Please note: this is an illustrative example. The business is fictional and the figures are made up, though realistic, to show how a campaign is planned and how to read the numbers. They are not results from a real client campaign. We keep real client results confidential.
It’s one thing to read how ChatGPT ads work. It’s another to see how a campaign comes together and what to do with the numbers once it’s running. This walkthrough follows a fictional local business through its first four weeks on OpenAI Ads Manager.
The business
Greenline Gardens (fictional) is a landscaping business in Melbourne’s south-east, serving Pakenham, Officer, Berwick and Narre Warren. Its average job is worth around A$6,000, and most customers spend weeks researching before they ask for a quote. That long research phase is exactly where ChatGPT ads can help.
Goal: get quote requests from homeowners planning a backyard makeover.
Test budget: A$30 a day for four weeks, A$840 in total.
Step 1: Check eligibility and verify the business
Landscaping is a local service, which is a supported category. The business verified its details in Ads Manager, making sure its business name, ABN and website domain all matched. Verification can take up to 7 business days, so this was done two weeks before the planned launch.
Step 2: Prepare the landing page and tracking
- A dedicated “Backyard makeovers” service page listing the suburbs served, recent project photos, real reviews and a short quote form
- The OpenAI pixel installed, with the quote form submission set up as a conversion
- UTM parameters on the ad link, so the traffic appears separately in Google Analytics
Step 3: Build the campaign
Campaign: optimise for clicks, A$30 daily budget.
Ad group: Backyard makeovers. Context hints:
- “Planning a backyard makeover or new garden in Melbourne’s south-east”
- “How much landscaping, a new lawn or a retaining wall costs in Victoria”
- “Low-maintenance garden ideas for a new estate home in Casey or Cardinia”
- “Preparing a backyard before selling a house”
Two ads:
- A: “Backyard makeovers in Pakenham and Berwick” / “Local landscapers. Free on-site quote and a fixed price before we start.”
- B: “Low-maintenance gardens for new homes” / “Designed and built by a local team in Melbourne’s south-east. Get a free quote.”
Both ads used a real photo of a finished backyard and linked to the same service page.
Step 4: The results, week by week
| Week | Spend | Impressions | Clicks | CTR | Avg. CPC | Quote requests | Cost per request |
|---|---|---|---|---|---|---|---|
| 1 | A$210 | 5,600 | 48 | 0.86% | A$4.38 | 1 | A$210 |
| 2 | A$210 | 5,300 | 55 | 1.04% | A$3.82 | 2 | A$105 |
| 3 | A$210 | 5,100 | 62 | 1.22% | A$3.39 | 3 | A$70 |
| 4 | A$210 | 5,200 | 68 | 1.31% | A$3.09 | 3 | A$70 |
| Total | A$840 | 21,200 | 233 | 1.10% | A$3.61 | 9 | A$93 |
Illustrative figures only.
Step 5: What changed each week, and why
Week 1: no changes. Spend reporting in Ads Manager can lag by 7 to 8 hours, and the system needs time to learn which conversations suit the ad. The main job this week was confirming the pixel recorded the first quote request.
Week 2: ad B, the low-maintenance angle, had a clearly higher click-through rate than ad A. The team wrote a third ad using the same angle and paused nothing yet.
Week 3: the landing page was updated so its headline matched the low-maintenance message in the ad. Keeping the ad and the page consistent usually lifts both relevance and conversions. The context hint about preparing a house for sale was reworded to be more specific.
Week 4: the weakest ad was paused, and budget stayed the same to keep the test fair.
Step 6: Reading the numbers
- Click-through rate rose from 0.86% to 1.31%. The ads and context hints were getting more relevant as they were refined.
- Cost per click fell from A$4.38 to A$3.09. In a relevance-weighted auction, more relevant ads usually pay less.
- Cost per quote request averaged A$93. With an average job worth A$6,000, winning even one or two of the nine requests would more than cover the test.
- Nine requests is a small sample. It’s enough to justify continuing, not enough to declare a winner.
The key comparison is cost per quote request against the business’s other channels, such as Google Ads, not cost per click on its own.
What we’d do next
- Keep running for another four to eight weeks at the same budget to build a bigger sample
- Track which quote requests become paid jobs, not just enquiries
- Test a second ad group for a different service, such as retaining walls
- Compare the true cost per job against Google Ads before moving budget between channels
Lessons that apply to any business
- Set up tracking before you spend a dollar.
- Write context hints as real situations your customers describe.
- Give each change a week to show its effect.
- Judge success by cost per enquiry or sale, and by what a customer is worth to you.
For the full setup process, read how to advertise on ChatGPT, and see what ChatGPT ads cost to plan your own budget.
Want this done for your business?
Madexify runs ChatGPT ad campaigns exactly this way: verification, tracking, context hints, creative and weekly optimisation, for a flat A$200 a month in management with ad spend billed directly to your own OpenAI account. See our ChatGPT ads management service or get in touch.
This article is part of our Online Advertising Guide. Browse all our guides on the guides page.
