University Case Study: Using Bots to increase Leads & Sell
My name is Joseph again. I was just up here co-founder of smart Loop where chatbot platform for right now. We really focus on sales marketing and customer service. You could build anything. But like I said, those are the use cases that we really focus on I've been in digital marketing for about 10 years. Probably more really started out when Facebook launched business pages and around that I started helping like my family they grow spices and vegetables and we started selling them online. I build up all these As I started working at agencies helping them really understand how to build communities sell products through Facebook pages and for five years, we've been building startups smart Loop has been around for about three years. so, smart Loop is a platform to help you easily build chatbots. You don't have to write any code you just sign up. It's just pretty much like Squarespace for chatbots. We have a lot of people that are not developed. Purrs, they don't have to write any code and they can build out these chatbots really easily. Some the customers we work with we work with Coca-Cola, AOL National Geographic some government clients, some telcos and some universities. Okay, but today I'm really going to focus on how to how University is enrolling more students using Facebook Messenger Bots. so, a little bit about the client that we work with is University of the people. They were founded in 2009 their nonprofit. They have students in over 20,000 or they have 20,000 students in over 200 countries. so, these are a lot of students overseas. They have some in the US but it's primarily overseas. so, their goal was really to enroll more students and we're working with them through Facebook Education team. They reached out to us. They said hey, we have this client University the people they use Facebook. They're doing a really good job of helping students like get education at a really cheap rate, but they have a problem. They want to enroll more students and they're getting bombarded with messages on messenger. so, their problem was Educating and qualifying students to get them to actually submit an application and enroll in their University. So, this is what their problem looks like. They have over 1.3 million likes on Facebook people would come to their Facebook page and they would ask him just simple questions like what degree programs you guys offer. How much does it cost to enroll? What do I have to do like to be qualified to enroll? so, this is what their problem is people would come to their messenger and they showed it to me once we started working with them and it's just had probably a couple thousand of just unread messages and once they started having a conversation with the student it would span a Is they would say? Okay. Well Judy programs you have they would send them then the user would see it. They would reply in four hours and then the agent would see it in another four hours. so, just became this very long process where a lot of students would fall out of their funnel. They would just get disenchanted or maybe go to another service. So, the solution was to create a messenger chat bot that was able to answer these questions for them educate them about their University and push them to submit an application to the university. so, this is what it looks like people would come to their messenger bot from the very beginning. They would say okay. Hey, I'm Michelle the mess the University of people bought I can help you. Do these things. And first we started asking them questions. so, initially, the very first question was like are you a student because they had students that were already a part of the university and they had students that were interested in being a part of the University so, we had different for those people so, from very beginning we would understand like who is this person? Are they a student? Are they not a student based on what they would answer we would start to tailor the content based on that so, you can see here? Someone says, I'm not a student and we show them degree programs fees. If somebody continue through say the degree programs flow. They would tell us okay. I want a bachelor's degree in computer science. and then once they got to so, once they started going through the flow, they would start answering questions. Like okay. Do you have a high school diploma? Did you graduate from high school that spoke English and if they met this criterion, then we would allow them to submit an application. so, these are results for three months. And again, Facebook came to us and said hey, we need to get this launched. We need to have three months of data so, that we can go to Facebook fa and present this use case. so, this is the first three months of the so, students came to the bot and the bot had around a million messages processed in those first three months. so, if you think about something like that, it would be very hard for their live agent teams to actually provide this type of service because they didn't I mean they probably could in India pay for all these agents, but theoretically the bot was control was controlling most of the conversation for them. So, in those three months 51% of the people who use a bot starter start an application so, they went through the whole flow. They qualified themselves and of those 51 percent 20 percent actually completed the converse the application and these applications are done on their website via their like just forms that they have so, I imagine there's some optimization that can happen here with the completion rates, but that's outside of messenger and then the biggest thing and I think this is the biggest Away from this whole thing is that they saw 62 percent decrease in their cost per lead versus just going Facebook ad to a landing page or Facebook add to their website. They saw 62 percent decrease in the cost per application through Facebook advertising and we believe this is because the bot was able to further educate them and engage with them around their needs and then once we fully understood who they were and if they met if there are qualified lead then at that point they can Actually go to the application versus I'm targeting someone with a Facebook ad I'm going to send them to the website. And then at that point all that traffic is what they're actually calculating is like cost per lead vs. Qualified leads going to an application. So, how do we do this in two weeks? And I really did when they came to us and said hey, we need this in two weeks. I was like, there's no way we can do this. Like I don't our team we have a good team but there's no way you can do this, but we figured out the process and I think this is a good lesson for other companies to learn is that you can actually if you're clearly Define your goals and your processes and you can launch agents like this and really fast process. so, the steps to launch since this was