What we learned taking a website to a million impressions

Does content marketing actually work? Yes, but the honest version looks nothing like the overnight-success story, and the gap between the two is where most businesses give up. This is what actually happened when we took one of the sites we run from zero to over a million monthly Google impressions in about five months, with steady publishing and no ad spend. It is our best case on a high-volume topic, not a promise of what any one business will see, and the useful parts are the months that looked like failure and the tactics that flat-out did not work. We are happy to walk any prospective client through the live graphs rather than ask you to take the numbers on trust.

How long did it really take, month by month?

The short answer is about five months to a million monthly impressions, and the shape mattered more than the total. Impressions went from zero in the first month to a few thousand by month two, into the tens of thousands by month three, past two hundred thousand by month four, and over a million a month by month five. Clicks over the same stretch climbed from barely over a hundred a month to several thousand. On a linear chart the first four months look like failure, because the curve is flat then exponential rather than a straight line. That flat stretch is not the strategy failing; it is what a young site looks like while it compounds.

The more instructive site is a second one we run on the same method, in a harder topic. Its impressions started climbing from around month two, but clicks barely moved for the first four months, sitting at a few dozen a month while impressions grew into the thousands. Only around month four did clicks begin to lift. Same pipeline, same quality bar, a completely different speed. Every market compounds at its own pace, which is the single most important expectation to set: the trajectory is reliable, the timetable is not. We cover the full timeline question in how long SEO takes to work for a NZ small business.

Why did it look like failure for the first few months?

It looked like failure because clicks lag impressions by months, and a young domain sits suppressed while search engines decide whether to trust it. For the first stretch, a brand-new site can do everything right on the page, answer-first content, clean structure, consistent business details, and still sit well down the results, because the thing holding it back is accumulated trust and time, not the writing. Higher-stakes subjects, the ones touching money, health or safety, wait the longest.

The trap is watching the wrong number. Visitor count is the last thing to move, so an owner who judges progress by traffic concludes nothing is working, right at the point where the leading indicators say the opposite. The signals that actually move first are impressions climbing week on week, average position improving even before you break into the top ten, and the range of queries you appear for widening. When we watched those three, the payoff was visible months before it showed up in traffic. If they are flat by month four on a steady cadence, that is a genuine signal to change the plan; if they are climbing, the quiet stretch is exactly when to keep going.

What was the biggest hidden lever we found?

The biggest hidden lever was intent match, not ranking or click-through. When we diagnosed why pages with plenty of impressions were earning almost no clicks on one of our sites, almost all of the gap, around 98 percent of it, traced back to a single cause: the page ranked for a query whose intent it did not actually serve. Someone searched for a number, the page served a long essay, they bounced back to the results, and over the following weeks the page quietly slipped down. More than a hundred pages showed a single-pageview rate above 85 percent, a pattern invisible in Search Console and only visible in behaviour-on-page data.

The fix is to match the answer format to what the query wants. A search like a renovation cost or how long a job takes wants a number or a short table at the very top, not a slow build. Serve the exact thing the query asked for in the first line, then expand beneath it. This is why our whole house style leads with the answer, and it is the same craft we cover in how to write content that answer engines quote.

It also taught us to read click-through rate against the type of site, never against a generic average. One young, high-volume site of ours runs a click-through rate under half a percent at an average position of about seven, while a mature, brand-heavy site runs over twelve percent at almost the same position. Same apparent ranking, wildly different clicks, because a young site piles up impressions on broad questions answered above the click, while a mature site earns clicks on branded and buying-intent terms. Benchmark a business against what its own query mix and maturity should produce, not against someone else’s number.

How did we know it was actually working?

We knew it was working from the leading indicators and the behaviour-on-page data, not from the traffic, and learning to read them honestly saved us from two opposite mistakes: quitting too early and celebrating too early. Impressions, average position and query diversity climbed for months before clicks, so those were the real progress report. Behaviour-on-page data, the single-pageview and dwell figures, revealed the pogo-stick decay that Search Console simply does not show, so we watched both.

One episode taught us to read the query breakdown rather than the headline number. On one young site, a single page suddenly picked up a flood of impressions, roughly 85 percent of which turned out to be foreign, wrong-intent searches it had been mismatched to, earning zero clicks. The total made it look like a breakout; the query breakdown showed it was noise. A few weeks later Google quietly corrected the mismatch and those impressions vanished, which read like a traffic collapse but was actually a cleanup. If we had watched only the total, we would have first over-celebrated and then panicked, both for nothing.

