We talk to robots several times per day. If you have Amazon’s Alexa, talk to Siri on your iPhone, or chat with friends on Facebook Messenger, then you use chatbots – computer programs designed to talk to you as if you’re speaking to a human. While these products won the first chatbot battle, research shows that they might lose the war to small businesses.

In 2016, investors poured $188 million into 90 startups building text and voice-enabled bots. And Gartner predicts that small businesses will outperform big brands by building bots that do a better job of fulfilling customers’ unique needs.

There’s a big gap to fill if startups can get this right. GetApp research found that one in three shoppers are most impressed by having in-store salespeople. This suggests that although customers might stand in awe of new tech trends, they don’t trust tools like chatbots to replace people yet.

Analytics for chatbots?

So, let’s say you’re a small business owner who’s on board with building chatbots. There’s still a big question: how will you tell if they’re working or not?

To answer this question, we spoke* with Arte Merritt, CEO and Co-Founder of Dashbot.io. Think of Dashbot as Google Analytics for chatbots. A digital marketer might use Google Analytics to pull metrics like the number of visitors to a web page; the number of subsequent web pages the average reader went to; and a web page’s average conversion rate.

A small business owner can likewise use a tool like Dashbot to set key performance indicators (KPIs) for their chatbots. Then, they can track how well those chatbots perform based on how customers engage them.

Read on to learn:

  • Why traditional analytics tools can’t measure chatbot performance;
  • Which small businesses are using chatbot analytics today;
  • How to tell if you should build your bot in-house or outsource the development work.

In layman’s terms, what is a “bot” and how are bots used in small businesses today?

When we say bots, we’re referring to conversational interfaces. Bots are applications that users interact with via conversation. Bots can be text-based like Facebook Messenger or Slack. They can also be voice-based systems like Amazon Alexa or Google Home.

The use cases are quite varied. People tend to think that chatbots are only used for customer service. The truth is that these applications are much broader – we don’t see that many customer service bots. Think of all the genres – news, sports, travel, retail – that you see for mobile apps. Chatbots can be used across all of these genres as well.

Let’s take personal bots as an example. Some celebrities, like Redfoo of LMFAO, have chatbots that their fans can interact with. Every so often, they’ll pause the bot and interact directly with the fans. There are also virtual character bots that can be quite engaging.

How do bot analytics work? What’s the difference between using a tool like Google Analytics vs. Dashbot?

Traditional analytics tools don’t work well for bots for the following reasons:

  • The tracking mechanisms are different
    • Anything that is clickstream, or event-based tracking, loses the richness of messaging;
  • The data captured is different
    • Bots receive unstructured data because users send text, images, audio, video, etc. Most importantly, they also send their own voice and words to say what they want from the bot, and what they think of the bot afterwards. This is quite different from how users interact with online content by clicking links and buttons.
  • The processing is different
    • Bots can be asynchronous and have multi-user sessions. This is quite different from one-to-one interactions on web and mobile.
  • There are new types of metrics
    • In addition to traditional metrics like retention and engagement, there are bot-specific metrics like sentiment analysis, AI response effectiveness, and conversational message funnels.

A strong chatbot analytics service provides all the traditional analytics. But they’ll also provide the bot-specific metrics mentioned above – sentiment analysis, conversational analytics, AI response effectiveness, and even the full transcripts. And Dashbot includes features that let users take action based on the data they find. These include our Live Person Takeover of sessions and our Push Notifications for re-engagement.

Describe a case study where a small business used Dashbot. What quantitative impact did it have on their business?

One of our customers is Machaao, a top bot on Facebook for cricket fans to follow their favorite teams and players, and get score updates. Two of the founders spend 30 minutes a day looking at Dashbot analytics to see how their bot is being used to increase engagement and retention. They got this information via top message reports, message funnels, and the full transcripts. This let their developers traverse through the bot to see which messages users sent in and how the chatbots answered them.

Dashbot provides top message reports, message funnels, and even the full transcripts to enable developers to traverse through the bot to see the messages users send in as well as the bots responses. These reports are particularly useful for improving bot response effectiveness.

Machaao’s goal was to make their bot’s responses more effective. So, they created three new features based on insights from our reports:

  • Live Scores
    • Users kept asking for live score updates. Before, users had to send in a team name to subscribe and get notifications. Now, users can click a quick “Reply” button to get live score updates anytime.
  • “Follow Player” 
    • Machaao had wondered if they should build a feature that would allow users to follow specific players. The problem was that they weren’t sure how high user demand was. Based on the messages in their Top Message Report, they saw that users were asking the bot for player details. That knowledge moved this feature up on their priority list.
  • Mute/Pause
    • Previously, if a match was one-sided and a user’s team was losing, they would get upset and unsubscribe so they wouldn’t have to see score updates. This created a vicious cycle where Machaao paid to re-acquire users through ads. Adding the “Mute” feature allows users to pause notifications without unsubscribing from the bot. This helped Machaao boost user retention and save money.

Describe the role that bots will play in small businesses four years from now. Do you think they’ll replace employees (which could save small business owners money, but negatively impact the job market)?

We’re strong believers in conversational interfaces. We think it’s the natural evolution of human-computer interaction.

2017 is shaping up to be the year for voice-enabled interfaces. We see it in the signups on our platform and inbound requests from brands that want to build for Amazon Alexa and Google Home. There is a lot of demand. Bots can also free people up to work on new ideas and activities, which can result in new types of jobs and skills.

Which skills should small business owners acquire – and hire for – to use bots most effectively?

Prior to building a bot, you need to think through the “why” and answer this question: “What’s the use case and strategy?” A small business owner needs to think about their target audience, the use cases, and the product. Then, they’ll have to consider all of these in terms of user acquisition, engagement, and monetization. This is no different than the process that one goes through when considering a new business or product idea.

There are several options available to build bots depending on company resources. There are bot frameworks to speed up development, bot platforms to build bots, and bot development companies and agencies. You could also build the bot yourself from scratch.

Each option demands varying technical skills. If you don’t have a developer on your team, outsourcing will be the best option. But if you do have the resources to build a custom bot, you can save time and money by doing the work in-house.

Once the bot is built, the next step depends on your use case. You might have to think about marketing, user acquisition, engagement, retention, monetization, or some combination of these.

Once you’ve determined which metrics you’ll use to manage your bot’s success, you’ll need to track how users engage with it. This is essentially the same way that digital marketers use analytics to learn more about how users engage with their content.. And once you know how users interact with your bot, you can use those insights to improve the bot experience.

*This interview was edited for length and clarity.

Read these articles to learn how small businesses use chatbots for HR, marketing, and customer service: