Case study, a real account
An air-conditioning sales and installation business: the first three weeks of Google Ads, 140 clicks at 55 cents and 13 counted conversions
A sales and installation business in a regional Bulgarian city asked for one search campaign, capped at a fixed monthly figure in euros and paid to Google from its own card. The campaign went live on 15 September 2026 after a pre-launch audit. These are its first twenty-one days, read from the Google Ads API on 6 October, with the client unnamed, the currency kept, and the caveat beside every figure that needs one.
Written by Daniel Stoychev, who built and runs the account. Figures from the Google Ads API for 15 September to 5 October 2026; the owner sees the same numbers in the account.
The account in one table
Twenty-one days, every line from the API
| Line | Value | What it means here |
|---|---|---|
| Impressions | 623 | Search only; phrase and exact match |
| Clicks | 140 | Six ad groups, one shared budget |
| Click-through rate | 22.5% | High because the local service searches had almost no competing ads when read from a local IP before the build |
| Spend | 76.78 (euros) | The client pays Google directly |
| Average cost per click | 0.55 | Under the 0.45 bid cap on some days, over it on others; the cap is a target for Maximize Clicks, not a ceiling |
| Conversions, counted | 13 | Website call taps, forms and messenger taps that bidding is allowed to use; fractions are consent-mode modelling |
| All conversions | 34 | Adds the taps on the ad's own call button and the secondary actions, which bidding does not count |
| Search impression share | 68% | Lost to budget 5.9%, lost to rank 26.1% |
The gap between 13 and 34 is the most important line on the page, and it is explained below. The click-through rate is the least important, because it says more about the empty auctions than about the ads.
By ad group
Where the clicks and the enquiries came from
| Ad group | Impressions | Clicks | Spend | Counted conversions | All conversions |
|---|---|---|---|---|---|
| Installation | 211 | 44 | 25.90 | 4 | 11 |
| Maintenance and cleaning | 186 | 43 | 24.66 | 5.5 | 7.3 |
| Air conditioners, sales | 145 | 38 | 18.33 | 3 | 8 |
| Repair and service | 52 | 8 | 4.82 | 0 | 2 |
| Brand | 18 | 6 | 2.47 | 0.5 | 5.7 |
| Refrigerant top-up | 11 | 1 | 0.60 | 0 | 0 |
Installation and maintenance, the two services with the least competition, took two thirds of the budget and produced most of the counted enquiries. The sales group, where every regional retailer bids, bought clicks at the lowest cost of the three and produced the fewest counted enquiries per click, which is what the pre-build research predicted. The brand group exists because two competitors were already bidding on the business's name; six clicks and five all-conversions for 2.47 is the cheapest line in the account.
What was built
One campaign, six groups, and the decisions behind them
- One campaign with one shared daily budget, because splitting a few euros a day across campaigns leaves nothing that can learn.
- Six ad groups by service: installation, maintenance and cleaning, sales, repair and service, refrigerant, brand. Sixty-four keywords in phrase and exact match only.
- Shared negative lists from day one, including car air-conditioning, which sixteen months of Search Console showed to be one in eight of the local searches containing the word, and the sales terms the business does not want.
- Maximize Clicks with a cost-per-click cap for the first weeks, to be moved to Maximize Conversions once fifteen verified conversions exist. At 13 counted after three weeks, that move has not been made.
- Three conversion actions through Tag Manager behind a consent banner: a tap on the site's phone number, the enquiry form, and a messenger tap as a secondary action.
- A Monday to Friday schedule, because the business answers the phone on weekdays.
The cap, and what it taught us
Why the daily budget changed on 1 October
The owner's instruction was a fixed monthly maximum. The daily budget was set to that figure divided by 25 weekdays, which is the arithmetic most people do, and in the first full week the account spent 8.13 per weekday against a budget of 6: Google averages the daily budget across every calendar day, including the weekend the schedule switches off, and moves the weekend's money onto the weekdays. On 1 October the budget was reset to the monthly cap divided by 31, with a small margin, which gives Google's own 30.4-times ceiling room under the promise. The campaign kept its schedule and the impression share lost to budget stayed under six percent. The arithmetic is set out in the guide on capping monthly spend, linked below, because it is the same on every account.
What 13 conversions means, and what 34 means
Three kinds of call, two of which bidding can see
The counted column holds the website phone taps, the forms and the messenger taps, with fractions where the consent banner was declined and Google modelled the rest. The all-conversions column adds the taps on the ad's own call button, which Google reports as clicks to call. That action sat in the account's contact goal marked primary, and the counted column still did not include it; the API refused to make it biddable. So the honest reading is thirteen enquiries that bidding can learn from and about twenty more that it cannot, and the report to the owner carries both lines with that sentence beside them.
No revenue is claimed. The owner has not yet told us which calls became installations, and until that is known the account is judged on cost per counted enquiry, which over the three weeks was under six euros.
What changed on 24 September
The one structural change inside the window
All six ads launched with an ad strength of poor, for the reason Google gives: seven to ten headlines of a possible fifteen, and four sitelinks where it asks for six. On 24 September every ad was taken to fourteen or fifteen headlines, with the pinned headlines left untouched, and the sitelinks to six; every ad was approved the same day. The share lost to rank, 26 percent over the window, is the lever that change was meant to move, and it will be read against the next three weeks rather than claimed now.
Caveats
What this page does not say
- Twenty-one days is a reading, not a result. The account had no advertising before, so there is no before to compare with, only the first after.
- The figures are in euros and the auctions are Bulgarian; the click costs and the click-through rate do not transfer to a UK account.
- The counted conversions include modelled fractions from consent mode, which Google documents and which no account can avoid where a banner is declined.
- Two settings changed inside the window, the budget on 1 October and the ads on 24 September, so no single change is credited with the numbers above.
Where this leads
The pages that do this work
Tell us what you spend. We reply with a fixed fee and the first three changes.
Your monthly spend, what a lead is worth to you and the address of the page your ads land on. Daniel Stoychev replies within one working day with the tier, the fee, what he would change first, and whether your sector and area are still free.