Should You Fix Your Website Before You Spend on Ads?

Most businesses that come to us have already tried advertising. They ran Meta ads for three months, or they had someone managing Google for a while, and the honest summary is that it did something but nobody could say what. The budget went out, some clicks came back, and the relationship between those two facts stayed a mystery.

Then they do the reasonable thing and assume the ads were the problem. Different agency, different targeting, different creative.

Usually the ads were fine. The place the ads were sending people was the problem.

This is a story about a flower farm in Hamilton, New Zealand, and the six weeks we spent refusing to spend her money.

The situation

Harvest Bloom Flowers is a working flower farm at Peacocke, just south of Hamilton. Rach grows and sells seasonal cut flowers, and she had recently added a retail line of cut-flower seed kits that ship New Zealand wide. The seed kits are the interesting part of the business commercially, because flowers are local and perishable while seeds go in an envelope and travel anywhere.

She had a Squarespace site. It worked, in the sense that a customer who already knew what they wanted could find it and buy it. Across January to July, online sales came to NZ$723.

That figure is not a criticism of Rach. It is what a site looks like when it exists to confirm you are real rather than to sell anything. Plenty of small businesses are in exactly that position and have no idea, because a website that looks fine and a website that works are two different assessments and nobody ever runs the second one.

What we did first, which was nothing you could see

We have a rule at Infinity1 that we will not run paid traffic to a site that has not scored above 75 on an AI visibility audit. It has cost us work. People arrive wanting ads and we tell them the ads come later, and some of them go elsewhere.

The reasoning is simple enough. Advertising is the amplifier. Whatever your site does to a visitor, paid traffic makes it do that faster and to more people. If the site converts at half a percent, buying more visitors buys you more half-percent.

So the first phase was an audit, then a rebuild of the existing Squarespace site rather than a ground-up replacement. That distinction mattered. Rach did not need a new website. She needed the one she had to do its job, and rebuilding from scratch would have cost her months and money she could put into stock instead. We upgraded what was there, with a full rebuild on the table later once the revenue justified it.

The work itself ran across the last two weeks of July. Structure, product architecture, the path from landing on the site to completing a purchase, and the technical groundwork that lets AI search engines read and cite a business properly.

That last piece deserves unpacking, because it is where most small-business sites are quietly invisible.

The bit almost nobody has done

AEO stands for Answer Engine Optimisation. It is the work of making your business legible to the systems that now answer questions directly rather than handing over a list of links: Google's AI Overviews, ChatGPT, Gemini, Perplexity, Copilot.

When someone asks one of those tools where to buy cut-flower seeds in New Zealand, the tool does not run a search and read the results the way a person would. It assembles an answer from sources it can parse and trust. Structured data, clear entity definitions, content written in the shape of the questions people actually ask, and crawler access that lets those systems read the site in the first place.

Most small business websites fail at the first hurdle. They have no structured data at all, which means an AI engine has to guess what the business is, where it operates and what it sells. Sometimes it guesses right. Often it does not mention you.

By the third and fourth week of the campaign, Harvest Bloom was scoring 100 out of 100 on our audit. That is not a vanity number. It is the difference between being a candidate for an answer and being invisible to the fastest-growing way people find things to buy.

Then we turned the ads on

The campaign went live on 1 August.

Here is what August produced, taken off the Squarespace dashboard on 27 August:

One month delivered 84% of the year's online revenue to that point.

The conversion rate is the number worth sitting with. At 2.44%, roughly one in forty visitors bought something. That is a healthy number for a small ecommerce store, and it is a number that only exists because the site was ready before the traffic arrived. The same spend against the old site would have bought the same visitors and converted a fraction of them.

The honest caveats

The base was small. Going from NZ$723 to NZ$3,828.97 is a big multiple, and big multiples off small numbers are easy. The absolute figure matters more than the percentage, and the absolute figure is a good month for a small farm rather than a life-changing one.

It is one month. A single strong month is a signal, not a trend. The question that actually matters is whether September and October hold, and we will know that shortly.

Seasonality is real. August is when New Zealand gardeners buy seed to plant out after the last frost. That window was going to lift sales regardless of what we did. Anyone telling you otherwise is selling something.

What seasonality does not explain is the conversion rate, or why the same business sitting through previous seasonal windows on the old site did NZ$723 across seven months. The season brought people. The rebuild is why they bought.

The order of operations, in five steps

If you take one thing from this, take the sequence rather than the numbers.

1. Score the site before you budget anything. Run an audit and get an actual number. Not an opinion about whether the site looks good. A measurement of whether it is structured to be found and built to convert.

2. Fix structure before decoration. Almost nobody needs a redesign. They need the path from arrival to purchase shortened, the product architecture to make sense, and the pages that answer buying questions to exist at all.

3. Make it readable by machines. Structured data, entity definitions, content in question form, crawler access opened. This is the cheapest work in the whole sequence and the least done.

4. Then turn on paid. Once the site converts, every dollar of ad spend works harder, and it keeps working harder for as long as the site stays fixed.

5. Measure from the first day. Conversion tracking and analytics wired in before the first ad runs, so you can answer the question the previous agency could not.

The sequence compounds. The site work keeps paying after the ad budget stops, which is the opposite of how paid traffic behaves on its own.

What this looks like for an Australian business

Harvest Bloom is in Hamilton, but nothing about the pattern is specific to New Zealand or to flowers.

A Gold Coast homewares maker selling through Instagram with a Shopify store they set up in a weekend is in the same position. So is a Brisbane trades business paying for Google Ads that land on a homepage with no structured data, no service pages and no answers to the questions people ask before they call anyone.

In both cases the instinct is to spend more on traffic. In both cases the faster return is in the thing the traffic arrives at.

At Infinity1 we do this work in that order for every client, which occasionally means telling someone their ad budget should sit still for a few weeks. It is a harder conversation at the start and a much easier one at the end.

Let's revisit what we've covered

1. A working flower farm went from NZ$723 across seven months to NZ$3,828.97 in one, after the foundations were fixed.
2. No advertising ran until the site was ready, which took the last two weeks of July.
3. The AI visibility work took the site to 100/100 by the third week of the campaign.
4. August converted at 2.44% across 98 orders, which is the number that shows the site was doing the work.
5. Seasonality helped. The sequencing is why the season converted.

If you are about to put money into advertising, the useful first question is not how much to spend. It is whether the place you are sending people is ready to receive them.

We can tell you that in about a week, with a number attached. If you want to know where your site currently sits, book a strategy call and we will map out what your business actually needs before anyone spends anything.