There’s one that we’re actually going to integrate our platform into. I can’t remember it off the top of my head, but basically, what they’ve done with their chat bot is it’s a little bit smarter because it’s actually using AI in the background, but to leverage that solution, you’ve got to be a little more tech savvy to do it. So, you’d need a programmer to go do that. For the average person, there’s a couple different solutions out there that are easy to manage yourself. So, Flow.io is one where they’ve made it easy to create kind of step-by-step chat bot processes. So, you create kind of the keywords. You create the conversation and the responses, but they’ve made it super simple to go through and really kind of make it so that you can create new experiences with your customers and answer questions. I would be hesitant to use that type of technology unless you’re really going to invest some time in it. Where a lot of companies start is they start with the dumb chat bot where you’re actually talking to a person.

Crystal:
So, you can start with those, and those are plug and play technologies. You can take that log data, go reverse engineer to go say, “What are the common issues and questions that we’re getting?” And then, go implement your automated chat bot. So, that’s the way that I would go about implementing that, and there’s dozens of different plug and play chat solutions or messaging solutions out there that you can implement.

Larry:
Right. That gets into content creation in an AI world. It’s more like you’re creating that … What do they call it? Learning sets or something in machine learning where the set of data that … in a real artificial intelligence machine learning environment where it’s actually learning to do new stuff, but this is simpler than that. This is just a simple, basic conversation level bot that can answer common questions about your product or services or whatever it is you’re doing on your site, right?

Crystal:
Yeah. So, that’s actually what a lot of companies are doing is they’re taking their existing chat log history, and they’re creating their training data set, and so, that training data set is what AI uses in order to predict an answer. So, if we take a step back and talk about what AI is, basically, it’s a complex math formula that predicts an answer, right? So, if somebody says this key phrase or this combination of words, here’s the right answer based off of your chat log history, right? So, that’s fundamentally how all of these AI programs work. Every single application, it’s just a statistical probability that this is the right answer.

Larry:
Got it. Okay. Hey, so, chat bots is one application. There’s also . . . getting ready to … figuring out the scope of this. You talked about sort of predictive things that you can do, and I guess this probably goes back to the server logs and then learning about your customers and stuff, but tell me more about that, what kinds of predictive activities you can do with AI.

Crystal:
Yeah. So, I think, to me, one of the best uses of AI is really kind of churn prediction, so predicting whether somebody is going to become a customer or not. You can predict, based off of all of these different attributes. It can be their demographics. It can be their activity. It can be their engagement across all of your different touch points with them. Are they opening your emails? Are they going back to your website? How many times have they viewed your website? All of these different data points can be used to triangulate what’s your probability that they’re going to cancel or churn. So, that’s one of my favorite ones. That’s one of the easiest ones to implement. So, that’s why I like it so much, is it’s just a simple way to get a really productive introduction into leveraging AI as part of your business because once you have that type of data, then you can actually transform the customer relationship, right? So, you can say, “Hey, we noticed that you’re not engage. Here’s some things to help you move you along the funnel,” right? And so, you take those different data points.

Crystal:
So, you say, “Okay, we’re going to push you into the next phase of the funnel,” and you can start AB testing those different responses and then figuring out, statistically, which one’s likely to work based off of that customer segment, right? So, you take the power of what AI can do with your customer segment. So, you use AI to say, “Here’s my customer segments. They have these types of attributes,” and then, you couple that to how they’re responding to all of your different forms of engagement, and then, you’ve got a powerful tool that can actually help you drive real business decisions, not just kind of shoot in the dark and hope to serve everyone with everything that you’re building. So, [crosstalk 00:18:13].

Larry:
Right, and even as you’re implementing that, you’re like, “Well, let’s … ” Well, I can picture two scenarios, like, “Let’s start with the less engaged one because it’s low risk, and we’ll see if we can grab them in.” But the higher impact might be figuring out the more engaged people, how to really rope them in and really further engage them, and I guess those are all just business decisions that you make down the road [crosstalk 00:18:33].

Crystal:
Well, I think it’s even more fundamental than that. It’s who is most engaged and why, right? What are the segment drivers that make them the more engaged, right? And all the data exists in all of your different platforms that do it. It’s just pulling it out and understanding it, and that’s the power of what machine learning does.

