About Purna Virji
Named by PPC Hero as the #1 most influential PPC expert in the world, Purna specialises in SEM, SEO and the future of search. With over a decade in search, she is a regular keynote speaker at conferences across the globe such as MozCon, Hero Conf and BrightonSEO, and writes for Moz, Search Engine Land and Search Engine Watch.
In her Learn Inbound talk, Purna will explore how in today’s conversational world a landing page is no longer necessary for online conversions. Chatbots are seeing a marked rise in use and audience adoption, so advertisers need to be thinking about driving ads to chatbots, and how to measure the engagement that occurs there.
- Why chatbots are a fantastic tool for increasing audience engagement across multiple channels
- How to use bots across multiple stages in the consumer decision journey (esp. remarketing!)
- How to track the chat- the top KPIs to be measuring
Good morning. How are you guys? So happy to be back in Dublin. I am, like, obsessed with your country and I've been trying to find a way to tell my boss that, "Dear boss, I'm going to be working remotely for the rest of my life." So, we shall help make that happen. So I'm Purna. I work at Microsoft and two cool things you need to know, A, I have met and hugged every single member of Boyzone. I was 16, it made my life at that time. And the second thing is I'm super nerdy and obsessed with bots and AI and everything. So if you ever have any questions on the topics, just Bing me on Twitter. I am always on.
Now, when I talk to most people about bots, I always get this reaction, "Oh my God. The robots are coming." Right? Thank Hollywood for putting all these images of these dystopian movies into our minds. But the thing is, the bots aren't coming. They're already here, and they are super friendly and warm and lovable and child-friendly. Their goal is to learn all about us so they can help us and know our preferences and likes and dislikes and habits so they can make our life easier. Best of all, we can talk to them. Conversation is the most natural thing, right? Right from when you're a baby, it's the first thing you learn to do from crying to your first words to the bubbling, it is all about conversing and getting your thoughts out.
We are gonna enjoy talking to the bots so much that Gardener has predicted that in the next two-and-a-half years, the average person is going to have more conversations with their bot that with their spouse. My husband is equally anticipating this day, so... But I'm like, "No, honey. I'll just get a bot to nag you as well." So, it won't really work. Now, I know this number can seem, like, really unbelievable because we've all had those frustrating experiences with the bots, right? Anyone with me here? From IVR to any of those other frustrating soulless experiences, but that's not what the next gen of bots is about. All they wanna do is make our life super easy.
For example, I can talk to my digital assistant, whether is Alexa or Google Assistant or Cortana, and say that, "Hey. Ask Skyscanner if my BA flight to Edinburgh is on time." Or I can use it to order a pizza, "Hey, Cortana, ask Domino's to send me my easy order." Or I can even use a bot to help me fix my plumbing at home and I can say that, "Hey, ask Talk Local that I need a plumber to repair my broken water heater." It just takes the pain out of life admin. So we want to use it.
But from a marketer standpoint, this actually gives us this huge gift. For the first time in all of history, we can talk to our customers on a one-to-one basis. It takes our marketing from one-to-many to a one-on-one conversation that the customers can dictate the pace of, and then we can go ahead and give them the messages and learn from them in their own voice and serve them just what they want to see.
As you can imagine, because it's getting easier to use and the adoption rates are going up so much, it's changing the technology that we use. We are seeing that fewer users of screens, apps are becoming more and more unnecessary and unsupported, and websites as well because sometimes the website is not the fastest way to get the answer that you're looking for.
Instead, we're seeing people using things like voice search or digital assistants or, my favorite, chatbots here. And I was trying to look at some research into, like, the numbers and what people were thinking, and I found things that actually really surprised me as well. So first I was trying to look at people's reactions to bots here in the west. I mean, they're really common in Asia but let me see what people thought about it here.
So I found a Hotspot study that said 40% of people really don't care if they talk to a bot or a human as long as they get their work done. And I thought that was pretty fair. Like, think about your own behavior. Would you care if you found out if your flight was on time by talking to a human or a bot? Would it make a difference? Not really, right? It's super easy. And they may be new here, but 67% of people all around the world are using bots for customer service. It's just normal.
We're like, "Okay, that's going from a customer standpoint, but what about from a business standpoint?" So Juniper Research did this study on the banking and finance vertical to see how bots impacted their customer service times. They found that, on average, it saved four minutes for a call compared to a regular phone call. I mean, that's massive, and that's got to be worth a lot of money if you think about it. They're predicting eight billion dollars in cost savings in the next, four, five years. That's a lot.
