Most people who leave your website without contacting you were not uninterested. They had one question, could not get it answered, and left to ask somebody who would. That is a fixable problem, and it is fixable more cheaply than almost anything else in marketing — but only if you find out what the question actually is instead of guessing.
Why do visitors leave without asking?
Four reasons, in roughly the order they occur:
- The answer is not on the page. Price, delivery area, availability, whether you do the specific thing they need. Most sites answer the general question and not the specific one.
- The only way to ask is a form. A form is a commitment: name, email, phone, message, and then an unknown wait. For a question that would take ten seconds to answer, that is a bad trade and people decline it.
- They do not know when you will reply. A form that says nothing about response time is asking somebody to gamble their afternoon.
- It is after hours. A large share of browsing happens in the evening, when the business is shut and every route to a human is closed.
What does the evidence actually support?
Less than most articles on this subject claim, so here is the part that holds up. The Short Life of Online Sales Leads (Oldroyd, McElheran and Elkington, Harvard Business Review, 2011) found that firms responding to an online lead within an hour were nearly seven times more likely to qualify it than firms responding later, and that a five-minute response beat a thirty-minute one by roughly 21 times on qualification.
⚠ That study is routinely quoted as "100× more likely to convert within five minutes". That figure is about making contact, not converting, and it originates in a 2007 MIT / InsideSales dataset rather than in HBR. We repeated the wrong version on this page until August 2026. Treat any page still quoting it — including a vendor's — as a page that copied rather than checked.
What we will not do is quote a bounce-rate benchmark or a chatbot conversion figure. Bounce-rate averages blend businesses that have nothing in common, and every chatbot conversion statistic in circulation was published by somebody selling chatbots. We sell chatbots. Draw the obvious conclusion about how much weight our number would deserve.
How do I find out what the question actually is?
This is the part worth doing before you buy anything, and it costs nothing:
- Read your last fifty enquiries — email, phone notes, messages — and tally the questions. Not the topics. The literal questions.
- Check how many are answered on your website. In most cases fewer than half are, and the top three are missing entirely.
- Look at your timestamps. Count how many arrived outside 08:00–17:00 SAST.
- Then put the answers on the page. Prices, delivery areas, lead times, what you do not do. Half the problem disappears without any software at all.
Only what remains after step four is a job for automation. Buying a chatbot to answer questions you could simply have published is paying a monthly fee to avoid an afternoon of writing.
What is left for a chatbot to do?
| Visitor's situation | A contact form | A trained chatbot |
|---|---|---|
| One quick question at 21:40 | Fill it in, wait until tomorrow | Answered in about a second |
| Question specific to their situation | Cannot be answered by a static page | Answered from your own material |
| Not ready to give a phone number | Leaves | Asks, gets an answer, gives it later |
| Wants to know when you will reply | No indication | Tells them, honestly |
| Needs a person | Is a person, eventually | Hands over with the conversation attached |
| Speaks Afrikaans or isiZulu | Depends who reads it | All eleven official languages |
What does it cost to fix?
A website chatbot starts at R299 a month plus a R250 once-off setup fee — R549 at checkout, month to month, no contract. Whether that is worth it is arithmetic you can do: after-hours enquiries you never replied to, times your average order value, times an honest conversion rate. If the answer is under R299 a month, do not buy it.
What will not fix it?
A pop-up that appears after four seconds. An exit-intent overlay offering 10% off. A live-chat widget that is only staffed during the hours somebody could have phoned anyway. All three are interruptions rather than answers, and the visitor's problem was never that they had not been interrupted.
They did not leave because they were not interested. They left because they had one question and the only way to ask it was a form with an unknown reply time.
What is the shortest version?
Tally the fifty questions you actually get. Publish the answers. Automate what is left, and only what is left. Then measure after-hours enquiries, because that is the number the fix is supposed to move.
Where we do this
We run why website visitors leave without making contact for businesses across South Africa. Upington, Vereeniging and Bloemfontein each have a page covering what is different about doing it there, down to the suburbs and the languages.