Articles · Repair shop

Repair shop replies in nine minutes, yet website enquiries get lost

12 October 2026

A customer asks what a new screen would cost, hears nothing back and takes the phone to another workshop. The owner of a repair shop has the same explanation for every such case: the staff are glued to their phones and slow to reply. A vendor is offering a bot that replies at once. The owner is already wondering whether one person at the front desk would then do instead of two.

We will test that with a single question about a screen, sent to the shop by three routes. The shop is FixPoint, an invented device repair workshop from our sample reports; the report on its data was written by our AI audit agent. Over three months, 1,828 enquiries reached its CRM from WhatsApp, Instagram and the website. The numbers below come from the report; the three identical requests are our way of showing them.

Request one: WhatsApp, two in the afternoon. A reply in seven minutes

The CRM assigns WhatsApp and Instagram requests to one of the two receptionists automatically. During opening hours half of all enquiries get a reply within 9 minutes, and on WhatsApp and Instagram within 7.2 minutes. Of messenger enquiries, 80.7% get a reply within 15 minutes, and both receptionists are equally quick.

Messages are not all they do, either. They book devices in and hand them back — an average of 23.6 devices booked in per working day — and answer the phone, and messages take each of them about three hours a day. "Glued to their phones" did not survive the check.

Request two: the website form, the same hour. There may be no reply at all

The same question sent through the website lands in the CRM with no one assigned. A request like that can be seen only by applying a special filter. All 494 website requests over the three months arrived this way. For 231 of them — 46.8% — the CRM shows no reply at all, and 215 are still sitting unprocessed.

Of 494 website requests, 231 never got a reply. They have no one assigned and can be seen only through a special filter.

When someone does get round to a website request, the reply goes out after 257.5 minutes. That is during opening hours, and from the same receptionists who answer on WhatsApp in seven minutes. One of them puts it plainly: "sometimes we don't look at them for a week". The people are the same; the difference is one setting that gives the request no owner.

One caveat from the report. Whether website requests are followed up by phone from an ordinary handset, bypassing the CRM, neither the owner nor the receptionist could say. If they are, some of the 231 may have had an answer that the records do not show.

Request three: WhatsApp, nine in the evening. A reply tomorrow morning

In the evening, at night and on Sundays, 760 enquiries came in — 41.6% of the total. Not one evening enquiry was answered the same evening. The reply comes in the morning, typically 837 minutes later, which is about fourteen hours.

What happens next is something nobody knows. Of the enquiries made during opening hours, 42.5% end in an order; of the rest, 20.5%. That is a strong link, but it is not proof: the people who write in the evening may be a different kind of customer, one who would order less often in the daytime too. Only a trial can settle it.

Replies are quick wherever a request has been given to someone

The owner got the observation right and the address wrong. Customers really are kept waiting, just not where the receptionists are replying. Request two belongs to nobody and cannot be seen without the special filter, so there is nobody to chase it. Request three arrived when the shop was closed. The receptionists are quick exactly where something has been assigned to them, and everything else never shows up in their speed.

There is a third hole of the same design: nobody follows up with customers who go quiet. The log does not contain a single case of a member of staff messaging a customer a second time unprompted, and of 1,007 enquiries closed as lost, a reason is recorded for only 148.

At best, one request in three needs the bot

The bot costs 150,000 KZT plus 25,000 KZT a month. The vendor has not said what share of conversations the bot resolves on its own, and it is not clear whether it can be cancelled. In cash terms it will not pay for itself: not one of the firm's costs goes away with it. The receptionists are on a salary, and both stay — the shop is open 54 hours a week, customers walk in without an appointment, and the bot does not answer the phone or book in devices.

Now request by request.

• The first does not need a bot: it gets its reply in seven minutes as it is.
• The second needs an owner. Straight away and free: put one named person in charge of the website and clear the backlog of 215. Then a CRM rule for a one-off 20,000 KZT: an owner for every request, and an alert if there is no reply within 15 minutes. This adds work — up to 21.7 hours a month on a cautious estimate.
• The third needs a trial, and to begin with one that involves no software: a month of evening and Sunday cover, done from a phone, for 60,000 KZT. Compare the share of enquiries that end in an order with today's 20.5%.

If the share rises after the month of cover, software that replies in the evening will have a case it does not have now. The report names the place for it: an assistant on the website that answers from the price list and the frequent questions at any hour and passes the request to a receptionist. For WhatsApp and Instagram, only once the questions it would have to answer have been counted. The decision on the bot is worth coming back to a month after the rule and the trial.

A second bot, one that answers "is my order ready?", costs 40,000 KZT plus 15,000 KZT a month. It is too early to discuss until the firm has counted those questions for at least two weeks.

Diagnosis, any price not on the price list, complaints, warranty claims and refunds stay with people. The report puts no figure on extra orders: the trial will show it.

Three numbers a repair shop can pull today

1. Enquiries with no reply at all, by channel. Export the quarter's enquiries from the CRM, pick out those with no outgoing message and break them down by source. In the example 15.59% of all enquiries went unanswered, and 46.8% of those from the website. The warning sign is one channel failing several times worse than the rest. Open five of its requests and look at the "assigned to" field: if it is empty, the problem is a setting, not the people.
2. The share of enquiries outside opening hours, and how they end. The same export: the time each enquiry was created, and its outcome. In the example 41.6% of enquiries arrive outside opening hours, and 20.5% of them end in an order, against 42.5% in the daytime. The warning sign is the two together: a share above a third and a gap of two to one. That is not yet a reason to buy a bot, but it is a reason for a month of cover.
3. First reply time during opening hours — the median, not the average. The CRM's response-time report, or the same export: take the time within which half of all enquiries got their reply. In the example, 9 minutes. If yours is also a matter of minutes, "slow to reply" can be closed, and the search moves on to the requests that get no reply at all. If it is an hour or more, your case is a different one: speed really is the problem, and the place to start is what the front desk is busy with during those minutes.

All three numbers sit in an ordinary CRM. In our example they showed that requests were being lost not to slow hands but to an empty "assigned to" field and a front desk that closes for the night.

All articles