So it's not surprising that 80% of businesses really want to have a chatbot in the next couple of years. It's win-win for the customer. It's win-win for the business. So if the whole world is now looking to get bots and customers are engaging with bots, [inaudible 00:05:46] stuff coming to your website at times, what should you do? Well, I'm gonna walk you through your four-step process to making conversations convert for you.
So let's start with step one. Step one is trying to find the pinpoints in every step. Now, if you've ever tried to ask for your pizza by asking Domino's, "Hey, Alexa, Domino's for my easy order," versus going to the website, there's a massive time difference. So what you should try to do is observe the entire funnel yourself. Very often, even I do this when I'm going through some of my clients' websites. I spot things that are really annoying. For example, if I'm putting in my credit card details and I put in my zip code, it won't automatically populate state and country, which I'm like, "If you have my zip code, you should be able to populate it automatically. There's no reason to ask me for it again."
So if you go through this process by yourself, you will be able to spot the areas that could be improved, spot any frictions so you can iron it out. And one of my favorite tips is to try to find the least tech-savvy person I know, usually a grandmom or my mom, and have them look through the process themselves because they're just not that tech savvy. They won't know quick workarounds They'll be there and they'll be stuck, and you can learn so much from that.
So some of the top questions I try to look at when I have people go through the funnels, are things like, 'Hey. How do customers find information or seek help? How do they make a purchase and make a payment? How easy is it?" The easier it is, the more likely they are to make impulse buys. Where are they getting stuck in the process? What are the slow points that confuse them or delay them? And then do pinpoints different across channels? Maybe mobile is really easy, but the desktop is clunky or vice versa. You want to know this.
Once you have these answers, you can prioritize the interactions. And then you can go ahead and build your hypothesis. Now, this is one of my favorite hypothesis templates because it states the goals very clearly. So if I use this, I could figure out that, "Okay, a pinpoint for me and my client is that the customer service reps are often answering the frequently asked questions that people call in to ask, and as a result, there are long hold times for my other customers. What if I build a bot to help answer some of those frequently asked questions so that my customer reps could be freed up to answer more in-depth questions and reduce the wait times?" That would make customers much happier. So now I could find that out.
The next step is honing our conversational skills. In a conversational world, we're gonna need two things. We're going to have to build a bot and a skill. What's the difference? Well, think of a bot as an app that you can text back and forth and chat with. A skill simply teaches a digital personal assistant how to interact with your brand. So for example, Skyscanner had to teach Alexa or Google Assistant or Cortana how people should interact with it, or if they wanna ask for a flight timing, how to give it to them. So essentially, a skill is teaching a personal assistant how to use your brand, and a bot is directly your agent like an app.
It's not as intimidating as it sounds to create a bot. I've got a ton of helpful links for you here that include sources from Microsoft and Facebook and Google, all on this list. Most of them are absolutely free to get started, and you can use it as much as you like. And some of them will only bill you if you exceed a certain amount of usage. So it's really very low-risk to get started. But what if you're like me and you're like, "I don't know how to code at all." I was in that situation. Like, "Oh, my God, I hate coding." I don't know, that's not me. Here's a tool that you can use that requires zero coding skills and very little time.
So if you go to this website called qnamaker.ai, all you need to do is plug in the URL off your FAQ page. Now, most of you have an FAQ page? Exactly. Put that in that, and it will use natural language understanding to think of the different questions that could come up. And in less than 10 minutes, you could have a quick bot that is easy to deploy across any channel you like, Facebook, Skype, Kik, anything that your customers use. The other thing you must think of as people are using personal assistance a little bit more is optimizing for voice search. It's been one of my favorite topics, and it's a question I get all the time, like, "Hey. How should I optimize for a voice and how do voice queries rank?" Well, at the moment, our algorithm ranks voice searches pretty similarly to how we do the regular search results.
To understand a little bit more about where Cortana will pull her vocal answers for, I think it's good to look at where search engines pull their answers from. Now, most of you will be familiar with this. But if you think about it, there's three sources of data that we can use. The first one, if I was looking for, say things to do in Dublin, you'd see the column right up top, the carousel, which is provided by structured data. So you provide structured data to us via things like your schema markup, so it gives us more information about how to organize the info.