Larry:
Right, and if you’re planning a content marketing program, that’s golden information to have. So, that’s one classic example that you just alluded to. You talked about the funnel and the customer journey and all that and creating the content that . . . We talked earlier a little bit about the implications for SEO, and I’m a little bit of an SEO nerd as a hobby. So, I’m really curious about the implications you … because here’s why I asked that. Because the first thing I thought was like, “Oh, my God. Google is completely algorithmically driven. If we all become, as customers or as SEOs, algorithmically driven, that ever escalating thing that happens is just going to explode.” Am I wrong to worry about that, or are there different concerns? Just tell me, I guess, tell me about how you think this will affect that and how AI stuff will affect SEO.

Crystal:
Yeah, I mean, so, here’s my … I have a theory of how Google, specifically, is going to handle this. So, I think that Google is pro-customer, right? They are about the best customer experience and finding you the best answer. I would suspect that they will actually implement algorithms that reward AI-based dynamic solutions because, if you think about, if you’re customizing your website based off of the customer experience, then you are giving them a better answer than what they’re finding. So, Google, they’re not fully automated. They’re not 100% AI driven, and the technology is still in its relatively early stages. They are, by far, the most advanced AI company out there. So, don’t get me wrong on that, but the technology is still in its early innings. So, they still influence the algorithms. They still influence the rules around SEO rankings.

Crystal:
So, my bet is that they’re actually going to reward companies that are using advanced technologies. I think it’ll be challenging to do that, but I think that you’re going to see that within the next few years. They’re already rewarding companies who are faster, who have faster websites, and they’re doing significant ranking improvements if your site is super fast. This guy that I met, he implemented this tool called Gatsby, and basically, it just renders your site as plain HTML text, and so, it’s like the fastest that you can get. There’s no code anymore, and what he said was his lame blog increased up to the first page just because of the speed improvements that he made, and he didn’t have any content on there. So, he was like, “This is the way of the world.”

Larry:
Yeah. No, that’s interesting, but what you said before, I agree with you because I think Google is most concerned with how well you satisfy the user intent, and if AI is helping to do that, boom, you’re golden. And I bet that, that would be a more important ranking factor than speed. I think speed is one of those tiebreakers, but that’s my hunch. Anyhow, yeah. So, I can see that, and sort of like marketing management, I’m trying to think. All this data that you have when you talked about this ability to be more predictive, are there more sort of general sort of … just thinking how you can be a smarter manager of content and other resources with all this data that you’ve got.

Crystal:
Yeah. So, I think, really, that’s the challenge, right? So, when you look at marketing as the profession, marketers are creative, right? They’re creators. They’re creative. They’re interesting. They’re storytellers, and so, kind of with where things have been at, they’re forced to become bean counters, right, where it’s like, “All right, how many leads did you get? How many sales did you get? How much, how much, how much money did we make off of this campaign?” And that is so hard to get to today. You’re looking across 10 different systems, pulling it into spreadsheets, trying to figure out what the heck happened, just making crap up when you don’t know, right, like spinning a story, but to be fair, they don’t have all the data. They don’t have all the visibility. So, I think as these platforms … and I’m not the only one working on this. There’s hundreds of companies who are doing the same thing. It’s the biggest evolving space in marketing. As these tool sets evolve, their job goes back to more creating the stories and less about trying to understand the data, right? To me, everything that I’ve seen, every time I’ve done an analysis, it’s like we look at the data, and it totally makes sense. Until you see the data, it’s like, “Well, I don’t really know what’s happening.” We see the data, and it’s like, “Oh, well, that’s what happened. That makes sense to me.”

Crystal:
And so, with the change where they’re not necessarily doing all this heavy lifting to go understand what’s happened, they’re told a story with data. Then, they can use those stories with data to create stories that are going to attract the customers as opposed to creating stories to go ward off their senior leadership team on why they’re not getting more leads, right? Like, “Oh, we need more money.” Well-

Larry:
Well, that gets into that. I think whether you’re just trying to achieve a business goal with content struggles at some point with tracking back your business results to the content they’ve created and what it’s done. Do you see any hope? And you just set out the highest level dynamic there, that it’s like trying to quantify every little thing, that kind of bean counter mentality versus the creative mentality, and I think part of that underlies the creative sort of approach to things is that there is power in storytelling, and that’s probably more effective than a lot of data. Do you see data coming … Do you see data that can tell you effectiveness of content beyond those sort of clicks and other engagement metrics?

