Field Test: Using chatbots as an NYC tourism resource
Introduction
Every day new tools are designed to simplify the interaction between humans and computers. These tools can also be known as Chatbots or Virtual Assistants. A chatbot is an artificial intelligence software that can simulate a conversation with a user in a natural language. To date, chatbots have become a very promising interaction between humans and machines. The goal for a Chatbot application is to streamline conversations and enhance customer experience.
Studies show that nearly half of consumers prefer communication with chatbots for customer service inquiries. The best chatbot experience is one that strikes a balance between conversational, friendly, and helpful. Let’s take a look at brands that use them effectively:
Sephora offers two services, their Reservation Assistant and Sephora Color Match. The reservation assistant allows costumes to make appointments with beauty experts by seamlessly sending a message to their chatbot. The Color Match uses AI to help consumers select colors. All they have to do is hold a camera to an image or face and a shade will be identified for their natural coloring.
Starbucks is another brand that does a great job with chatbots. Coffee is such a competitive market but they do an excellent job standing out in the crowd. Their chatbot allows you to order by voice, pay, and tip. All the customer has to do is show up.
Brands use chatbots all the time to streamline simple questions customers have. They answer questions that a human doesn’t need to. What is a way we can push the boundaries of chatbots to guide a user toward improving even more aspects of their life? To answer this, we need to reflect on the troubling year we just had.
NYC + COVID
Shutting down the city that never sleeps. On March 17th, 2020, the most unimaginable statement came out of one of the world’s greatest cities. Mayor Bill de Blasio announced a shelter in place. The city went through an economic plight for the 16 months that followed. Restaurants and bars shut down, boarded-up storefronts were everywhere, tourism dropped drastically. Empty sidewalks in midtown felt like a deserted island. If there is one thing we can say about New York though, we are resilient. After 472 days New York has lifted all COVID restrictions. The city is coming back.
Restoring the city to economic health is a massive challenge for the next mayor as we get through this. At 10.9 % in May, the city’s unemployment rate was near twice the national average of 5.8 percent. Many businesses didn’t survive the brutal wave of the pandemic devastation. The ones who did, need our help in order to stay alive.
Foot traffic is slowly coming back to attractions, bars, and restaurants. New York is slowly reawakening, it just needs its tourists back. As tourism slowly starts to trickle back in, we need to show them the stunning landmarks, restaurants, and shops to return to. New York needs these visitors and their disposable income to fully recover. Preserving New York’s culture is at the core of its values but that mission is at risk when visitors don’t know how to plan a trip or where to go when they visit or that returning to this marvelous city is even an option.
It’s a slow burn, but to revive the tourist trade, we need everyone to do their part. We need the critical mass of attractions and places to flourish.
I hypothesize that by using chatbots as a virtual concierge, New York City tourism will increase for distant visitors providing a lively experience that also works to boost the city’s economy.
Field Test:
My plan to test this hypothesis through the following steps:
- Create a chatbot of Manhattan using SendPulse
- Divide Manhattan into sections tourist may travel to
- Research the best tourist attractions in those areas
- Include those areas in a chatbot
- Share bot with users
I would determine the success or failure of my hypothesis based on the responses of both locals and visitors who used the bot.
Creating the chatbot:
It is quite user-friendly to create a chatbot in SendPulse. All you need to do is connect a Facebook profile to SendPulse and choose a Facebook page. Once you’ve connected everything you can create message flows that imitate live communications with a representative. You can organize the conversational logic of the chatbot based on keywords in that chat.
Sendpluse also allows for a variety of visual content! Text, images, files, etc.
I wanted to think of my chatbot as a 24/7 concierge. Think about when you go to a hotel and the expert at the front desk knows the exact recipe for a fun day. What happens when that concierge isn’t around? Or you don’t have one accessible! That’s where this chatbot comes in.
It took some trial and error to identify that SendPulse was the best interface for me. My first question in researching bots was, is it possible to create a bot without coding a single line? Luckily I found the answer to be yes. This is HUGE for smaller companies who don’t have access to a computer scientist. There are plenty of chatbot-building platforms out there to help a business.
I tried Chatfuel, ItsAlive and MobileMonkey. How did I end up selecting SendPulse? After spending a few minutes on the site, it proved to be the most user-friendly for my needs.
Dividing Manhattan:
The next step while developing this concept was trying to identify the way I would break up my tourism attractions. Fortunately, Manhattan looks like a grid. Manhattan is long and skinny, it’s 13 miles long and 2.3 miles wide. It’s an island, bounded by the Hudson, East, and Harlem rivers. There are really three major areas within Manhattan: Midtown, Uptown, and Downtown. After that, we can cut the city in half to East and West. That breakdown gives me six major areas of town to work with.
Uptown East: 59th to 110th between 5th Ave and FDR drive
Uptown West: 59th to 110th between Riverside and Central Park West
Midtown East: 14th to 59th between 11th and 5th Ave
Midtown West: 14th to 59th between 5th and FDR drive
Downtown East: Battery Park to 14th between West Side Highway and Lafayette
Downtown West: Battery to 14th between Lafayette and FDR drive
Research the best Areas:
Once I cut up my corners of Manhattan came the fun part. I did research on the best attractions Manhattan had to offer. I made it my priority to explore dining, shopping, Broadway, museums, nightlife, landmarks, and tours. I also thought about being conscious of the weather/seasons and whether or not visitors could be outside. I had a stacked list of amazing NYC options. The only problem with this research (and exists in my bot as well) is the fact that New York has so many options. How is one to pick? NYC really is a concrete jungle of too many choices. With that in mind, I tried to select places with 5-star reviews that were also a New York-specific staple.
