AI Chatbot ROI for Small Business: A Simple, Honest Method

Chatbot ROI for a small business should start with the business outcome, not the number of messages. Estimate the extra qualified enquiries the chatbot produces, multiply by your close rate and average gross profit per job, then subtract software cost and the staff time spent reviewing. If you cannot measure "extra", report the result as directional, not proven.

Why message counts mislead#

"Our chatbot handled thousands of conversations last month" sounds impressive. It tells you how busy the bot was. It does not tell you whether a single extra job was won, whether staff saved any time, or whether the bot gave someone a wrong price.

A chatbot creates value in a lead-capture workflow in two ways:

1. More or better enquiries: people who would otherwise have left now leave usable details.
2. Avoided work: routine questions that no longer need a person.

Both need measuring, and both are easy to overstate.

The five inputs#

For a lead-capture use case, gather:

• Qualified enquiries from chat, using a written definition of "qualified" (valid contact details, in your service area, a service you offer, a stated need).
• Baseline qualified enquiries from the same traffic source before the chatbot, or from a comparison period.
• Lead-to-customer rate: the share of qualified enquiries that become paying jobs.
• Average gross profit per new customer, not headline revenue. A A$1,200 job with A$800 of materials and labour is worth A$400 for this purpose.
• Monthly cost: software plus the time someone spends reading transcripts, fixing answers and following up.

The formula#

Incremental monthly value = incremental qualified leads × close rate × average gross profit
Net value = incremental monthly value − software cost − operating cost
ROI = net value ÷ (software cost + operating cost)

A worked example (hypothetical numbers)#

These figures are invented to show the arithmetic. They are not results from any Pivra customer.

• Baseline: 20 qualified enquiries a month from the website.
• With chat: 26 qualified enquiries a month, over a comparable period.
• Incremental qualified leads: 6.
• Close rate: 40%, so 2.4 extra jobs.
• Average gross profit per job: A$350, so incremental value is A$840.
• Software: Pivra Starter is US$29 a month, or A$39 + GST in Australia (pricing). This example uses the A$39 + GST price before GST is added.
• Operating cost: 2 hours a month reviewing conversations at A$60 an hour, A$120.
• Net value: A$840 − the Starter price before GST − A$120 = A$681. ROI: A$681 ÷ A$159 ≈ 4.3.

Change one assumption and the answer moves a lot. If only 2 of those 6 extra enquiries were genuinely incremental, the net value falls to about A$121. That sensitivity is the point: be honest about which inputs you measured and which you guessed.

Is the lead really incremental?#

Not every contact captured in chat is new business. Some visitors would have called or used the contact form anyway. Ways to get closer to the truth:

• Holdout: if your traffic allows, show the chatbot on some pages or days and not others, and compare.
• Comparable periods: compare with the same weeks last year or a similar month, noting seasonality, weather events and ad campaigns.
• Separate cohorts: keep paid search and organic visitors apart. Paid visitors often behave differently.
• Keep the old paths open: leave the phone number and form in place so you can see whether chat adds enquiries or just moves them.

The comparison is far easier if you record a baseline of enquiries, missing details and response times for a couple of weeks before the chatbot goes live.

Count avoided work carefully#

Include time savings only when the work genuinely disappears. If the assistant answers 200 routine questions but someone still reads every transcript, the saving is smaller than it looks. Measure with a sample: time a week of answering the same questions by phone or email before launch, then time the review work after.

Subtract the downside#

Track what goes wrong, too:

• Wrong or unsupported answers.
• Spam or junk leads.
• Alerts that did not arrive.
• Poor-fit bookings.
• Follow-up delays because nobody owned the inbox.

One costly error, such as a wrong price honoured to keep a customer, can outweigh many routine conversations.

What Pivra can and cannot measure for you#

On Starter and above, Pivra's analytics show conversations this month, total leads, a lead conversion rate (leads captured ÷ conversations over the last 30 days), visitor thumbs-up and thumbs-down ratings and knowledge-source health (analytics). You can export leads as CSV from the Leads page and conversations as CSV from the Conversations tab (lead capture).

What Pivra cannot see is what happened after the enquiry: whether the job was quoted, won or lost. Join your exported leads to outcomes in your own job-management system, CRM or spreadsheet. If you use ServiceM8, Pivra can create the enquiry as a job there so the outcome lives in one place (ServiceM8 integration).

Analytics and lead capture are not on the free plan. The free plan lets you test answer quality on your own content; to measure leads, use a paid plan, which can be tried free for seven days with a card at checkout.

For a broader view of what chatbots and the alternatives cost, see how much an AI chatbot costs a small business in Australia.

A realistic first goal#

In the first month, the honest goal may simply be proving that the assistant captures usable enquiries outside business hours, not a positive ROI after seven days. Write down your decision date and criteria before you start: continue, change the workflow, or remove it.

Monthly ROI checklist#

• Same definition of "qualified" as last month.
• Leads exported and matched to job outcomes.
• Gross profit used, not revenue.
• Review time recorded.
• Errors, spam and failed alerts listed.
• Assumptions written next to the result.

Next step: calculate value from qualified outcomes and gross profit, not message volume, and write your assumptions beside the number.