an admissions flow. We really wanted to talk to the admissions team to fully understand. Okay, if you do talk to a person in real life or they come to your website or you visit their High School. What does this admission slow look like and they were able to kind of write out a decision tree or conversations around how they actually talked with the with the prospective student and guide them to submit an application. Yeah. so, inside of that we want to understand League qualification, like what is a qualified lead? What does that look Like to us, how are we going generate traffic? They had organic and we have ads and then what is going to be like our fastest way to build this. so, here you can see the user. This is just a sample of our user Journey that they helped us that we worked with their teams to build. And again, this is like their admissions team to fully understand. Okay, if the first time you talk to user how do you get them to summon application? You need to understand like what degree programs are interested in Bash? Where's and then from there? You can actually move down to the lead generation steps. So, once we got into the lead generation, this is where I spoke about earlier in the very beginning. We said are you a student? Yes or no that helped us really Define like okay if this guy is already a student maybe just has questions about his class or he has payment questions something like that. We can direct them other way if they say they're not then we can actually move this down to where we're trying to generate a lead or get them to fill out an application from there. We go into understand like what degree programs and then from The Next Step we go in and we understand like the real qualifying questions. And again, this is an online university and they work all over the world. So, we have to understand like are you did you graduate from a Spanish speaking or a Spanish/English-speaking country or high school? And are you like a do you meet a certain age requirement? so, here you can see like a lead profile that we generated for this person. Thankfully Facebook already tells us like their first and last name so, we don't have to ask them that but we can as you can see, we go through the questions here and we understood like okay this person is not a student. They are interested in a Computer Associates computer science. so, we're able to build these profiles around users and over time. You can tailor the bot to show them specific content or if you understand that this person is interested in Maybe based on like your CRM data, maybe later. You can say okay. You got your Associates now, it's time to get your bachelor's or Master’s something like so, when came to traffic the good thing about universe of the people is that they are their messenger and boxes already bombarded with messages. so, that's a good problem to have I guess but there are some things that we did around there. We made sure that the page settings are correct. so, when somebody comes to your Facebook page the messenger window pops up automatically and you could chat with them. They did social posts around their chatbot saying like hey, we have this chat bot you can talk with it and we're driving traffic that way and then we're also, running it on their website we This is just one thing on our side as we were using the Facebook Messenger plug-in to run on their website. But since somebody might not be already logged into Facebook Messenger. We are seeing a drop off there. so, we just changed it to a native Facebook widget. so, you don't have to be logged to Facebook to use the chat Bots and then the biggest driver for a lot of the traffic was Facebook ads. so, this is a quick to messenger add right here. We can see has a video somebody can watch the video. It says this was going to open in messenger. Connect a messenger. so, we're saying the expectation that if you click this it's going to go to messenger so, that people aren't caught off guard so, they can watch the video. There's a call to action. They click Send message and it seamlessly goes right into the messenger experience and you can take them from there through your funnel. so, there's a lot of really good conversion. If you can Target people correctly you have a good dad. You can push them into your Bots. so, this is our platform smart Loop and this made it really easy. I remember so, when we first started this company three years ago, we did not have like WYSIWYG builder. We're just writing code and we worked with Coca-Cola for a very simple project and maybe had two flows, but just the amount of time that it took us to actually build the bot because we had to write the code, we had to send it may be the head changes. We had just a lot of back and forth in actually iterations and writing. Code so, now we have a builder super simple really fast. so, this is why we can build huge bought in a week and a half two weeks versus writing code and I think that's really what helped us launch this in the timeframe that we needed. so, just a couple key points about our platform. It's a point-and-click conversation Builder. You don't have to write code. You can write code we have code sections where you can create your own custom features. You can pull in apis. You can pull an API and then do some type of action or show some type of Carousel. We have ai analytics around the eye side right now. Just the biggest thing is we have natural language processing built into the bottle ready. So, if a user types something we can understand the intent and serve some type content to them or perform some action multi-channel support so, you can deploy this on Facebook messenger website Viber telegram, we chat a bunch of other different channels again one-click deployment so, you don't have to worry about setting up infrastructures. You just go log in Click connect your Facebook page and we'll go live we offer push notifications. so, you can send out broadcasts to your people on Messenger. A lot of people do it on vibrate as well for upcoming campaigns or notifications and then a big thing that initially when we all kind started this we were like, well bots going solve everything we don't need to Pass to a live agent. It's just going to do it. It's going to be perfect. Well that wasn't the case. so, we have a bot to live agents or live chat so, you can seamlessly transfer somebody from your Bot pause the conversation go to live agent talk with them close the conversation and rese the bot. so, in conclusion, I think the big thing I wanted to focus on is that lead generation and using Facebook messenger. Facebook ads is a really good use case for a lot of businesses and really driving home the point that it's not going to take two three four months. It's you can actually if you plan it right you talked you define your goals and you really focus in on the use case and what you're trying to do, then you can actually build something in two weeks and get it live. Thank you. Questions