Building the bot:
Once I had my prime tourism locations I took to the bot to create it. This is certainly where I ran into the most issues as SendPulse doesn’t allow for as many buttons as I was hoping. Once I saw this issue I learned I need to reroute my plan a bit.
That being said, I still started the way I always intended to. First, with a warm welcome and then an ask for what part of town this visitor was from. The first three options were what we identified up above, Uptown, Midtown or Downtown. I also thought about if a visitor is coming to NY for the first time ever. Maybe they are directionally challenged and have no idea where they are. That’s when I realized I should provide a map for assistance. Once the location was identified my follow-up was if they were on the East or West side of the island.
From there, I had even more categories. It occurred to me while I was picking locations for visitors that there are many types of people visiting, not one size fits all. A family trip is very different from a weekend with friends. That was my next question for the bot. Are you with family (kids) or friends (adults)? From there, the user needs to decide what sort of activity they are looking for. Is it food? A museum? Shopping? I allow for all those responses. The more details I receive from the user the more tailored my response and suggestion can be for them.
Here is when I ran into two issues: character limit and button limits.
SendPulse has a 500 limit word count for every message shared. This means I couldn’t provide a ton of details about the destination. I had to wordsmith my copywriting in order to convince the user. The solution I found for only three buttons available per message was to add more messages. This meant I had to get more specific and thoughtful with my bot setup. When I was listing dinner options I added a follow-up message listing dessert. When listing attractions/museums I listed a follow-up option for something to do outdoors.
The last step for my bot was always a link for the attraction. Since I was showing users options outside of my realm, the website was the best option for leaving my bot and getting tickets, the specific address, the menu, or learning the hours of that destination.
Surveying my audience:
The field test was intended to see whether having a 24/7 attraction resource would encourage people to learn and visit new parts of the city they wouldn’t know otherwise. This bot was made for both locals and people who have never visited. It’s all an effort to just get people back out there. I envisioned my target audience to be any gender or age but probably on the younger side (under 50) as New York is a bustling walking city. At the end of the day, anyone can benefit from these recommendations about our history.
I sent 20 people my chatbot and encouraged them to share with others. After they played around with it, I asked them a few questions about the experience. If they shared with people who weren’t comfortable speaking with me, I asked the sharer to bring back any insights. The people ranged in age from 16 to 55. Here is the message I sent:
Hi there! I generated a chatbot about NYC attractions. My goal here is to encourage people both near and far to jump back into the city that never sleeps. I hope you can use this tool as a resource to be reminded of the amazing things this city has to offer, or learn of a new destination you never knew existed! Please click the following link and engage with my bot: LINK
Some questions I asked users after they engaged:
- What did you enjoy about the interactivity?
- Was there anything choppy or not so great about the user experience?
- What did you feel was missing?
Below are my key findings
- 50% of users wanted more options
- 40% of users wanted a natural flow
- 30% of users wanted to stay in the bot
- 75% of users loved the options available!
My biggest learnings were that the bot felt like an expert, which is a great goal I achieved and the users loved the options presented. That being said, the users wanted more options and didn’t want to leave the bot.
Here are some positive quotes from two test subjects. One said “these recommendations were great, so specific, creative, and fun,” another said, “I have lived on the UES for 3 years and never knew you could go on a gossip girl tour”. That captures my two goals of my bot being an expert and capturing good suggestions for users.
On the other hand, one user said “some of these didn’t seem like a perfect fit for me, I wish there were more options” another said “I didn’t like getting directed to an entirely new site once I picked my destination. I wish I could stay with the bot to make a reservation or buy tickets.”
The only main solution for more options is my own user input. In a perfect world if I were executing this for a business I would include as many options as I could and constantly be updating/swapping them out. For the trial's sake, I kept it small. As for leaving the bot I know it’s possible to set up reservations or tickets within the interface that just bleeds more into coding which I would need an expert for or more education on as it’s a bit more advanced.
Conclusion
After generating my bot I learned how important it is to understand my users. Without a feedback loop, I’m limiting the intelligence my bot could provide. Looking at what users are actually saying to the bot helps me understand how my bot can anticipate turns in the conversation. I think some classic issues among all bots are missing intents, poorly handling of failure cases, or repetitive/unnatural language. User feedback is a valuable resource in solving those problems. Turning feedback into improvements comes from two places:
- Manual improvements: This is on me. Taking a look at feedback data and making improvements accordingly. For example, I can make manual adjustments by adding more options for my users and changing the language my bot uses.
- Automated bot improvements: Using the technology of the bot. This one is harder for me as I am still learning how sophisticated a chatbot can be. As much as I know about bots, there is still so much to learn. Some issues seem out of my realm of expertise and I do not know how to automate a more seamless chatbot experience (yet!).
Overall, I think New York City tourism will increase for locals and distant visitors as many users expressed how educational my bot was. If this were implemented for a business or brand I can see the success that could come out of it. Especially if it were linked with already popular tourism resources on Instagram and Facebook such as Time Out New York.
Sources:
“Chatbots: What Is a Chatbot and Why Is It Important?” Expert.ai, 26 May 2021, www.expert.ai/blog/chatbot/.
Streck, Hannah. “New York Events and Things To Do All Year.” Time Out New York, Time Out, 5 Feb. 2018, www.timeout.com/newyork.
“NYC Boroughs: Manhattan.” NYCgo.com, www.nycgo.com/boroughs-neighborhoods/manhattan/.
Mcgeehan, Patrick. “New York City to Reluctant Tourists: Please Come Back.” The New York Times, The New York Times, 24 June 2021, www.nytimes.com/2021/06/24/nyregion/nyc-tourism.html.
Nelson. “New York Faces Lasting Economic Toll Even as Pandemic Passes.” The New York Times, The New York Times, 20 June 2021, www.nytimes.com/2021/06/20/business/economy/new-york-city-economy-coronavirus.html.