CaddieOpen in CaddieDownload for Mac

Chatbot Marketing to Decrease Complaints & Improve Product

Shared by Caddie Library · Updated
App Masters16:37Open on YouTube
Chapters show their main points

The bottom line

A layered in-app support flow of FAQs, AI bots and agent handoff resolves most issues without an agent. Users only want their problem solved, so automation cuts wait times and raises satisfaction while agents handle the hard cases.

Key takeaways

  • Users care about getting a fix, not about whether a bot or a person gives it. 2:29
    • During COVID, automation rose and agent numbers fell, while response time and CSAT kept improving.
    • Remote agent teams handled the growing ticket volume faster than agents in physical offices.
  • A strong knowledge base of FAQs is the first thing to build. 3:57
    • Customers who keep updating their FAQs reach up to 90% self-service, and some reach 99%.
    • Reports show which FAQs get read and marked helpful, so top issues get new articles quickly.
  • Bots handle each step of the journey and pass on what they learn. 5:45
    • Smart intents use ticket history to offer the top problem types, then a quick search bot shows matching FAQs. It deflects about 9% of tickets on average and up to 30% for some customers.
    • Custom bots collect name, email and game level, which would take an agent three to five messages.
  • Complex cases and premium users still go to a human. 7:41
    • Legal and policy questions, or VIP users, can be routed to an agent after the bot finishes.
    • The agent sees what the bot already collected and tried, so the user does not repeat themselves.
  • Automation shortens waits and gives agents more meaningful work. 14:19
    • A case study shows a 17 hour drop in average customer wait time and a 20% improvement in resolution.
    • Setup takes about 30 days, and the SDK runs on over three billion mobile devices.

As long as they're able to get their problem resolved they don't really care whether they're talking to a bot or an agent.

Samantha Pang, Helpshift

Chapter summaries

0:00 Why users accept bots

Support tickets grow with the user base, and bots handle that growth because users only care that the problem gets solved.

  • Rapid growth brings a flood of support tickets that automation can absorb. 0:15
    • Helpshift launched its chatbot offering a couple of years before this talk.
    • Lockdowns brought more players to mobile games, including people who had never played.
    • Customers deployed bots because agents could not get into physical offices.
    • Ticket volume and the share of tickets automated both rose together.
  • Bots do the verification so agents can start on the actual problem. 2:41
    • The bot collects name, email and the last level played.
    • The user just taps or types details they already have at hand.
    • CSAT keeps rising every year as the automation percentage goes up.
    • Complex problems still need a human, so the bot hands over after collecting data.

3:33 The layered in-app support journey

Support works as a chain of steps inside the app, each one trying to solve the issue before the next one starts.

  • A published help center is the first layer, so users try to help themselves. 3:57
    • Common questions need neither a bot nor a person.
    • The whole journey stays inside the app and is personalized.
    • FAQ read and helpful data is tracked across languages.
    • This data guides constant work on the deflection rate.
  • Smart intents and the quick search bot personalize the next step. 5:45
    • Patented AI reads ticket history to find the top problem types.
    • Intents appear when a user says they need more help after an FAQ.
    • A new customer can bring ticket history over from another tool.
    • Each FAQ a user picks for their intent makes the AI more accurate.
  • Custom bots tied to the chosen intent decide whether a bot or an agent finishes the job. 7:05
    • The decision depends on the intent category, the user type and other traits.
    • A premium user may still get a human after the bot finishes.
    • The goal is to confirm the problem is fully solved and the user is happy.

8:19 The feedback bot and a live demo

The user steers the whole flow, and a feedback bot closes it by sending happy users to the app store and unhappy ones to a reason form.

  • The feedback bot starts when the user says the problem is solved. 8:39
    • The conversation can end at any point, on the user's own timing.
    • A five-star rating prompt comes up, and the company sets what each score triggers.
    • Four or five stars can ask for an app store rating.
    • One or two stars can ask why the score was low and how to improve.
    • The user then returns to the app where they left off, in the customer's own branding.
  • The demo shows the full flow in a fintech trading app. 10:02
    • The in-app help center lists popular articles, ranked by user clicks and searches.
    • Typing my wallet is missing brings up matching intents. The user picks lost bank card, and the quick search bot shows the top three FAQs.
    • If the FAQs don't help, a custom bot blocks the card, and a new one ships the next day.
    • The bot knows the authenticated user, Eric Ashby, and lets him pick which card was lost.
    • The user can still ask for an agent, for example to check transactions made while the card was lost.
    • A three-star rating brings up a list of reasons that the customer wrote themselves.

