Chatbot ROI for a Store With 1,000 Monthly Questions

A store owner opens Monday's support report: 1,000 questions came in last month, and a chatbot vendor says it could answer a quarter of them. What would the chatbot ROI actually be? In the fictional example below, the answer is 25% only if 250 questions are truly resolved, each saves $4 of avoidable work, and the bot plus review costs $800 a month.

That is a worked assumption, not a HoverBot customer result or a promise about your store. The owner can replace every input. The important question is which costs change when a bot handles a conversation, because a fluent reply and a saved support expense are two different events.

The chatbot ROI formula in plain words

Start with eligible questions, not all website visits. Multiply by the share that end without a person having to finish the same job. Multiply that count by the support cost you can actually avoid. Subtract every recurring bot cost and the time spent reviewing wrong or incomplete answers. Divide the remaining value by those costs. Put as a formula: ROI = (avoidable support value − bot and review cost) ÷ bot and review cost.

This is a deliberately narrow operating calculation. It does not assign imaginary revenue to a nicer chat or count a shopper who left silently as a successful sale. If a bot helps staff answer faster without reducing paid hours, the benefit may be capacity, not cash. Record that separately instead of putting it into the savings column.

Zendesk defines cost per resolution as total support cost divided by verified resolutions. That is a better starting point than a vendor's broad deflection claim, but it still needs a local adjustment: rent and a salaried team's fixed hours do not disappear every time one chat is automated. Ask what cost or capacity can change in the period you are measuring.

A worked example with fictional store numbers

Suppose a fictional homewares shop receives 1,000 eligible support questions in a month. The owner samples the chat history and plans for 250 of them to be resolved by the bot without a teammate doing the same work later. The owner values the avoidable handling at $4 per resolved question. For this exercise, the bot and its integrations cost $600 a month, while a teammate's review and correction time costs $200. These are invented round inputs for the calculation, not HoverBot prices, support benchmarks or observed results.

Fictional inputs and arithmetic to replace with your own figures

Step	Calculation	Result
Verified resolutions	1,000 × 25%	250
Avoidable support value	250 × $4	$1,000
Full monthly cost	$600 + $200	$800
Net monthly value	$1,000 − $800	$200
ROI	$200 ÷ $800	25%

The break-even point is more useful than the headline percentage. At $4 of avoidable work per verified resolution, the shop needs 200 such resolutions to cover $800. That is 20% of its 1,000 eligible questions. If only 15% resolve, the value is $600 and the month is $200 short. At 30%, the value is $1,200 and the net is $400. A small change in the true resolution rate changes the buying decision.

Count resolved questions, not just chatbot answers

A bot can answer a question and still leave a shopper needing help. The reply may cite an old returns page, miss the order's actual carrier status, or answer only the first half of a request. A person then completes the work, sometimes after the shopper repeats the story. That conversation belongs in the review pile, not in the savings total.

Intercom's reporting definitions distinguish answered conversations from resolved conversations and describe confirmed versus assumed resolutions. That distinction matters even if you use another platform. A customer who explicitly confirms the answer is stronger evidence than one who simply stops replying. Sample both groups, then look for repeat contact on the same issue. Keep handoffs visible: a bot that sends the right exception to a person can be useful without pretending it resolved the case alone.

The fictional store should therefore audit a sample of its claimed 250 resolutions. If the bot's answer needed human cleanup or the shopper returned with the same question, revise the 25% input. Count the reviewer's time and any extra contacts as cost. This is why we would judge a pilot on labelled conversations and their follow-up, not on the number of messages the model produced.

Decide whether the gain is cash or capacity

For a store paying an outside team per resolved ticket, fewer eligible tickets may cut a bill. For a small team on fixed salaries, 250 resolved chats may instead give people time for complex cases, product advice or faster replies. Both can matter. Only call it a cash saving if a cost really falls. If the value is capacity, record the hours freed and what the team did with them, then decide whether that outcome justifies the expense.

Also inspect the mix. Order-status questions with a connected tracking source may be easier to verify than refund disputes. A single average rate can hide the expensive cases a bot hands back. Group questions by job, write down which ones need a person, and estimate the avoidable cost for each group. Start with one repeatable group before applying the percentage to the whole inbox.

Our verdict: a chatbot ROI claim is worth using only when it names the denominator, the evidence for resolution and the costs that actually change. A vendor percentage may help frame a pilot; it cannot fill in your shop's spreadsheet. Try the numbers in the ROI calculator, then ask a fictional store bot a real catalogue or policy question on our industry demo hub to see what an answer and a boundary look like.

Five checks to run this week

1. Count one month's eligible support questions by job, such as order status, returns and product advice.
2. Calculate the avoidable cost of handling one verified resolution for each job; separate fixed staff cost.
3. Sample bot conversations and label confirmed resolution, assumed resolution, handoff and repeat contact separately.
4. Add subscription, setup, integration and ongoing review time to the monthly cost, even when a vendor leaves them out.
5. Find your break-even resolution rate in the ROI calculator, then run the same calculation at a lower rate before deciding.