The other one is unstructured data, which is what we pull from your website and try to organize. And then the knowledge graph, or satori knowledge graph as we call it, is where we pull in information from either the structured data or the non-structured data and we try to create an encyclopedia of nouns. So it's all about information on the people, places, any of that like. So this is where we pull in all the different information from. If you think about how Cortana will answer your queries and where it comes from, she'll do one of two things. So if you ask her a very direct question that has only, like, one possible answer, so for example, "Hey, what's the capital of Berlin?" She's gonna read out that answer to you and say, "Okay, the capital of Germany is Berlin." But what if I asked her a question that was highly debatable like, "What's the greatest rock band in history"? Now, what would your answer to that be? Boyzone. I know, Boyzone. I know I love it.
Well, Cortana would be a little bit more fair. So she would either choose to give you a verbal answer or not, and she would sometimes just give me web results that would allow me to choose from. What you can do to help impact... By the way, most search engines will work in this exact same way at the moment. What you should do is try to optimize and give us more schema, put in more structured data to help us understand you better, and also try to optimize for position zero. If you are looking, Rob Bucci from STAT Search and Pete Meyers from Moz having really good articles on that topic.
Two quick reminders, if you are a local business especially, you want to pay attention to voice because we find that if you are using voice search on your mobile device, it's three times more likely to be local than if you were using it on any other device. So what you wanna do is make sure your citations are absolutely accurate, so things like name, address, phone. And pay attention to reviews because, very often, they might not want to come to come to your website, they'll look at you for your reviews. And if you are highly rated, they'll engage with you or not. The other thing is you want to think of having content on your website that optimizes for different types of questions. Think of informational ones like, "How long is the train ride from Dublin to Cork" or, "What is Dunne's return policy," or, "What's the best camera for a child?"
If you try to have answers to these questions, it makes it easier for us to pull it from your website and serve that answer. Okay, you've verified the pinpoints, you've optimized for voice, and you've created a chatbot and a skill. Now you want to add the chatbot across the entire customer journey. When we think about the customer journey, I find it's easier to have three distinct points that a chatbot can act as. So your bot can act as an introducer where, if people are not familiar with your brand or your offerings, it can come in and introduce it. Or, it can act as an influencer where it's guiding the purchase decisions and trying to guide them into adding more things to their cart or maybe wanting to buy from you from the first place. And then, of course, we have the bot acting as a closer at the end, which is where it converts right on the spot.
So let's look at how a bot can act as an introducer. And for that, let's look at Adidas as an example. Now, any of you in from London are a member of the Studio LDN community that Adidas has? Okay. Figured I'd ask. What Adidas did was they are trying to focus on their females fitness community in London called Studio LDN. And as a way of getting more brand engagement and introduce some of these free fitness classes that they offer, they built a bot. They use the bot for people to book appointments for any of the free fitness workout classes that they had across the city. And then once people booked the appointment, it would confirm the time, and then it would engage them in conversation by sharing links to different videos or helping them ask questions to fitness experts.
What they found was that in just the first couple of weeks, they had over 2,000 people interact with their bot, and they had an 80% return rate for people to come back and keep using the bot. What they even found was that even going forward a couple of weeks, their retention rate for engaging their customers was over 60% more than they've ever had with any of their apps. That's pretty cool for them to see that the engagements with bots is really immediate. Because think of your own behavior when you get a text message. How many of you, when you get a message, will check it within 10 minutes? All of us, right? It's such a human instinct. So if you're getting a personal message that's one-on-one that's meant for you, you will want to check it.
Now, the only way that you could book these appointments was through the bot, but they did a really smart thing. They tried to promote their bot through all their other channels. So they used their Facebook channel, they used their website. If you went in the store, they mentioned this bot that you could go and interact with. So they used all their other channels to draw attention to their bot. Like, "Oh, that's really clever." So let me share with you what we can learn from Adidas and we want can do.
Now, how many of you here run Facebook ads? Any of you? Okay, quite a few of you. Now, most of the time, what do we do? We run ads to websites. What if we had a chatbot and then we ran ads to bots instead? It would be so much better engagement. You could be like, "Hey message us for a 20% off coupon." That's enough incentive for me to want to engage with the bot. And then once I start conversing with them, then you opt in to receive messages. It's so clever. Another thing you can do if you're running YouTube ads. Any of you running YouTube ads here? Okay, great. You can use your ads to create a little instructional video on how to use your bot. It's super simple, simple. Let me show you an example of a really good video that's less than 15 seconds, but it's still showing how you can use a skill that Fitbit has created.
Male: Hey Cortana. Ask Fitbit how did I sleep, what my heart rate is today, to log that I had 12 ounces of water. Hey, Cortana, ask Fitbit how many steps I took today.