13:04 Setup speed and what teams report

Teams can start quickly on common frameworks, and the gains show up in agent routines and in users returning to the app.

  • Setup is quick because support problems repeat across similar industries. 13:20
    • Known common problem types for each vertical give new customers a head start.
    • The demo flow took about two minutes, against a five-hour phone call during COVID.
    • Supported frameworks include iOS, Android, Unreal, Unity, Cocoa, Xamarin and React Native.
    • The case study customer deployed and integrated in only three weeks.
  • Automation improves agent routines and brings users back. 14:30
    • Agents spend less time on mundane questions that a chatbot can solve.
    • Higher CSAT makes agent performance look better.
    • The case study customer said prioritized queues helped optimize the team's routine and response time.
    • Brands see users return to the app because the experience feels personalized.

In the growth library

Samantha Pang, VP of Customer Success and Growth at Helpshift. In-app support can be automated end to end with a help center, intent-based quick search, custom bots, human handoff and a feedback bot, so users self-serve and agents handle only the hard cases. Helpshift customers report up to 90 to 99 percent self-service, a 17 hour cut in average wait time and a 20 percent resolution improvement.

More from App Masters · The whole library, and ask it anything

Transcript455 lines
0:00

what is up app nation it is cfp young founder of appmasters.com the place you go when you want action-packed content related to helping you grow your app downloads and more importantly those revenues and to grow those revenues you want to have some customer feedback so i know we talk a lot about scale and growing that user base but you might get more customer support so what's the best way of handling the influx of new support tickets coming in well we believe it's chatbot marketing we're here to break it all down the benefits how to really set it up and do people prefer a real person i brought on samantha payne she is the vp of customer success and growth at help shift the at first customer first support platform samantha sam welcome to the show thanks steve it's great to be here sam let's break it down at first like when i'm thinking about chat bot marketing i'm like all right it's gonna

1:00

be a bot it's gonna be automated i'm talking to a robot not a real person is there a benefit do people want to talk to a real person what have you guys seen on your end yeah you know at the end of the day users just want to get the problem resolved and what we found over time since we launched our chat bot offering a couple of years ago and especially during covid we definitely saw a huge uptick in our customer base starting to deploy those chat bots because they had trouble getting agents into physical office spaces what we found is that customers that actually had remote agents were actually able to handle the growing issue volume much faster than uh customers who had agents in physical offices at the same time we saw that the number of customers and the amount of tickets they were automating actually started going up as people were locked down in their homes um and playing mobile games right

1:58

and even people that didn't play mobile games before started to try mobile games and so that was kind of uh their their hobby during covid and we really saw that at the same time a lot of our other kpis that we look at like response time csat um all started improving and the number of agents started going down as automation ramped up so that that was the kind of the trend that we saw during the two years of covet and we found that really interesting and we're able to basically validate that at the end of the day you know as long as they're able to get their problem resolved they don't really care whether they're talking to a bot or an agent now there are some more complex problems that uh still need human involvement so what our bots really do a good job at is collecting that information your name your email you know maybe the last level you are playing those type of verification questions and then actually handing them

2:55

over to agents if you think about an agent trying to collect each of those pieces of information it would probably take them three or four or five interactions to get the complete list but with a bot you're basically tapping or you're entering your verification information that you already have at hand and then you're just handing that over to an agent who can then help you focus on solving your problem we found that really effective and that's kind of what we're seeing among our customer base and the csat just continues to go up every year as the automation percentage goes up at the same time i like what you said customers just want solutions they don't care exactly where it comes from hey sam one of the ways when i think about chatbot marketing i'm just like oh man i don't know where to start i have to come up with all these logic flows what should i ask what are some of the best practices for someone

3:44

who's just like starting to do support like took off went viral let's say the app went viral and now they're getting influx of users you know for for most apps they start out with a smaller user base as they're testing and preparing for global launch what we would really recommend if mobile and in-app is is really where you're focusing having your users engage with your brand is a really good knowledge base that is your faqs so that they can help themselves right like they can find answers to very common questions that really don't require a bot or and human to give them answers to if you have a really strong set of faqs that you can publish into your knowledge base you can really start to create that self-service best practice that will allow most of your users to at least try to find answers on their own before going down a journey where they need more help that's what we're focused on

