What happens when you fix the website before you spend on ads

Rachel grows flowers. For the first seven months of the year her website was doing very little. The product was good, the photography was good, and almost nobody was buying, because almost nobody was arriving.

We rebuilt the site across the last two weeks of July and switched the campaign on 1 August.

August closed 1,083% up on July. One hundred and fifteen orders against eleven the month before. Three hundred and seventy units out the door, up 849%. Year on year the business sits 2,676% up.

Here is the number that actually proves something. Traffic went up 601%. Revenue went up 1,083%. Revenue grew at nearly twice the rate of the traffic, which means the site was converting better at the same time as it was getting busier. Her conversion rate moved from 1.47% to 2.41%.

That is the opposite of what normally happens. Turn ads on and conversion usually falls, because you are reaching further out and buying colder visitors. More traffic, worse traffic. Everyone accepts that trade.

Roughly 82% of everything the business has earned this year landed in that one month, on 63% of the year's traffic. The extra nineteen points is the site doing its job.

What we actually did

The rebuild came first and the ads came second, and that order is the whole point.

We rebuilt the site around the way people actually buy flowers, which is quickly, on a phone, usually for someone else. Then we took the AI search visibility score from 87 to 100, so that when someone asks ChatGPT or Google's AI where to buy flowers locally, the site is structured to be the thing that gets quoted. Schema markup, a real answer library, question-format headings, the lot.

Only then did the Meta campaign go live. By that stage the site could convert what the ads sent it.

Why it worked

Most people run this backwards. They spend on ads first, send traffic to a site that cannot convert, and conclude that ads do not work for their business.

The ad budget is the shot at goal. The website is who is taking the kick. You would not hand the kicking tee to a middle forward when you have a fullback landing 85%.

Stack the odds, then kick.

Warner Pool Care sell pool and spa equipment out of two shops in North Brisbane. Two physical locations, a real product range, and a website that search engines could read but AI assistants could not properly quote.

The first audit scored it 42 out of 100. Not broken, just invisible to the half of search that now runs through AI.

Today it sits at 100, and the Google Shopping campaign returns fourteen dollars for every dollar spent, on a budget of thirty dollars a day.

What we actually did

Sitewide schema so every product, location and opening hour is machine-readable. A blog answering the questions people actually type, published on a schedule, bylined by the owner rather than a faceless team account. Both branches structured properly so the Brendale shop and the Albany Creek shop each read as a real place rather than one business with two addresses stapled on.

Then Google Shopping, kept deliberately small, because at 14:1 the constraint is stock and staff rather than budget.

Why it worked

Thirty dollars a day is not a marketing budget, it is a rounding error. It returns 14:1 because the site behind it was fixed first.

That is the same lesson as the flower farm, in a completely different industry.