Cortana: You've walked 11,246 steps today.
Purna: So short, so easy. Would you find that interesting? Would you be more tempted to get the Fitbit skill afterwards? Exactly, yes. It works, it's effective. But, hey, we're at a search conference. Let's talk about search. Both Google and Bing let you surface bots if you just type in the type of bot it is. So for example, if I was looking for travel bots, all of these would come up on both Google and Bing. The really cool secret, though, is how easy they are to rank.
If any of you have ever done old-school SEO where you've created a meta tag, have you ever done meta title descriptions? Yes. Circa 10 years ago. That is exactly what you need to do to give the search engines information about your bot. So you have a display name, which is a title. You have your description, put in the website. You can even put in tags to describe what your bot does. It is really simple to actually rank. So definitely do this. It's so simple. And what I realized about Adidas was that they showcased their bot through so many different channels, and you've seen how easy it is, because it helps keep the brand front and center.
What a difference would it be if I did, "Hey, Cortana, how much are flights to Paris this weekend?" She'll give me a list of all the different search results. But what if I evoked the Skyscanner skill? What if I said, "Hey, Cortana. Ask Skyscanner how much are flights to Paris this weekend." Look at the difference. What do you think the odds are of me buying from Skyscanner on the second example? Pretty high. What about the first example? It's a crapshoot, right? We don't know who will get it. So if they remember you and they know that your skill makes it very easy to complete their goals that they're looking to do, then they'll want to remember you and ask for you. So I would say the key lesson from this, from looking at the introducer step is to give your audience plenty of ways and reasons to chat.
Let's see how it can act as an influencer. And for this, I wanna give you the example of Alibaba. So Alibaba in China is a giant marketplace. Brands like Nike and all the big ones sell through them. What they did was they created this concierge chatbot called Dian Xiaomi, which is also called Little Shop B in English. And what it did was if you were browsing, say the Nike store, and you uploaded a picture of a t-shirt, it will recommend the shorts or the track pants that could match that. Or if you uploaded your height and your weight, it could recommend the right size that you should buy. So it was just really helpful, like a shop assistant, essentially.
But think of the ways that we could use it to up-sell. Let's say I am Brown Thomas and somebody's bought a black dress. They could recommend that, "Hey, why not complete the look with these heels or with these earrings?" Complete the look. That's a way to up-sell and help increase the average order value. The bot is smart enough to know what matches with the outfit, so it's not gonna recommend a completely random pair of, like, flip flops if you're buying a formal dress. It wouldn't do that. You can train it. So that's smart. What you want it to do is to learn from the conversation and remember, and then use it to shape future conversations.
I'll give you this example. Let's say I am a bookstore...or let's say I frequently go to this bookstore and they know about the kind of books that I like. Now, I've always been a big fan of Marian Keyes, and so imagine if they knew that. So the bot could ask me, that "Hey, Purna, the latest book by Marian Keyes is out. Would you like a sneak peek at the first chapter?" Now it's a slam dunk because I'm a big fan of hers. I would be like "Yes, please. I absolutely want to see this." And then it would also know that Purna tends to buy more when things are on sale because I'm cheap. But... So it could be like, "Hey, I'll email it to you now. But you could also save 15% if you order it today with the code MammyWalsh," because she writes a lot of books about the Walsh family. "And hurry up, the code expired in 24 hours." So, obviously, I would be like, "Thank you so much," and look for that email and buy the book immediately with the discount.
Isn't this easier? It's all about changing the conversation from, like, the, "Buy now, buy now, hit people over the head," to, "What do you need right now? How can I help you? How can I give you things that you would be interested in and seeing as helpful," as opposed to being like, "Oh my God, there is that pushy brand pushing things on me again." One more thing that you can do to kick it up a notch further is if you can make your customer experience throughout so easy and convenient that they'll want to choose you over any others.
Now, you've probably seen this before, where in the search results, we're now integrating chatbots. So if I wanted to find out more information about this restaurant, I could just say, "Hey, chat with me now," and it would bring up a little bot and I could ask it that, "Hey, what are your parking options? What are your vegetarian options," or any of the like. The goal of this is it saves me, the consumer, time from browning on the website to find the answers. I can go there and get it quite easily. This is still a pilot and it's really only available for restaurants in Seattle. But what's coming out soon that's gonna be available to more people, it's still in pilot, is a bot extension in your ads. And when I looked at this, you get charged only if you click that and then the chat interface comes up, and then any conversations you have are not charged anymore.