4:36

is starting the experience inside the app so let me talk a little bit about our modern in-app support journey and how we provide this experience end-to-end inside the app and it's also very very personalized a user will contact support through the app so this is the beginning of the journey we talked a little bit about the help center that is really important for somebody starting out inside an app experience this is where the app experience comes in what we're finding is that a lot of our customers that uh really have a strong knowledge base are able to provide up to 90 self-service to their users and some as high as 99 because they're constantly iterating on the faqs that understand what the top issues are any given time so that they can just create those new faqs and publish them very quickly we also provide a lot of reports along this journey that will enable our customers to identify where they can improve on

5:32

this experience so the help center we have a ton of data on you know what faqs are getting which ones are marked as helpful and helpful um across languages as well so that they can constantly find ways to improve that deflection rate so step two here is where our personalized journey starts if the help center doesn't help them out this is using our patented technology that's done by ai called smart intents this is where we will use the customers data history or if they're a new customer coming from another tool we can use their ticket history to run ai models to identify the top problem types once we're able to see that the problem types will kind of build out their intent models and will surface those to their users when they say i need more help after going through an faq so once they select their intent then um a quick search bot which is also using ai and natural language processing will show the most common faqs that

6:33

match their intent type and because this is done by ai the users who select this faq that matches the intent will continue to increase the accuracy of the ai over time we do see this deflection rate improve and currently we're seeing about on average nine percent deflection across all of our customers who are using this quick search bot with some customers getting as high as 30 percent and all it takes is really a good quality set of faqs so that they can try to enable the users to self-serve now the next step here is a custom bot so if the quick search bot doesn't help them out then the custom bots will trigger and these custom bots are very use case specific tied to the intent that they selected at step two so the quick search bot is there to help them to start to collect some information or answer some of the questions that the quick search bot couldn't and at that point our customers can decide based on the intent

7:31

category based on the type of user you are based on any other traits that they're trying to look at how to provide the service to them at what level of service they need to provide then they will decide whether the bot can fully handle the problem or if it should get handed over to an agent for example if i'm a premium user when the custom bot is done they might prefer to still have a human engaged with that premium user to ensure that their problem is fully resolved and that they're happy with the experience or if there's a more complex problem that custom bots can't solve related to legal questions or you know policy questions that they still want a human to handle that then the custom bot can hand over that type of issue to an agent who can then use the information collected earlier on by the bot to try to resolve the user's problem and because some of the upfront leg work has already been done with um steps two to

8:27

four then the agent can easily understand how far they've gone in trying to solve the problem on their own and then where they need to continue to provide support to help them to get to resolution once the problem is solved then our feedback bot kicks in so at any given time the conversation can end and then feedback bot can kick in if the user says okay my problem has been solved so this is all controlled by the user in the user's own time and to the preference of where they want to go next so once the conversation ends our feedback bot will kick in and it will ask them how their experience was and that five-star rating will come up and then based on the star rating they can set up different workflows like for example the four or five star could ask them for rating at the app store one two can ask them for more feedback on why they gave such a low rating and how they how they can improve things like that so

9:18

that is all being controlled by our customers based on the personalization and the information they want to collect along this journey once feedback bot is done interacting with the user then they just exit the support experience and they are brought back inside the app where they left off and you know they can continue with whatever they want to do inside the app or go somewhere else inside the app to continue exploring on their own so all of this is done inside the app and it looks like it's branded to our customers own branding so that everything looks very seamless and connected for the user damn what i really want to see next is all the stuff that we've been talking about the flows the automation the custom bots can you show me how easy this is or how difficult it is to set this up definitely let's let's get into the demo here's one of our fintech demo apps it's just a simple trading app that we

10:08

created here's our help center this is branded to look like the app you can see the faqs are inside the app there's popular articles and this is generated based on user clicks on what they're searching for as well and they can search as well using the search terms if they need to do that now if the help articles don't help them then that's where they can click on the contact us icon and the smart intents experience is what you're seeing here from here you can type in a problem or you can select your problem for the demo we have typed in the problem and we just said my wallet is missing so before i even type anything else it's showing me okay here are popular intents that match what you're saying how would you categorize it and i basically say i'm lost i lost my bank card here here's the quick search bot so you can see the top three most relevant faqs are showing up here based on the lost bank card problem and

