Car buyers research for weeks before they contact a retailer, and many enquire outside opening hours. This study tested whether UK automotive retail websites are set up to respond when a buyer arrives with a question. We mystery shopped 33 dealer and leasing broker websites across three test windows in a working week and scored each against a five-stage chat maturity model. Twenty-three sites offered no visible chat at any point. Ten did, and five of the 33 showed human or hybrid agent capability. The average benchmark score was 10.0 out of 45. The sample is small and the findings are descriptive, but they point to response, not traffic, as the weak point in the buyer journey.
The full report, with every chart and the scoring model, is available to download below.
1. Background: car buyers research for weeks before they enquire
A car purchase is now a long, self-directed process. The figures cited in our report put the average journey at 87 days from first trigger to purchase. Of buyers who bought in January, 74% had started browsing the previous November or December. And 73% research their next vehicle entirely or mostly online before they set foot in a showroom. The retailer usually becomes aware of the buyer only at the point of enquiry, by which time most of the comparing and shortlisting is done.
Timing matters too. Auto Trader data cited in the report shows that 50% of Auto Trader Deal Builder enquiries are submitted outside typical 9am to 5pm hours, and that after-hours enquiries convert at around twice the rate of standard leads. Auto Trader handled 15 million enquiries in FY25, and a typical search involves 3.9 websites. A buyer who does not get an answer has another option one click away.
Retailers have become good at paying for the visit. This study asks what happens next. When a buyer reaches a retailer's website with a question about price, specification, part-exchange or finance, is there a way to get a response?
2. Method: 33 websites, three test windows, five stages
We mystery shopped 33 UK automotive retail websites, covering automotive retailers and leasing brokers. Each site was tested across three windows in a working week, including business hours and the evening.
The first test was deliberately simple. Is anyone visibly there to help? We did not count a form or a phone number. We looked for a live, real-time way to respond. For sites that passed, we then scored what each conversation could actually do: who answered, how fast, and whether it moved the buyer forward. Sites were placed on a five-stage maturity ladder and given a benchmark score out of 45, covering availability, speed, quality and capture.
| Stage | Description |
|---|---|
| 1. No visible chat | Nothing to engage with |
| 2. Bot-only route | Automated answers only |
| 3. Basic response | Answers questions, but makes no progress |
| 4. Human or hybrid support | A person can join the conversation |
| 5. Sales-ready conversation | Moves the buyer forward |
3. Findings
3.1 Availability: 23 of 33 sites had no visible way to respond
Twenty-three of the 33 sites offered no real-time way to engage a buyer at any point in the test. Only ten had visible chat. Those same ten sites had chat in business hours and in the evening, so coverage did not drop after hours.
This changes how the problem should be described. Out-of-hours enquiry volume is real, but it was not the main failure here. Most sites had no real-time response channel at any time of day.
3.2 Quality: only five sites showed human or hybrid capability
Having chat did not mean being ready to sell. Five of the 33 sites showed human or hybrid agent capability in any window. The other sites with chat were bot-led or limited. They could answer a question, but they could not hold a conversation that progressed a sale. Taken with the 23 sites that never got past the first rung, most of the market sat on the first two stages of the ladder.
3.3 Capture: seven of ten chat sites captured leads well
Of the ten sites with visible chat, seven achieved strong lead capture. We defined this as the conversation doing a sales job: taking contact details, offering a callback or live transfer to a person, or creating a clear next step before the buyer left. The strongest implementations treated chat as a sales channel, not a help function.
From tested to captured
Strong capture meant the conversation did a sales job: details taken, a callback offered or a clear next step created. Implementations varied. The strongest treated chat as a sales channel, not a help function.
The average benchmark score across all 33 sites was 10.0 out of 45. That figure is pulled down by the 23 sites that scored zero because no chat was present.
4. Discussion: the gap is response, not traffic
The results support a narrow claim. In this sample, the typical retailer had paid to bring a buyer to the site and had no means of responding to that buyer in the moment. The measure that matters is not whether a site has chat. It is whether the conversation captures the opportunity.
The strongest experiences in the study shared a pattern. They detected likely buying intent, opened a conversation straight away, qualified the buyer with real questions and answers, recorded contact details and intent, and then created a next step, whether a connection, a booking or a handover. For this to work in practice, the response has to be available when the buyer is active, fast enough to keep them engaged, useful enough to answer real questions, and structured enough to capture the opportunity. For most retailers it also has to be managed, because few dealers or brokers want to staff an evening enquiry team.
Illustrative. What happens after a buyer shows intent depends on whether anyone can respond.
One CommVersion customer shows what this looks like. Select Car Leasing already had website traffic and wanted more of it to become high-intent lease enquiries. After SmartChat was introduced, website leads from existing traffic rose by 28% in two weeks, and the site now generates more than 500 lease enquiries a month from the same traffic. This is a single case, reported by the supplier of the system, and it is not a controlled comparison. We include it as an illustration of the mechanism, not as proof of a general effect.
5. Limitations: what this study cannot show
Readers should weigh the findings against five limits.
The sample is 33 sites. We have not shown that it is representative of the UK market, and small samples give wide margins of error. The study is descriptive. It measures response capability, not conversion, so it cannot say how many sales a missed response costs. It looked for live chat on the website. A retailer that handles enquiries well by phone, WhatsApp or email would score low on this model, and we did not measure those channels. The scoring was carried out by CommVersion, which sells a chat system. Finally, the buyer journey and enquiry-timing figures in the background are third-party figures as cited in the report. We have not independently verified them.
6. Implications: three questions to test your own site
Retailers can test their own position with three questions. When a buyer lands on the site at 8pm with a question, can anyone respond in real time? If someone does respond, can they qualify the buyer and create a next step, or only answer a question? And does a qualified opportunity reach the sales team with contact details and context, not just a chat transcript?
The report includes a way to answer these for your own site. CommVersion will benchmark a dealership or leasing brand against the same five-point model used for the 33 sites, at no cost and with no obligation to proceed. You share one link, we run the benchmark, and you receive a short, plain-English review in a matter of days.
7. Conclusion
In this sample, the next opportunity for UK automotive retailers was not more traffic. It was a better response to the traffic they already had. Twenty-three of 33 sites had no visible way to respond, and few of those that did could take a buyer forward.
.png)
.png)
.png)