Crystal:
Yeah. So, to me, I think that our mission is to enable you to be able to track or rely on every single post, page, and tweet, right? So, with that type of information, though, it’s at a surface level, and then, you take the data to the next level, which is, okay, who did I attract? What did it attract them? So, really we create these customer archetypes or personas on who we’re actually targeting. Well, those personas, you can actually tie that back to a real customer in your system. You can actually tie that back to a real person or a real group of people, right? Because we’ve got similar kind of attributes, and so, with that type of information, you can actually go into each type of persona, and then, you can say, “This type of person was attracted to this content. This type of persona was attracted to this content.” And then, you can customize all of your levers based off of what’s working for your best customers. So, the arena of personalization is actually going to become transformative in every customer experience here within the next couple of years. This isn’t like 10 years away. This is like months way.

Larry:
The way you said that, I’m all of a sudden picturing an intermediate level between mass marketing and full on personalization. A lot of what you’re talking about there, that sort of persona base marketing, but not just in that you actually kind of put in an actual sort of aggregate personality that emerges that you can still target campaigns and things, too. Is that an accurate way to think about that?

Crystal:
Yeah. So, if you look at Facebook, right? So, Facebook knows more … Well, Google and Facebook know more about you than anybody else in the world, right? You’re on Facebook. You’re a foodie. You’re a mom. You’ve got two boys. You’ve got all of these different attributes. You like wine. You like your son’s soccer games on Saturdays, right? So, you’re the soccer mom on Saturdays who works for a marketing firm, right? And so, knowing that information, you, as a marketer, can actually target that type of person, right? So, you can actually create better look-alike audiences saying, “Facebook, go find me these specific types of people” and get very narrow and laser-like focused on who you’re targeting because they match with the types of people that are successful conversions.

Crystal:
And so, with kind of that new level of information, you actually have a lot of power on how you’re spending your marketing spend, right? So, instead of spending $500 to try to attract 100 people, you can spend $250 to attract the 10 people that are really going to be a close fit, right? So, it really optimizes what you’re doing as opposed to kind of the scattershot marketing that most people do. It’s kind of the “set it and forget it” campaign, and then, that’s like, “Oh, well, we’re burning too much cash. Now, we’re going to go refresh the campaign.” So, you can measure and attract those as you’re going along and tweak it if it’s not working.

Larry:
Again, a lot of what you said seems to be in that realm of helping you work better and smarter, and hey, I just noticed, Crys, that these always go so quickly. We’re coming up on time, but I want to give you one last shot. I always give my guests the opportunity. If there’s anything last, anything we haven’t talked about, about content strategy or AI or anything of interest, anything last that you would like to share with my folks?

Crystal:
Yeah. I think, like I said, AI is really a journey, right? So, you first focused on measuring things, making sure that what you’re doing is working and categorizing your data because that’s what you can use to actually enable your organization. To me, the second level is really personalization. So, how are you personalizing your experience across every single part of your sales funnel? And then, the third piece is really kind of we are around the corner from amazing technologies like personalized video content. Chat bots are soon-to-be smart, right, where you’re not going to have to do that much work to go implement those technologies. So, I think the modern marketer is always used to kind of these new tools that are constantly just on the market. We’ve grown from 1100 MarTech solutions in 2011 to over 10,000 this year, right?

Crystal:
And so, you guys are always bombarded with all these solutions that all have their own learning curves and things like that. So, the question, as a marketer, that you should be asking now is, “Aligned with my content strategy, what are the AI technologies that I should be looking at and evaluating that are going to evolve my game within that strategy?” So, if you’ve got a video content strategy, how can you create personalized videos, right? If you’ve got a dialogue strategy where you’re creating conversations with customers on Twitter, right, how can I evolve that strategy and make it smarter? And now’s the time to start adopting those technologies because it’s kind of like if you make a mistake, you can be forgiven, right? But if you’re making mistakes three years from now when the technology’s evolved, it’s going to be disastrous, right?

Larry:
That’s great because I’m a late adopter traditionally, but this is one of those places like, “Yeah, maybe I should go make all my mistakes right now.”

Crystal:
Now’s the time to make the mistakes, absolutely.

Larry:
Cool. Well, thanks so much, Crystal. This was a great conversation. I really appreciate you taking the time.

Crystal:
Yeah. Well, thanks so much. I enjoyed it.