This is an illustrative example with stated assumptions, not a client case: it shows how the math is built so you can redo it with your own business's numbers.
Many chatbot evaluations add up new sales and savings as if all of it were profit. Here the math goes piece by piece: how much margin the additional sales leave, what the freed-up time is worth, what it costs to keep the chatbot running, and how much goes to third parties. Every assumption is in plain view so you can replace it with yours.
The scenario
An apparel e-commerce business that sells through its online store and handles purchase questions on WhatsApp. Today a single person answers, and at peak hours questions wait.
The chatbot being evaluated in this example:
- Answers frequent questions (shipping, sizes, payment methods) using catalog information.
- Reports an order's status by checking the store.
- Asks the qualifying questions and passes conversations with clear purchase intent to a person, with a summary of what was discussed.
- Hands off to a person, following agreed rules, whatever it can't answer, complaints and exchanges. The team reviews the answers periodically.
- Sends a reminder to anyone who left an abandoned cart.
The example's assumptions
These are round values chosen so the math is easy to follow. None of them is a StriqTech rate or a measured result.
- Sales conversations per month: 1,000. Different people asking about a product. These are not messages: a conversation usually involves several back-and-forths.
- Conversion without the chatbot: 8%. Out of every 100 sales conversations, 8 end in an order.
- Conversion with the chatbot: 12%. This is the most uncertain assumption: nobody knows it before trying. Further down you'll see what happens if the improvement is half as large.
- Average ticket: USD 60.
- Gross margin: 40% of the ticket. What's left of each sale after the cost of goods and the variable costs per order (payment processing fees, shipping you cover). If your margin doesn't deduct those costs, deduct them before using it.
- Chatbot setup: USD 3,000, one-time payment. An assumed quote: replace it with the one you have.
- Chatbot monthly cost: USD 220. The provider's maintenance and hosting. It's an assumption, like the setup: replace it with the one in your quote. Every proposal should spell out what it includes.
- WhatsApp Business Platform fees: USD 40 per month. A third-party cost billed by Meta. Replies to someone who wrote to you first, within the 24-hour window, are not charged; templates you initiate, such as the abandoned cart reminder, are, at a price that depends on the category and the country. The details are in how much WhatsApp Business API costs.
- AI model: USD 70 per month. A third-party cost for using the model's API. It grows with conversation volume and depends on the model you choose.
- Support hours the chatbot frees up: 80 per month. Half of a full-time schedule of 160 hours.
- Hourly cost of the person answering: USD 5. A monthly cost of USD 800 including payroll charges, divided by 160 hours.
The math, step by step
Step 1: additional orders
- Without the chatbot: 1,000 × 8% = 80 orders.
- With the chatbot: 1,000 × 12% = 120 orders.
- Difference: 40 orders per month.
Going from 8% to 12% is an increase of 4 percentage points. In relative terms it's 50% more (4 out of 8). Both ways are correct; what matters is saying which one you're using, because they describe the same change with very different numbers.
Step 2: additional gross margin, not revenue
- Additional sales: 40 orders × USD 60 = USD 2,400.
- Additional gross margin: USD 2,400 × 40% = USD 960.
The USD 2,400 is not profit: USD 1,440 goes to goods, payment processing and shipping. What the chatbot adds to the bottom line is the margin, USD 960 per month.
Step 3: value of the freed-up time
- 80 hours × USD 5 = USD 400 per month.
This line is money only if it changes a decision: not hiring the next person, no longer paying overtime, or moving those hours to work that brings in revenue. If the salary gets paid anyway and nobody reassigns that time, it's freed-up capacity, not cash savings.
Step 4: monthly costs
- Chatbot (provider's maintenance and hosting): USD 220.
- WhatsApp Business Platform (Meta): USD 40.
- AI model: USD 70.
- Total: USD 330 per month.
Step 5: net monthly benefit
USD 960 + USD 400 − USD 330 = USD 1,030 per month.
Step 6: setup payback
USD 3,000 ÷ USD 1,030 = 2.9 months.
| Item | Per month |
|---|---|
| Additional gross margin | USD 960 |
| Value of freed-up time | USD 400 |
| Chatbot (provider) | −USD 220 |
| WhatsApp Business Platform (Meta) | −USD 40 |
| AI model | −USD 70 |
| Net benefit | USD 1,030 |
With these assumptions, the setup is recovered in 2.9 months. It's the result of a calculation, not a promise: if an assumption changes, the timeframe changes.
The shortcut that inflates these numbers
If step 2 used the additional sales (USD 2,400) instead of the margin (USD 960), the net benefit would come to USD 2,470 per month and payback would drop to 1.2 months. That's the gap between a revenue calculation and a profitability calculation.
What changes the result
The math depends mostly on three things: how much conversion really improves, how much margin each sale leaves, and whether the freed-up hours turn into savings. With every other assumption unchanged:
| Scenario | Additional margin | Freed-up time | Monthly costs | Net benefit | Payback |
|---|---|---|---|---|---|
| Example baseline | USD 960 | USD 400 | USD 330 | USD 1,030 | 2.9 months |
| Half the improvement (8% → 10%) | USD 480 | USD 400 | USD 330 | USD 550 | 5.5 months |
| 20% margin | USD 480 | USD 400 | USD 330 | USD 550 | 5.5 months |
| Not counting the hours | USD 960 | USD 0 | USD 330 | USD 630 | 4.8 months |
| Half the improvement, no hours | USD 480 | USD 0 | USD 330 | USD 150 | 20 months |
| 20% margin, half the improvement, no hours | USD 240 | USD 0 | USD 330 | −USD 90 | Never pays back |
What the table shows:
- Conversion improvement and margin weigh the same. Additional margin is orders × ticket × margin: if either one is cut in half, that line is too. With half the improvement (from 8% to 10%, 20 more orders instead of 40), payback goes from 2.9 to 5.5 months; with a 20% margin, the same.
- The hours carry the case when sales fall short. In the baseline they contribute USD 400 of the USD 1,360 coming in. If they don't turn into cash, the net benefit drops to USD 630.
- The effects compound. With half the improvement and no hours, payback stretches to 20 months. If the margin is also 20%, the chatbot costs more than it returns.
The math also leaves out things that are hard to put into numbers: your team's time to define answers and hand-off rules, reviewing conversations at the start, customers who prefer to talk to a person, and wrong answers that need correcting.
Plug in your numbers
- Sales conversations: count them over a normal week and multiply by 4.3.
- Current conversion: orders over sales conversations in that same period.
- Ticket and margin: from your actual sales, with the margin after variable costs.
- Setup and monthly cost: from the quote you have. Ask what maintenance, hosting and third-party services it includes.
- Meta and AI model fees: ask the provider for an estimate at your volume.
- Hours: how many your team spends answering today and what you'll do with the ones that get freed up.
- Conversion improvement: run at least two scenarios, the one you expect and one with half of it. If the case only works in the optimistic one, it isn't proven yet.
If you want to run this math for your business
Tell us what process you want to solve. In a first conversation we understand the need and propose the next step. If we move forward, the proposal defines scope, stages and timelines, separating what depends on us from what depends on your team and on third parties, and spells out what maintenance, hosting or third-party services it includes.