Many of my coaching clients have at some point looked at building Facebook Messenger Bots. The 1-click integration provided by Dialogflow has many limitations, but even with that, the possibility of reaching a billion plus people with your chatbot is a very appealing one.
In this post, I am going to discuss why a website chatbot is a much better option for many businesses.
You might be thinking, “Well, Aravind, you help people build website chatbots. You will obviously be biased in that direction”
My view is, I am actually choosing to concentrate on website chatbots over Messenger bots because I see a lot of potential. And frankly, I might be wrong. Who can predict the future? From the knowledge I currently have, this is my viewpoint.
With that said, here are my reasons:
You control the deployment process
As you know, on Facebook, you will need to get approval before you can deploy your bot. If your bot needs special permissions (e.g. sending daily notifications to your users), you need to apply for additional permission with a longer and more challenging vetting process.
With a website chatbot, you completely control the deployment process, and no one is going to tell you what to do and what not to do.
You can make it infinitely flexible
This one is a straight forward reason. There is very little you cannot do on your own website, so you have the option to make your chatbot’s user interface infinitely flexible.
For example, in one of my website chatbots, I allow people to change the color of the chat boxes. Now, this is quite a trivial example, but you can see the potential of having this option.
While people are interacting with your messenger chatbot, odds are they are constantly being pinged with other updates from their friends. Whom do you think they would prefer to communicate with, their friend or your bot?
Of course, chat bots are asynchronous, meaning they can always come back to the conversation later. Now imagine that you are one of the hundred thousand plus bots on Facebook. How certain are you that they will actually come back to your bot?
On your website, though, you actually control the distractions (to a large extent).
SEO benefits from user engagement
Once people actually start engaging with your website chatbot, it will start helping you with your SEO. For example, your bounce rate will probably go down. The time on page is likely to increase. And not to mention, if people are telling you about their problems, you can write an article about it later.
For example, I have a survey bot on some of the pages on my site right now and I have already figured out that quite a few people want me to write more about webhooks, which I am planning to do over the next few weeks. This creates a positive feedback loop, obviously, because when you start writing about topics people care about, you will probably get more search traffic to your website.
A lot of people have pointed this out and they make much more persuasive and articulate arguments: choose to give attention to your website over social media. Unlike almost all other marketing channels, it actually gets easier as your website gets more popular, and you don’t have to subject your business’s success to the whims of the social media platform you use (See next point).
You will not be penalized for your bot’s success
When Facebook was initially promoting its Messenger bots, you had the ability to send daily notifications to your users from your Facebook Messenger bot, without applying for permissions from Facebook (phase 1). But at some point it changed to require permission from Facebook (phase 2).
Odds are, once enough people compete for this permission, Facebook will be charging them money for it (phase 3). This is pretty standard playbook for Facebook anyway:
Phase 1: Facebook initially needs users for their platform – make everything free and simple
Phase 2: A critical mass is reached, and investment into the platform increases
Phase 3: The bot creators are already heavily invested and need Facebook’s help to reach the same people they could reach very easily before. At this point, Facebook starts a bidding war.
Who will actually compete for the permission? Those who have successful bots.
Maybe you are wondering if there is a small window where you can get the benefits of this Facebook marketing without paying a high cost. I am sure it is possible, but even then you will have to ask: what is the long term play?
The web page already sets a context
This is something which is probably unique to a website. Let us say you are a consultant and the user is interacting with the bot on your website on a page talking about a specific topic.
If the bot is at the bottom of the page, they probably only noticed it after reading the article, meaning you already have some context for what they are looking for. In this case two things are possible
- Your bot can provide better answers because it has some context
- You can use the context and ask your reader better questions too. E.g. A bot located at the bottom of this page might start by sending the following message “I see that you are interested in building a chatbot on your website. Would you like to learn more about how Dialogflow can help you do this?”. I haven’t added this feature yet, but I know it is possible. 🙂
Needless to say, if the bot was on a product page on your e-commerce website (say), you now have a really large quantity of page specific “contexts” which you can use to better serve your customer and directly measure your bot’s ROI.
One of the really cool things about a website chatbot is the much superior analytics you can get (although obviously it requires more work on your part).
A simple example: the web page the user is chatting from can already tell you a lot about the visitor. If you combine this with conversational analytics, you already have a far better picture of what the user wants.
I have quite a bit more to say on this specific topic. I will be updating this post with some examples from my own bots, so you may want to come back to this post at some point in the future.
A/B testing (and the Law of Large Numbers)
A chatbot which resides on your website has something very important going for it. (this does assume you have at least 100s of visitors a day). When your site gets to a certain scale, because of the law of large numbers, you will already be getting useful and somewhat predictable data from how people interact with your chatbot. E.g. How many people visit a given page? Of those, how many interact with your chatbot? How do they interact with your chatbot? This means you can set up some A/B tests if you want to improve your bot engagement numbers.
For example, here are a few things I could test quite easily with the chatbot on the page below. Again, I haven’t actually implemented them, but I am talking about things which are doable simply because I have complete control over my website
- does an “online” indicator light make someone more likely to interact with my chatbot?
- does it help to show that the conversation has ended?
- will users be willing to click on a navbar menu to choose options (e.g. Start over, Go back one step)
- Are buttons and other rich controls more engaging than the default text response? (although it is a little more challenging to set up a test for this)
You simply don’t have that much flexibility to run such tests in your Messenger bot.
So, what do you think about this viewpoint? Leave a comment if you feel strongly one way or the other 🙂