USE CASE: How to Use GenAI to Build an Ideal Customer Profile That Isn’t a Demographic in a Car Dealership
A real-world GenAI marketing use case: how a multi-brand car dealership replaced “anyone who needs a car” with four ideal-customer profiles — new-car, used, service and fleet — defined by what each customer needs and how they buy, not by age and postcode.
An ideal customer profile isn’t a demographic; it’s a description of need and behaviour, and confusing the two is why most targeting misses. This is what building one with GenAI looks like in practice: how a dealership turned “anyone who needs a car” into four distinct customers it could actually market to.
The Context: a Customer Base Defined as “Everyone”
A multi-brand automotive dealership and service group whose answer to “who is your customer?” was, in effect, “anyone who needs a car”. Marketing was built for that imaginary average buyer (broad messages, broad offers, broad spend) on the reasonable-sounding logic that the wider the net, the bigger the catch.
The Challenge: “Everyone” is No One, and a Demographic Won’t Save You
“Anyone who needs a car” feels generous; it’s actually the absence of a customer strategy. And the usual correction makes it worse: reaching for demographics. “Our customer is 30 to 60, mostly male, within twenty miles” describes who walks in, not why, and why is the only thing you can sell to. The real problem was that the group served four fundamentally different customers and treated them as one: the new-car buyer chasing aspiration and finance; the used-car buyer weighing value and risk; the service customer who wants reliability and convenience on repeat; and the fleet buyer, business running a rational procurement decision about uptime and total cost. They share almost nothing in what they need, how they decide, or what they’re worth. A single demographic blob can’t tell them apart, so the marketing spoke to an average customer who doesn’t exist.
The same person is four customers: The same 45-year-old man can be four different customers – the new-car aspirant, the used-car bargain-hunter, the loyal service client, the fleet manager doing his job. A demographic can’t tell them apart; only what they came to do can. An ICP isn’t who your customer is; it’s what they’re trying to get done.
The GenAI Workflow: 4 Profiles, Built on Behaviour not Birthdays
Instead of one fuzzy “car buyer”, the group built four distinct ideal-customer profiles, and deliberately built them on the right axes. Using GenAI, each was defined not by age and postcode but by the job they were hiring the dealership to do, the trigger that brought them in, how they actually decided and who else was involved, what proof they needed to trust, and what they were worth over time. The new-car, used-car, service and fleet profiles came out looking nothing alike – different motivations, different timelines, different proof, and in the fleet’s case an entirely different, B2B buying process. GenAI drafted the structure and the first articulation; the team corrected it against the customers they actually see, and cut every line that had quietly slipped back into demographics.
You are a customer strategist for a multi-brand car dealership and service group. We serve four distinct customers: new-car buyers, used-car buyers, service customers, and fleet/business buyers.
For EACH, build an ideal-customer profile defined by BEHAVIOUR and NEED, not demographics: the job they’re hiring us to do, the trigger that starts their search, how they decide and who else is involved, what proof they need to trust us, their rough value over time, and where they look for us.
Do NOT define them by age, gender or income unless it is a genuine buying factor rather than a label. Keep the fleet profile clearly B2B. If two profiles start to look alike, sharpen the difference. Mark anything you’re assuming about our customers, rather than drawing from what I tell you, as VERIFY WITH OUR DATA.
The caveat that decides whether this works. GenAI has a demographic reflex: ask it for an ideal-customer profile and it reaches straight for “males aged 35–55, household income…”, because that is the pattern it has seen ten thousand times and it sounds like an answer. It isn’t; it’s a description. It also doesn’t know your customers, and will assert buying triggers, decision criteria and value with total confidence on no evidence at all. And it tends to blur 4 profiles back toward one average, because averages are easier to write. The controllable variables: build from your own sales data and the people who actually serve these customers, hold every profile to behaviour over demographics, and keep the fleet buyer firmly in its own B2B lane. GenAI drafts the profiles; your customers define them.
The Result: 4 Real Customers Instead of One Average
One average customer who didn’t exist became 4 real ones who did. The new-car, used-car, service and fleet profiles each got marketing built around what that customer actually needed and how they actually bought: aspiration and finance for one, value and reassurance for another, reliability for the third, and a rational B2B case for the fleet. Budget could be aimed instead of sprayed, because the group finally knew who each pound was talking to. And the fleet line – a different business hiding inside the dealership – got the distinct B2B treatment it had never had while it was lumped in with walk-in buyers. No invented figures here: the change is that “anyone who needs a car” became 4 people the group could actually recognise and serve.
Recommended KPIs to Follow
An ICP is a targeting decision, so it’s judged on how well your marketing now matches a real customer rather than an average. These are the metrics to watch, where the industry sits, and the direction this work should push them. The point is the direction of travel, not a promised number.
ICP Fit Rate
The share of your leads and spend that actually match one of the 4 defined profiles, rather than the undefined “everyone”. It’s the simplest read on whether the profiles are governing your targeting or just sitting in a document.
Benchmark: No universal figure, an internal metric; a common rule of thumb is that if under ~60% of your pipeline matches a defined profile, the targeting still needs work (Sybill). Set your own baseline.
Conversion / Win Rate by Profile
How each profile converts once marketing speaks to its actual need, and how much better than the old untargeted approach. Track the four separately; the fleet (B2B) line especially should be measured on its own terms.
Benchmark: Much-cited B2B research puts win rates roughly 68% higher for organisations with a well-defined ICP, and ICP-aligned leads often convert 2-3× non-ICP ones – strongest for the fleet profile, with the consumer profiles gaining through sharper targeting (SiriusDecisions / Forrester; Sybill).
Cost per Acquisition by Profile
What it costs to win each type of customer once budget is aimed rather than sprayed. Splitting CPA by profile reveals which customers are cheap to win and worth chasing, and which were quietly draining a blended budget.
Benchmark: No clean public figure, an internal metric; baseline CPA under the old “everyone” approach and track it per profile as targeting sharpens.
The 68% figure is a much-repeated B2B benchmark, useful as direction, not gospel. What matters is your own before-and-after: untargeted spend against four aimed ones. Track your own trend.
Why this Transfers
Almost every “our customer is everyone” really means four or five different customers no one has bothered to separate. The transferable move is to define them by what they’re trying to get done and how they decide – not by age, gender or postcode – because demographics tell you who is in the room, and only need and behaviour tell you why they came, and what to say when they do.
Recommended Articles:
Diana Socaciu