Measuring Chatbot ROI: The Definitive Formula

To measure chatbot ROI you do not need expensive analytics software, just an honest before-and-after calculation. This article shows the formula for setting a chatbot's benefits against its costs properly, works through an example for a trades business and names the cases in which a bot simply does not pay for itself.

Why many chatbot ROI calculations do not add up

Most presentations calculate the same way: number of chats times minutes per conversation times hourly wage. The result is an impressive figure that has little to do with daily life in the business. Not every chat would otherwise have kept an employee busy on the phone. Many visitors ask about opening hours that are also on the website, and some of the conversations end up with the team anyway.

These mistakes come up again and again in ROI calculations:

• There is no baseline. Without figures from before the bot, no change can be proven.
• Every conversation is counted as saved, including those that are then passed to a person after all.
• Internal time is missing from the cost side: maintaining answers, checking conversation logs, clearing up questions from the team.
• Additional revenue is estimated instead of counted, because evening enquiries are not marked as such anywhere in the order management system.

A solid calculation therefore usually turns out more modest than the one from the sales pitch. In return, it holds up when your tax adviser or your bank asks questions. What such an honest calculation looks like for customer service specifically is shown in cutting customer service costs with AI chatbots.

The formula: how to calculate chatbot ROI

The basic formula is the same as for any other investment:

ROI = (benefit in the period minus costs in the period) divided by costs in the period

A value of 1 means that every euro invested has brought in one euro of surplus. If the value is below 0, the bot costs more than it brings in. Always calculate for the same period, ideally one month, and spread one-off costs over a realistic period of use.

What belongs on the benefit side

• Working time saved: only enquiries the bot has handled completely, multiplied by the time your team used to need for them and by the hourly rate including employer's on-costs.
• Additional contribution margin: orders or appointments that demonstrably came through the bot and would probably have been lost otherwise, such as enquiries in the evening or at the weekend. What counts here is the contribution margin, not revenue.
• Avoided no-shows: appointments that take place thanks to a reminder or are reassigned after a timely cancellation.

What belongs on the cost side

• monthly licence or usage fee
• setup, spread over twelve or 24 months
• internal maintenance time for the knowledge base, spot checks of conversation logs and coordination
• where applicable, costs for integration with the CRM, calendar or order management system

Our article on chatbot analytics and the right KPIs describes how to capture the individual metrics during operation.

Example calculation for a trades business

The following figures are assumptions, not measurements. They show how the calculation is built. Plug in your own values.

Suppose a plumbing and heating business with eight employees receives around 300 enquiries a month by phone, website and WhatsApp. An AI employee answers standard questions, schedules maintenance appointments and takes fault reports. Of the 300 enquiries, it handles 120 completely, each of which used to take the office worker six minutes. In addition, there are around 40 enquiries outside office hours, some of which the business used to lose to competitors who responded faster. We assume that these result in four additional orders, each with a contribution margin of €250. For costs we use the Neurobots Basic plan: €399 a month for one AI employee with up to 500 conversations, plus a one-off €1,299 for setup.

Item	Assumption	Per month
Office time saved	120 enquiries at 6 minutes = 12 hours at €35	€420
Additional orders	4 orders at €250 contribution margin	€1,000
Total benefit		€1,420
Licence	Basic plan	€399
Setup	€1,299 spread over 12 months	€108
Internal maintenance	3 hours at €50	€150
Total costs		€657

This gives (1,420 minus 657) divided by 657, an ROI of around 1.16. On these assumptions, every euro that goes into the bot would come back as about €2.16, and the setup costs would be recovered after about a month and a half.

The cross-check with only one additional order instead of four is more revealing. The benefit falls to €670, costs stay at €657, and the ROI is practically zero. In this example, then, the office time saved hardly carries the bot on its own. Whether the investment pays off depends on the additional orders, and that is exactly why you should count them carefully.

Measuring chatbot ROI in five steps

1. Record a baseline. Count for four weeks before launch: calls, missed calls, enquiries by email and form, enquiries outside business hours. A tally sheet in the office is enough.
2. Time each type of enquiry. Have your team note on a few days how long it takes to book an appointment, give a status update or answer a price enquiry. Gut-feeling estimates are often wide of the mark.
3. Tag the source. Every enquiry that comes in via the bot gets its own tag in the CRM or order management system. Only then can you see later which orders actually came from it.
4. Separate resolved from handed over. Each month, analyse how many conversations the bot completed and how many ended up with the team. Only the first group counts as time saved.
5. Calculate after three months. The first month is distorted by setup and fine-tuning. From the third month on, you have figures you can work with.

If you then have to defend the result to shareholders or the bank, the article Measuring AI ROI and communicating it convincingly will help. To look beyond the chatbot, our article on the ROI of sales automation is worth reading.

When a chatbot does not pay off

There are businesses for which the formula almost always comes out negative. If you only get a few dozen enquiries a month and your phone is reliably staffed anyway, you simply lack the volume. It is similar with enquiries that need a lot of advice, such as individual custom work, where every conversation ends up with the master craftsman anyway.

A full order book also changes the calculation. If your fitters are booked three months ahead, extra evening enquiries bring little, because nobody can work through them. Then the only benefit left is the time saved, and as the cross-check shows, that is often not enough. Another point is easily overlooked: if nobody in the business has time to maintain the bot's answers, the quality of the conversations suffers noticeably after a few weeks.

Conversely, a bot is more likely to pay off when many similar enquiries come in, a noticeable share of them outside business hours, and each additional order brings a decent contribution margin. Many businesses underestimate how much depends on phone availability alone.

Frequently asked questions

How long does it take before I can judge chatbot ROI properly?

Allow three months. In the first month things get fine-tuned; in the second, conversation volumes settle. Only then do you have figures you can compare with your baseline.

Should I calculate with revenue or with contribution margin?

With contribution margin. An order worth €2,000 where materials and labour cost €1,700 brings you €300. If you calculate with revenue, you inflate the benefit artificially and end up making a bad decision.

How do I know whether an order would have come without the bot?

You can never be completely sure. A useful approximation is enquiries outside business hours and calls that used to go unanswered. If in doubt, count cautiously and only what clearly came through the bot.

Do I have to tell customers they are talking to an AI?

We clearly recommend it. The EU AI Act requires transparency when people talk to or write to an AI system. A short notice at the start of the conversation is enough, and customers usually take it more calmly than a surprise later on.

Conclusion

A good ROI calculation for a chatbot fits on one page: baseline, time saved, demonstrably additional orders and all costs including your own maintenance time. Calculate in advance with cautious assumptions and do the cross-check with worse figures. If the calculation still holds, the bot is a sensible investment. If it does not, you save yourself money and trouble.

The page on the digital assistant for trades businesses shows how an AI employee works in a business like the one in the example calculation.