10:59

the faq is there you can take a look at it you can also go back to it and see another faq if you want to and there's a selected response at the bottom where it says okay these faqs help me or i want to continue to get help so at this point uh it knows where i am because i'm authenticated right i'm running inside the app it knows that the person's name is eric ashby and they and they've lost their bank card and they want to be able to help them out because we have these bots that know the who the user is their bank cards are also connected to the backend systems with the bot and they can select which card they lost so in this case they're asking what date you lost it and what would you like to do with it i want to block this card is a response i can select in the custom bot message here the card has already been blocked the new card will be sent out the next day and it's asking is there anything you can help with and here you

11:48

can say no i'm all done for now in which case the feedback bot will come up because your problem has been resolved if you wanted to chat with an agent because you want to maybe check whether any transactions were made on the card while it was lost you could do that or if you really that's all you need and you just need a new card you know there's no other transactions that have been made and no other concerns then you can say no i'm all done for now and the feedback comes up so this is reading the experience did we answer all of your questions if you say yes then the star rating will pop up if you say no it can redirect you back to a flow that sends you to an agent or sends you to the bot again so you can select another problem that you need help with based on the rating here you can see we gave somebody of three stars it can ask you the main reasons so what are these reasons and these reasons are all created by our

12:35

customers as well based on how they want to collect this information how they want to measure things and you can select a response there and then at that point it will confirm again is there anything else we can help you with is there anything else we can improve on and if you say nope that's it the conversation ends right there you can start a new conversation if you have another problem or you can just press the x and it will take you right back into the app that was a quick demo if there's any follow-up questions i'll be happy to address yeah it's a very nice experience right very seamless and you probably did that in like two minutes instead of waiting on a five-hour phone call which i experienced during covet it was just like insane hey sam the anything i missed that you want to make sure we cover to our viewers you know trying to build this journey is really not as hard as you think we tried to get

13:23

our customers set up in 30 days we know what are the most common problem types across some of these of you know verticals that service customers in in the same way and we can definitely get you uh started up and running in in 30 days or less our sdk is on over three billion mobile devices we provide support on uh the most common development frameworks um that you you know of like ios android unreal unity coco xamarin react native so we are very much experts in the space and would love to help anyone who wants to get started on this journey one thing i want to hit on is the app first support for app first companies and there's a case study if you guys want to check it out too so they'll just take our word for it our customers are getting quicker responses chatbot system solves many issues without an agent we've been able to focus our agents on more complex problems i love that and so check out

14:15

i'm going to have a hard time saying this this yeah financially okay so 17 hour reduction in average customer wait time 20 improvement resolution only three weeks like sam said to deployment and integration with all that stuff sam the other thing that is like your agents are probably happier look improving the productivity of our agents help shift allowed us to work with prioritizing cues of customer issues which helped us a lot to optimize the team's routine and improve our response time and i bet you the agents are actually more happier having to solve these more complex problems than some of these mundane ones that can be easily solved with a chatbot so i love that quote exactly yeah yeah i know absolutely we're seeing a lot of our customers say this um you know the agents feel like they're they're doing more meaningful work instead of answering repetitive questions their c-stat is also going up so it makes

15:08

their impor improvement uh and makes their performance look better and i think that uh the the brands are seeing that their users are coming back to their app because they're able to provide this personalized experience once again it is helpshift.com right here at first for at first companies you can request a demo and you can check out all the cool companies that have already been using help shift sam if the audience wants to connect with you in any other way do you want to send them anywhere else uh yeah feel free to reach out to our team um we have a 30-day free trial that uh you know we have very personalized experience from an onboarding specialist that will guide you through the entire experience you can connect with me on linkedin if you want and you know i'll be more than happy to have a conversation and and you know share some more information on on how our customers are using us to keep their users happy

16:02

and inside their app well all that is linked up into the youtube description as well with helpshift.com and sam's linkedin join us for part three where we're gonna show you how easy it is to get your service desk and user feedback tool ready to go so eric's gonna be joining us for that video to join us then sam thank you so much for coming on doing this you're welcome thank you for having me thank you all for watching and i'll see you on the next video