The general rule we now follow: read the query breakdown before the headline total, treat impressions and position as the early signal and enquiries as the real result, and never trust a sudden spike until you have checked what it is actually made of. Traffic sits in the middle and misleads the most, because it moves last and hides both the junk and the genuine early progress.

Where does the traffic that actually converts come from?

The traffic that converts is the intent-qualified traffic, and there is much less of it than the impression count suggests, which is exactly the point. On one of our sites, visitors who arrived from an answer engine converted to the main contact action many times more often than the site-wide average, because they had already read a synthesised answer and clicked through deliberately, further along and readier to act.

That connects to a mechanism worth knowing. The answer engines your customers use, ChatGPT and Microsoft Copilot in particular, largely draw their live results from Bing’s index, so being named in one of their answers runs through ranking in Bing. On the best of our sites, almost every page that earned answer-engine traffic was already ranking in Bing’s top few results for its target question, and a single well-structured page drove a large share of a younger site’s answer-engine visits. That is our own data, not a promise, but the lesson is consistent: as ordinary informational clicks get absorbed into answers, the clicks you keep are worth more, not less, so the job shifts from chasing volume to being the cited answer and converting the deliberate visitor with an obvious next step. We go deeper in how ChatGPT search differs from Google and how to get your NZ business cited in ChatGPT, Gemini and Perplexity.

What did we try that did not work?

Plenty, and the failures are more useful than the wins. The honest negatives, all from our own attempts:

• An original-research survey aimed at earning backlinks flopped. We ran a detailed survey with sustained distribution and got fewer than 20 responses in twelve weeks against a target of several hundred, while a single direct, personalised email to people who already knew us hit a nearly 49 percent open rate. The lesson: for a small brand, the tidy original-data-to-journalist-citation thesis is far weaker than assumed, and relationship channels beat anonymous forms every time.
• Podcast appearances and expert-quote outreach produced nothing measurable. Across sixty checks of the questions we wanted to be cited for, our podcast and quote mentions turned up in none of the answers. They scatter your name around, which is not worthless, but they did not earn a single citation on a query that mattered.
• Monitoring online forums for leads failed decisively. In one week of watching for buying signals on community forums, 87 signals narrowed to 27 worth a look, then to 4 worth surfacing, then to zero that became real opportunities. People asking for help on a forum are usually pre-budget, not funded buyers. Actively reaching the right businesses first beat passive monitoring so completely that we rebuilt the whole approach around it.
• Chasing raw traffic numbers misled us more than once. Client-side analytics undercounts badly: on one small site a JavaScript analytics tool logged zero sessions over a week while the edge logs showed roughly 1,300 pageviews, because bots do not run JavaScript and ad-blockers zero out part of the human traffic. Trust server-side and Search Console data, and read the query breakdown rather than the headline total.

None of this means those tactics never work for anyone. It means that in our own hands, on real sites, the simple, direct, focused approach beat the clever one, and the cheapest information you can get is a well-documented failure. It is also the hardest thing for a competitor to copy, because you cannot fake having actually run the experiment, which is exactly why we bother to write the failures down rather than quietly bury them and only report the wins.

What does this mean for a New Zealand small business?

For a New Zealand small business the takeaway is practical: pick one subject you can genuinely own, answer the specific questions your customers actually search, and do not quit during the quiet months. The winnable early ground is the low-competition long-tail, the cost and how-much and how-to-choose questions, not the crowded head terms. In our testing, the searches a young site can realistically win are the specific, buying-adjacent ones like how much a bathroom renovation costs in a given city, not best builder in the main centre, which is walled off by aggregators and entrenched sites.

Depth in one cluster beats breadth across many, because authority accrues inside a topic and does not transfer to unrelated ones, which is the case we make in what topical authority is and how to build it. Get the local fundamentals right alongside the content, the same ones in how a NZ service business gets found on Google. And measure honestly: watch impressions, average position and query diversity, not just visits, because those move first and they are what tell you the compounding has started.

The single most valuable thing five months of real data taught us is that the businesses that win are simply the ones that do not stop during the stretch that feels like nothing is happening. The curve is flat, then it is steep. Almost everyone who quits does so during the flat part, which is exactly when the work is starting to land. If you would rather have that engine run for you, steadily, week in and week out, that is the work YakkaDesk does.