How many inquiries per month justify a WhatsApp chatbot (and when they don't)

There is no universal number of inquiries, but there is a threshold you can calculate in two minutes: an AI WhatsApp chatbot starts paying for itself when the margin you recover per month exceeds the USD 250 it costs to sustain during the first year. That happens at 75 to 100 monthly inquiries if your margin per new customer is USD 180 or more (dentistry, real estate, law firm), at around 230 if the margin is close to USD 50, and only above 350 if your margin is USD 30 or less. Below that threshold a chatbot is not the right move: a simple automation that costs a tenth as much is.

The question "how many inquiries do I need?" has no answer because half the data is missing. Two hundred inquiries at a real estate agency and two hundred at a takeaway food shop are two completely different businesses facing the same investment.

The real threshold is not measured in inquiries: it is measured in recoverable margin

A chatbot gives you money back through two channels, and it is worth separating them because one is much smaller than people assume:

1. Inquiries that today get lost outside business hours and the bot rescues. This is the line that decides the case. It is new margin, not savings.
2. Hours of your team that stop going into answering the same thing. It is real, but it is smaller and it comes with a condition I explain further down.

There is a third channel that always gets quoted and that I am not going to put into the math here: the conversion lift from replying in one minute instead of three hours. It exists and it is usually significant, but you cannot measure it before having the bot running, so putting it into the purchase calculation means building a number that returns whatever you want it to return. If the chatbot does not work out with the first two lines, do not buy it hoping for the third.

The full math: USD 250 per month against two lines of return

On the cost side, the lowest tier of an implementation with support, in the first year and prorating the setup:

• USD 499 setup divided over 12 months: USD 42 per month.
• Platform, hosting, monitoring and support: USD 199 per month.
• Meta Cloud API at 150 conversations: service conversations started by the customer are not charged within the 24-hour window, and around 70 utility ones at USD 0.028 come to USD 2 per month.
• AI with GPT-5 mini at about USD 0.01 per conversation: USD 2 per month.

Total: around USD 245 per month. The uncomfortable detail in that breakdown is that Meta and the AI are noise: 98% of the cost is fixed. That is why at low volume the cost per inquiry handled is brutal (USD 1.63 at 150 inquiries) and at medium volume it collapses (USD 0.42 at 600). A chatbot does not get cheaper because you negotiate better, it gets cheaper because you use it more.

On the return side, the formula with the five assumptions in plain sight:

return = (inquiries × 0.35 outside business hours × 0.25 truly lost × 0.20 close rate × margin) + (inquiries × 0.70 repetitive × 3.5 minutes ÷ 60 × USD 5 per hour)

Those percentages are typical starting-point ranges, not audited results from any specific company. Swap them for yours as soon as you have them. The two that move the result most are the margin and the 25% of "truly lost": the other 75% of the after-hours inquiries writes again in the morning or calls you, and counting them all as a loss is the most common trap in this kind of math.

Scenario A: 150 inquiries a month at a food business (it does not pay for itself)

Average ticket USD 20, gross margin 35%, that is USD 7 per new customer. It is an illustrative scenario, not a client.

• Outside business hours: 52 inquiries. Truly lost: 13. Close rate of 20%: 2.6 sales × USD 7 = USD 18 per month.
• Repetitive: 105 inquiries × 3.5 minutes = 6.1 hours × USD 5 = USD 31 per month.
• Total return: USD 49 against a cost of USD 245.

That leaves USD 196 per month missing, USD 2,350 over the year. And there is no reasonable volume improvement that turns that around: to break even with a margin of USD 7 you would need around 780 monthly inquiries, five times the current volume. Here the chatbot is not an investment with a slow return, it is a subscription that never pays for itself.

Scenario B: the same 150 inquiries at a dental practice

Same volume, different business. Average initial treatment USD 300, margin 60%: USD 180 per new patient.

• The same 2.6 inquiries rescued per month × USD 180 = USD 468.
• The same 6.1 hours freed up = USD 31.
• Total return: USD 499 against USD 245 of cost.

Without prorating, month 1 closes in the red by USD 203 (USD 499 setup plus USD 203 of recurring cost against USD 499 of return) and month 2 is already positive by USD 93 cumulative. Break-even in month 2, with exactly the same inquiry volume as in the previous scenario. The variable that changed was not the WhatsApp: it was how much a patient is worth.

Scenario C: 600 inquiries a month at an e-commerce store, where everything depends on one number

E-commerce on Tiendanube (the dominant e-commerce platform for small merchants in Latin America) with an average ticket of USD 45 and a 33% margin: USD 15 per sale. Four times the volume of the previous cases.

• Outside business hours: 210. Truly lost: 52. Close rate 20%: 10.5 sales × USD 15 = USD 157.
• Repetitive: 420 inquiries × 3.5 minutes = 24.5 hours × USD 5 = USD 122.
• Total return: USD 279 against USD 255 of cost. USD 24 per month left over.

