Chatbots can be a gold mine, a mini side hustle or a waste of time. Like everything you get back what you put in. Hopefully this post will inspire and put some money in your pocket along the way.
We started cashbot.ai in earnest over the summer, but enabled self service in late October, so I consider us to be at our 45 day mark right about now. We thought folks might be interested in hearing about what we learned, so in this post we’ll share some of the more interesting observations & what we’re doing about it.
What is cashbot.ai?
If you’re wondering, cashbot.ai is a monetization platform for chatbots. It does 3 things well, 1) provides a path to monetization for a developer or operator of a chatbot 2) Simplifies the process of making a product recommendation 3) Improves the product recommendation itself IE improving CTR & conversions.
We have also dramatically simplified integrations into Chatfuel, your nodeJS & REST bindings. If interested in receiving a Chatfuel & Cashbot template email us at email@example.com and we’ll get right back to you with a link.
(shhhhhh Telegram is next)
What We Learned.
General observation: Chatbot users are very receptive to product recommendations, when done right.
The most non-invasive option is adding a section called in your flow called “Support me” where a user can click and see products you recommend. A user can chose to purchase as a form of support for your bot. This is a totally transparent way of monetizing your user experiences and doesn’t take away from the experience in any way. If you’re providing value why not ask for some support?
Another common route is to weave into your user experience the recommendation of tangential products to the theme of your chatbot. Have a fitness bot? Recommend supplements & training gear. Running shoes, Under Armor.
A cooking & recipe bot? Recommend flatware, pans and other items needed to cook those meals you’re suggesting. A weather bot? Umbrellas when it rains, winter coats when it snows, shorts when it’s sunny. A dentist bot? Toothpaste and toothbrush.
Key here is to get creative and try different ideas. You don’t want to disrupt your existing users, you want to enhance their experience by showing them things you actually think they should buy. I love Seinfeld Bot!
Then there’s bots that are intended to help find you products, facilitate product discovery. Gustav the gift giver is a great example of this, a fun chatbot that’s always surprising you with all sorts of products for all sorts of people in your life. Gifty does a great job of this as well.
The team at Chatsecrets.com are designing educational courses meant to show you how to make the transition from creative, to data driven marketing & product recommendations 100% chatbot based.
What We’re Doing About It.
Observation #1: Developers naturally segment their chatbots into categories & domains.
You got a Fitness chatbot, Joke Bot, Trivia Bot, Lipstick Bot, Weather Bot? One of those, all of those? So why not make things easy by allowing users to pair their chatbot category with product categories?
You can do that now.
Change #1: You can now train your agents by category. Every category Amazon classifies its products by.
Observation #2: Our customers search for products differently.
Some search using Brands as the keyword, some the manufacturer and some search by title. Everyone goes about finding what they want a little differently and if it takes too long, we found our users would give up.
We fixed that.
Change #2: You can now apply your key words to everything, or just to a product Title, Brand, Manufacturer, etc…
Observation #3: Price is really important. If you’re operating a Horoscope chatbot you don’t necessarily want to recommend a $4,000 TV. If you’re Kushnode, you probably want to recommend snacks, Dorritos or even a streaming movie. Users want to control the price range of the products recommended and filter accordingly.
You can do that now.
Change #3: You can now filter by price range. Only want items that cost between $5 & $20? No problem.
Observation #4: Sometimes our customers want to recommend all sorts of things, so they’re less concerned with relevance and more concerned with popularity. They seek variety and surprises. Whatever’s popular, that’s good enough.
Change #4: You can now sort by Most Popular. The default is relevance, but if you want to include popular items that are commonly purchased, we can help you there too.
Observation #5 Sometimes our users wanted to recommend something very very specific. Building a bot that recommends throwback Nikes? Build an agent that you want to recommend Air Jordans for men, all years and styles? It simply wont cut it if a recommendation for a pair of female Addidas slips in there. Our users requested precision, specificity and ultimately sometimes total control.
Very excited about this enhancement… My Air Jordan bot is killing it!
Change #5: You can now delete individual products from your agent database – see picture! This is HUGE!!! You can see how many products are in your agent database and one by one delete them if they don’t fit with what you wanted.
Observation #6 Every detail counts. We discovered that sometimes the product recommendations would have weird titles or the product description was unclear or didn’t make sense.
Change #6: You can now change the product description & title for each product. Sometimes Amazon returns a weird description, even a funny name. Now you can use these fields as calls to action, or summarize exactly what you want to say when describing the product. Got a birthday bot? Describe all your products as “perfect for birthdays!”
Observation #7 Price matters.
We received a lot of requests from our users to include the ability to recommend items on sale, or to start at least let users know what products were on sale. If you’re building an Ecommerce bot and you have a interaction flow that’s 100% about deals, you have to be positive what you’re recommending has a discount.
Well now you can.
Change #7: We now show price, discount and sales in the product training page. Note: price tracking and advertising via recommendations coming soon but not yet available.
Observation #8 The chatbot community is truly an international one.
We’ve received feedback from all over the globe and have users in over a dozen countries, none the less our initial hypothesis was always that we’d be focused on US based developers. What has surprised us has been the international flavor and interest levels coming from Eastern Europe, Scandinavia, India & Korea. Truly inspiring and while initially a challenge, we’re thrilled to be working with such a global user base.
Change #8: For our international users… We will be rolling out support for every Amazon locale later this week!
Our last observation was more about self reflection than product enhancements… After 45 days we are amazed at how much innovation takes place daily & how creative this community is. Despite working in this space for the last 3 years and having built multiple chatbots ranging from bespoke to massive, we’re still learning something new every day. It’s truly exciting to be part of this technology boom at an early stage and help the ecosystem expand
But that’s not all, we’re really interested in hearing what you’ve learned about Chatbots in the last 45 days, or 3 months or in 2017?!?!
P.S. If interested in receiving a Chatfuel & Cashbot template email us at firstname.lastname@example.org and we’ll get right back to you with a link.
If you enjoyed this post please support with a share or an upvote or whatever. Sharing is caring!
Thanks and we love you all!