This is you submit to us 50 to 100 questions, and then we build a bot for you. So it's actually even easier. You don't need to do any coding. But when I was looking at this, I'm like, "Oh my gosh, the possibilities are endless." Because of just the nature of PPC, even with that same bot, I could upload different questions for every stage of the decision journey. Let's say I was a camera store. I need that at an earlier stage when people aren't that sure on what they want to buy, they could ask me a question like, "Hey, what's the best camera for a 16-year-old?" Or they could ask me, "What's the difference between DSLR and point-and-shoot?" It's just really high in the discovery process questions.
Or, once they've sort of narrowed down their selections, I could have questions such as, "Hey. Which brand is better," or help with the common comparisons that we get asked, or, like, usage questions that, "Can I use this to shoot at night," or, "Can I use this to shoot sports? Does it have a fast enough frame rate?" This helps them narrow down the selection. And then for when they're close to buying, their questions tend to be more about things like, "Do you price match? What is your return policy?" There's not that many questions. With 100, you can easily fill all of these type of questions. And what are you doing with that? You're making it really easy for your customer to find the information that they're looking for. They don't need to go on your website and do tons of searches and find the answer. It's right there, instantly, when they want it. So the key lesson is you want to think about being helpful and adding value.
Now it's time for conversions. Let's talk about the closure stage. Have any of you ever bought flowers from a website like this? Like, in the U.S., there's 1-800-Flowers, but we've got, like, InterFlora, and there's tons of them all over the world. Personally, I always find these sites really overwhelming because there's so many choices, and then they all look the same to me. So maybe I'm just dumb. But let's say if I went to this website, so this is 1-800-Flowers. I'll go to there, maybe I want to buy flowers for somebody's birthday. So I'll go look, and they'll be like 8 gazillion options. By some miracle, I'll latch into one, I'm like, "Okay, let me...I'll go with this one." Again, multiple options of what I should buy, different price points, and say I finally chose one, super long form interrupted by a pop-up, super long form all over. Pretty long, and I think with how technology has evolved, that's not quite sustainable. What 1-800-Flowers did last year was really clever.
So they worked with Facebook, and they created a Facebook messenger chatbot. So I tried to play around with it, and what it did was it asked me, "Hey, let's get started. Do you want to select from on of our curated collections?" Instantly, it's giving me far fewer choices, but they're the ones that tend to be the most popular, what they've selected. So I'll say, "Okay, for birthday, let me view arrangements." So then again it shows me, like, five or six different options, and I can just choose any one.
So now when I'm messing around with things, I make them call me different names. So here I chose the bot to call me Queen Purna, because, you know, so it should be. So then I picked one and it was like "Great choice, Queen, fabulous. So please pick one of the recipients." So because I've sent flowers before through the bot, it recognized, it was like, "Oh, do you want to send it to yourself, to Queen Purna in New York?" And I was like, "Yes." Select this, super easy, one button. Then it asked, "When do you want it delivered," and it gave me the date selections at the bottom. I chose the date. It asked me if I wanted to give a message. I'm like, "No, it's to myself. I don't need to write a message to myself." And then it says, "Check out if everything is easy," and it just used my natural phone way of checking out.
So again, it just took me a couple of minutes. What you guys think, much easier? So much easier. And the result that... So their CEO gave an interview to Digiday, which it's really interesting reading if you are looking for it. They said they found that 70% of the people who use their chat box were, like, brand new customers. They were able to reach this younger, different demographic that wouldn't shop on their website, but they would shop with their bot because it's so much easier. And, hey, maybe I'll send myself flowers more often if this is so easy to use. So think about this. Actually, Mark Zuckerberg gave this really funny quote. He says, "Well, the irony is that you don't need to call 1-800-Flowers to order flowers anymore. You can just go to their bot." I'm like, "Yeah, that's quite clever."
What we should remember is that we don't have to make the user get off their couch. They should be able to interact with us no matter what technology they want to use. Do they want to use something that has a screen? Something that doesn't have a screen? Do they want to use voice, text, or a mixture of both? Let them just dictate how the conversation should go. I'll show you this example.
Female 1: Cortana, I wanna dress like this. Can you help me find it? I like this one. Yes, please. Can I get it in a size six? Cool. Cortana, please give them my payment and shipping info.