With four times the volume, the case barely breaks even. What turns it around is not selling more: it is which margin you put into the math. If that customer you rescued buys again 2.5 times a year, the relevant margin is not USD 15 but around USD 37, and those same 10.5 sales come to be worth USD 389: the return jumps to USD 511 and the case stops being doubtful. If your e-commerce sells something people buy once every three years, do not do that math.

Hours saved are only money if they change a decision

In all three scenarios the hours line is the smallest one, and on top of that it is the easiest one to inflate. Six hours a month you take off the receptionist's plate do not show up in your cash if the receptionist earns the same salary next month.

Those hours turn into real money only in three situations: if you were about to hire someone and now you are not, if you were paying overtime or a weekend shift to cover messages, or if that person shifts those hours into something that bills (calling pending quotes, collecting overdue accounts). If none of the three applies, cross out the hours line from the formula and decide with the lost-leads line alone. In scenario C that pruning is lethal: without hours you are left with USD 157 against USD 255, and the case falls apart.

How many inquiries you need based on your margin per new customer

Monthly threshold to cover USD 250, using the formula above. The first column applies if the hours saved translate into cash; the second, if they do not.

• Margin USD 15 (low-ticket retail or food service): 536 inquiries with hours, 950 without hours.
• Margin USD 30 (mid-ticket e-commerce, first purchase): 343 / 476.
• Margin USD 50 (gym, hair salon, recurring service): 232 / 286.
• Margin USD 100 (accounting firm, repair shop, event catering): 128 / 143.
• Margin USD 180 (dentistry, physiotherapy, professional service): 75 / 79.
• Margin USD 300 or more (real estate, law firm, machinery): 46 / 48.

Read it backwards and it is more useful: if you are in the USD 15 row and you get 200 inquiries, you already know the answer without asking for a quote.

If you are below the threshold, this is what actually works

Below the threshold the problem still exists: you lose inquiries at night. What does not work is solving it with an AI chatbot. The cheap version captures a good part of the value:

• Stay on the WhatsApp Business app, do not migrate to Cloud API. Migrating is what triggers the cost, and on top of that you lose native features: the away message, the greeting message and the quick replies exist in the app and do not exist in the API, there you have to program them.
• Away message with your real hours and one action, not a "we will get back to you shortly". Put in the hours when you do reply and a scheduling link (Google Calendar appointment schedule or Calendly, both with a free plan) so whoever cannot wait books on their own.
• Quick replies for the five questions that keep repeating: price, hours, address, payment methods, availability. You configure them once and they cut more time than any badly trained bot.
• An n8n workflow on a VPS of USD 5 to 10 per month that alerts you by email or Telegram when a message comes in outside business hours and logs it in a spreadsheet. That also gives you the data you do not have today to recalculate the threshold in three months.

It is an afternoon of work, with no platform subscription. When volume or margin cross the threshold, the chatbot will still be available.

There are businesses with plenty of volume where we still do not deploy a chatbot

The threshold is a necessary condition, not a sufficient one. Four cases where the math works and we still say no:

• B2B with corporate clients who write between 9 and 18. If only 5% of your inquiries arrive outside business hours instead of 35%, the main return line disappears and what is left is an expensive bot answering what your team already answers in five minutes.
• High-ticket consultative selling where the first contact defines the relationship (machinery, construction, custom software). There, a bot filtering out someone who was going to sign is a bigger risk than the lead lost at 11 PM.
• Businesses with no system behind them. If stock lives in a notebook or appointments live in one person's head, the bot will confidently answer false things. First you tidy up the data source, then you connect the bot.
• Inquiries whose substance is regulated advice (accounting, legal, diagnosis). The bot can take data, qualify and schedule; it cannot answer the substance, and that limitation has to be accepted before signing, not after.

How to get your own numbers this week without installing anything

Three data points, seven days of measuring:

1. Real inquiries. Count new conversations, not messages: one inquiry is the 6 to 9 back-and-forth exchanges with the same person. Take a normal week (not a holiday week or the peak-season one) and multiply by 4.3.
2. Which ones arrived outside your business hours and, of those, which ones never got an answer within 12 hours or never bought. That subset, not the total night volume, is your loss number.
3. Margin per new customer: average price of the first order or treatment times your gross margin. Do not use revenue. And hourly cost: gross monthly salary of whoever answers divided by 170.

With those three numbers, the formula in this post gives you the answer without anyone selling you anything.

What we do with these numbers in the 15-minute audit

Bring the three data points above and on the call we do the math together. If the threshold does not work out, we tell you so and leave the simple automation version set up for you; we do not automate everything, and by now you know why. If it does work out, you will know in which month you recover the investment before signing anything.

Write to us at info@striqtech.com or book the assessment at striqtech.com/#servicios.