Hello, my name is Arun. from nuanced. I have a quick question how you guys handle in comprehension on that specific University example, like how many in comprehension did you see that people go through before they have just like to like an agent. Can you see how many in comprehension of those are something like instead respond? Yes or no, they say something completely out of scope. How many do you see going through the comprehension before transferring to like a live chat? Yeah. so, the question is as people are going to the flow if there is start typing stuff or they don't answer the questions. How do we accomplish that correct? Okay. so, the way we build the box right now is if like we're asking a simple yes or no question. And again, this is why earlier I talked about like how do we bring someone back into the flow if they go off on a tangent? so, right now what we do is if we say a simple yes or no question, we kind of force people to continue on that process, but we can identify if someone says yes, then we can let them pass if they click the button then they can go to the next step so, we can identify that but then we also, over time as things went on. We started looking back at all the analytics and Did they like maybe this guy got halfway through the flow, and then he wanted to ask about pricing. so, we that's where the natural language processing for us, and these use cases comes in. We're okay now, we can understand that this person is asking about pricing we can direct them back to this pricing flow. They can learn about that and then we have to try to get them back on track. Thank you. All right and more questions on back. Yes. so, Question the users are working their way through the flows, right and at a certain point in time you either capture some information about them or you ask them to authenticate that information can those inputs then be shared or integrated with existing CRM resources, for example, maybe an eloquent database. Can you say that last part your question was once you have that information captured in the Bottom by the platform can it be integrated with eloquent or other CRM databases? Yeah, exactly. And I think that's where we're taking. This next step is like for the University. I know they have like this big CRM that helps them manage their and how they kind of progress through their University. so, I think taking that data and pushing it back to the CRM and at creating that initial customer leader syncing up with a customer lead and kind of rich Data is going to be very good. We haven't really been doing that yet with this University, but I specifically like this Telco client that we have they take all this they send it to the CRM and then the CRM kind of becomes the Hub where if a person didn't make it to the application step maybe in a couple days. They can trigger back the bought the say hey, here's a here's a discount or hey, maybe you didn't you didn't qualify for this program. Here's a secondary program that you can use. so, I think yeah, I think that's where Are lot of this is going is really use either like inside of our platform or an existing CRM is pushing that out and kind of syncing up and enriching so, that you can kind of upsell or resell to people. Another one back here. Sorry. So, you mentioned you mentioned the transition of bought two agent hand off. Can you describe the flow of how that interaction takes place? And how are you essentially benefiting a live agent providing in the context of a conversation or is Agent expected to kind of read that previous history of conversation and get that context in that manner. Okay. so, the question is if you escalate to a live agent, how do you handle that process? And then is there a way to kind of enrichment? That process for live agent so, that they kind of get the context of the conversation. Okay, that's a good question. so, specifically for like the university they have an external service that they use for all their live chats and we had to integrate our platform our live chat handoff to go to them. so, there's two ways that you can do that. so, when it comes to the user side if they get lost or they get stuck it hits the fallback content, aren't we call the default where it says? Hey, I'm a chatbot. I didn't understand your question. You can go back to the main menu. You can see frequently asked questions or you can talk to live in. At that point then they can escalate to live agent or we just give them the option to kind of escalate some of the clients. They just given the option like in the main menu or in the persistent menu. so, those are two options if it hits like the default or fallback it fails, then you can escalate or just give them the option when it comes to enriching or letting the agent catch up to where they're at since for like the university client. We had to integrate with an external live chat service that they use it was kind of hard for us to capture all the transcripts and push it to their live agent service. so, we do is we say, hey, we're going to connect you with a live agent. Please describe your problem or please ask your question and at that point, we push all the information and then and then the agent can read it and catch up inside of our platform. We have all the conversations running. We have a tab that says open conversations. And then from there you can kind of look back and see. I think it's just kind of either if you can't summarize or give all the transcript information just kind of ask the user to summarize or ask their question and then the agent can pick up from there. Any question that's couple of questions about the flow that you have. It looks very much like an ivr kind of flow very outgoing very step by step and people may not want answer the question in that same order. so, my questions are can do the skip a step. Up, could volunteer more information or if you ask question can it give a different answer and let the floor change in Direction? Yes, I think the question is how do you handle a user that might jump around or ask different questions and I think we're trying to get better at that because by Nature people want to have conversations with bot. They just don't want to follow like a just a stream line clicking flow. so, we try to identify these questions that people are going to ask and we catch that with natural language processing. And then once we kind of feel that we've answered their question, we try to nudge them back and I think over the next year's we're going to start seeing a lot of better systems that and allow a user to go off but we still understand. This is our main goal and then drive them back to us against our internal goal of generating a lead or selling a product or something like that. All right. I think that's it. I actually have to run to Facebook headquarters to do a similar presentation to this on this. so, I think that's going to end the talk, a round of applause for Joseph, please.