Purna: So simple, right? She used texting, she used talking, she used a mixture of both. She took an image search. It was so easy and so quick for her to buy the dress that she just saw in a magazine. That's the experience that we have to aspire to. Make it as easy as possible to interact with you consumers however they want to interact with you. And personal assistants can make it easier because of the intelligence that they have.
Both Google and Amazon have said that their assistants can recognize different people, and Cortana's working on that as well. And she can also tell the difference between a child and an adult, and she telling responses to children differently. A, is it's really great because my son likes to say, "Hey, Alexa, please add poop to my shopping cart." And so that'll be really good if they recognize not to do that. But in all seriousness, children have different needs and desires. So if you can talk to them in a different way, it is helpful. But regardless of whichever personal assistant it is, they're going to start influencing our behavior as consumers because they're gonna give you much more intelligent, streamlined recommendations. As a brand, we should think that the fact is that we're gonna get fewer impressions, but when we get them, they're gonna be far more targeted.
Think about this, if I did a search on the desktop for, like, "What's the best soap for men," I'd see a regular soap. But what if I asked my assistant like, "Hey, Alexa. What's the best soap for men?" She would say that, "Hey, there are five top soaps rated by Men's Health. Would you like to hear what they are or do you want to try the new one from Axe?" Now, that's pretty cool. If it knew I tend to like Axe or if Axe wanted to promote themselves, very interesting to see how this could all shape up. Additionally, it would understand our behavior and serve us the right thing.
So if I said, "Hey, Cortana. Make reservations at this restaurant for two people as close to 8 p.m. as possible." And Cortana knew that I used Lyft instead of Uber in the U.S., so then it could say like, "Okay. I'll make a reservation for 8:30. Lyft is offering a 20% discount at that time." Do you see? As Lyft, if I was the brand and I told Cortana that, "Hey, on this Wednesday night between these timings, I'm gonna give a 20% discount," then Cortana could tie that information back to me who likes to use Lyft because she knows that from my preferences, and will surface me Lyft. So in this way, you can give different offers and services and information to the personal assistants so they serve you at the right time. Your goal here is to remove any friction and increase the relevance because it really works. And it's not scary that we'll get fewer impressions because each one will be very targeted.
Now, as the last step, you want to measure the conversation with the right metrics. And this is one of my favorite quotes. It says, "Rather than tracking users with pixels and cookies, why don't we engage with them and learn about them and provide value that actually meets their needs?" We can do this because the bots and skills give us information in our customers' own words, and we can see from the one-to-one what they are responding to. So some KPIs that you'll want to look at could be things like delivery and open rates. You'd want to look at click rates. If you gave them a link in your chat, did they click on it? The retention, how often did they come back and chat with you? What are some of your top messages? Some conversion rates as well as sentimentalists.
I want to show you examples of what these reports look at. So say I had a chatbot and I was plugging it into multiple different channels. So for example, I had my bot that worked on Facebook as well as Skype. I could pull reports for each of the different channels. So it's not like one bot overall stats because different channels may perform differently. So I could look at that. You can even customize the time zone and the time range, of course. You can look at the retention report, which really measures when people talk to you once and then when they come back again to interact with you on a different day.
This is a really good metric to look at, is my bot sticky? Is it bringing people back? You can also break down, like, how users behave across channels. Where are they more active? And even break it down by messages. So I can look at which messages tend to be most popular on the different sources because each of them would be, like, email or GroupMe or Skype or Slack, I could see all of that there. Isn't this cool? Such cool analytics. Same thing with skills, it's very much the same thing, which is the audio voice app [inaudible 00:34:18]. You could recommend all of these KPIs. Also, you'd want to look at sessions and the number of turns. So the turns is I talk to you, you talk back to me, that's two turns. So you want to check that as well.
The big thing to ask yourself when you're looking at this conversation is, "What can I learn from my chatbot that I can use to help other channels?" There's so many rich leanings here that you can use to improve. So, let me recap the four steps. Step one, you want to identify the pinpoints that you need to make easier with conversation. Ask the right questions. Step two, you wanna use your voice well. Build a bot as well as a skill and optimize for voice search. Then you wanna enhance the entire customer decision journey with the help of conversation. It's super effective. And then you want to measure the right KPIs and go beyond clicks and impressions to better engage your audience. Because remember, what were once really just dumb machines are now smart enough that we can engage with them on a very human basis. We're turning to it for recommendations. We're turning to it for servicing our needs. And so much so that Accenture predicted that in the next five years, more than half of your customers will select your services based on your AI instead of your traditional brand. So let's get friendly with the bots now. Thank you so much.