How to Calculate ROI for 3 Types of Chatbots

Clearly define your KPIs

While it might sound like an obvious one, you can’t measure ROI without KPIs (key performance indicators).

If you’re building a customer service chatbot, your KPIs might include:

• Number of requests resolved by the chatbot
• Reduction in average response time
• CSAT
• Deflection rate (how many conversations are diverted away from reps)
• Cost per query

An HR or IT chatbot might have KPIs like:

• Number of requests diverted from employees
• Number of work hours saved by funneling requests to a chatbot
• Time it takes an employee to book vacation days

Your company’s chatbot KPIs will change depending on the type of bot you design.

Tie your KPIs to monetary values

Each KPI should be tied to a specific monetary value. It’s not enough to say a project will ‘save time’ or even ‘save 10 hours per week’. Calculate how much money it will save every month or year. 

If your AI chatbot will quell your need to hire 2 extra employees, ensure that these numbers are taken into account, too.

Start low, then increase

While it’s tempting to maximize your chatbot’s impact from the get-go, our Customer Success team recommends aiming for a minimal value ROI at first.

Focus on incremental gains. Once the chatbot has proven effective with this initial load, gradually increase its scope – and optimize along the way.

This measured approach not only allows you to track ROI with greater precision, but ensures that your chatbot’s performance remains high as its responsibilities grow. By focusing on incremental improvements, you can scale the chatbot’s capabilities in a way that minimizes risk and maximizes long-term success.

Starting variables

Start with:

• The hourly salary of a call center representative
• The average length of a customer support call
• How many calls the center takes each month

And find the average cost of a customer support call for your organization. If a representative makes 20 an hour and the average support call is 5 minutes, then the average cost of each support call is $5. 

Better chatbots bring higher value

Then add in the value of a chatbot. Identifying the right expectation can be tricky – we’ve seen plenty of clients vastly over- and underestimate the value of a chatbot. 

The value add with depend on a few factors, including:

• The similarity between customer support calls (repetitive calls = higher value add from a chatbot)
• The complexity of your chatbot (dumber bot = less value add)

At Botpress, our Customer Success team sits down with clients and prospects to help them identify realistic value estimates.

Calculating your ROI

For our example, we’ll assume a call center receives 3000 calls per month, and that their chatbot will take on 2000 calls per month. The chatbot will save its company 2000 x 5 (number of calls x price per call) each month, adding up to $10,000.

This number will grow as the company expands – instead of hiring 3-5 people, they can hire 1-2 to tackle the same volume of calls when assisted by the chatbot.

Starting variables

Start with:

• The hourly salary of the HR representative and the average employee seeking their services
• The average length of an HR interaction, including overall communication time (e.g. including time spent between other parties, like when an HR rep consults an accountant or manager)
• How many interactions occur per month that can be handled by the chatbot

This latter step will require some thought on which services can be taken over by AI. Vacation requests or inquiries about policies can definitely be handled by an AI chatbot, but sensitive conversations likely won’t be.

Data-backed design

Identifying the right use cases is paramount to designing a high-value internal chatbot. Identify the pain points that take the most time from both parties in order to get the strongest impact from the chatbot. 

These will likely be repetitive processes like onboarding, scheduling, sick days, feedback, performance reviews, or queries about policies and procedures.

Number of leads

Often clients will choose to measure lead gen bot ROI by how many people are talking with the chatbot. Picking the number of leads generated by the bot can be a tempting KPI, but this negates the importance of each lead’s quality. 

We’ve seen clients with led generation chatbots that brought in lower overall numbers, but a far higher quality of leads. This isn’t uncommon – human error can push employees to include or exclude leads based on information that isn’t backed by data. 

An AI chatbot, on the other hand, will identify the ideal lead based on past data and stick to their framework.

If your sales reps qualify 7/10 leads, aim for your bot to qualify 5/10.

Starting variables

Start with:

• Number of monthly website visits
• How many visitors are interacting with the chatbot
• How many leads the bot is qualifying

The first step is to come up with a realistic number of potential leads – those who will converse with your chatbot. 

If you have 100 website views, you likely only have 40 people taking time to look at the content – a lot of website visitors are using your homepage for navigation alone. With these numbers, you might want to aim for 10 out of these 40 people interacting with your chatbot, with 8 of those 10 having a real back-and-forth conversation.

Decrease in emails and forms

A strong metric to track to see if your chatbot is providing value is the number of forms and emails your employees receive about information answered by the bot. 

If the chatbot is successfully providing this information to website visitors, the number of emails and form submissions should decrease.

You should aim to see a 20-30% decrease after proper implementation, and a full 60-65% decrease after maximizing the potential of your chatbot.

1. How long does it typically take to see a return on investment from a chatbot?

You can typically see a return on investment from a chatbot within 3 to 6 months, assuming it handles a meaningful volume of interactions and is integrated with your key business systems like CRMs or helpdesk. The speed of ROI depends on usage volume and how much manual work the chatbot automates.

2. What are common pitfalls or mistakes when calculating chatbot ROI?

Common mistakes when calculating chatbot ROI include failing to assign monetary value to saved time, ignoring deployment and maintenance costs, or assuming unrealistically high adoption rates. It’s also easy to overlook hidden savings like reduced churn or faster resolution times.

3. How can I forecast chatbot ROI over multiple years?

To forecast chatbot ROI over multiple years, estimate year-over-year increases in user interactions, efficiency gains, and expanded use cases. Include expected costs for improvements like model fine-tuning or new integrations, and project cumulative savings compared to baseline operational costs.

4. How does chatbot ROI differ between B2B and B2C companies?

Chatbot ROI differs between B2B and B2C in that B2C companies typically see quicker returns due to higher support volumes and automation of repetitive tasks, while B2B companies often see higher value per interaction, especially in sales qualification or client retention.

5. What data do I need to start tracking in order to measure ROI effectively?

To measure chatbot ROI effectively, track metrics such as total interactions, resolution rate, deflection rate (questions handled without a human), average handling time, cost per conversation, lead conversion rate (for sales bots), and reductions in customer support workload or staffing needs.