Half the leads came from the channels nobody else was running

The challenge

A New Zealand hiking and adventure tour operator selling multi-day guided trips, almost entirely to people on the other side of the world. Their customers were in the United States, Canada, the United Kingdom and Australia, researching a trip months ahead and booking something expensive and considered. Nothing sold itself off a single click.

That makes the marketing harder than it looks. The sale did not happen online. Someone requested a brochure or picked up the phone, and the business closed it from there. There was no revenue figure on a screen to steer by. The job was to produce enough good enquiries, cheaply enough, from four markets at once.

What we did

We started with a full digital review of the site before touching an ad account. That review fed a site rebuild, which we had a lot of input into. Then we took over Google Ads in January 2017 with three campaigns running.

Over the next three years we built that into thirty. Search across each market, split into non-brand, brand, competitors and dynamic ads. Display for remarketing and for prospecting against audiences that looked like people already converting. YouTube. Then Discovery campaigns, which we were testing while most of the industry was still ignoring them.

The measurement had to carry all of it. With no online transaction to count, every brochure request, contact form, magazine request and phone click in four countries was tracked as its own conversion.

Then we went further, because a cheap lead is worthless if it is a bad lead. We split the brochure request into four graded conversions, from below average through to strong, with a separate one for people who only part-filled the form, and a single combined lead goal above them for the headline number. That meant we could look at any campaign and see not only how many enquiries it produced, but what kind.

Why this was unusual

That level of segmentation was rare in 2019, and most accounts still do not run anything close to it today.

Discovery campaigns had only just launched, and we were already running a properly funded test rather than a token one. YouTube was running conversion-focused remarketing broken out country by country, at a point when most advertisers still treated video as branding nobody could measure. Display was not a dumping ground for cheap impressions: remarketing was segmented by whether someone had already converted, in-market audiences were layered with keyword targeting, and prospecting ran against audiences modelled on people who had genuinely enquired. The biggest market had moved onto cost per acquisition bidding.

Underneath all of it, the conversion tracking graded leads by quality, which very few accounts of any size do even now.

None of this was standard practice then, and most of it still isn't. We built it that way because the evidence pointed there, and the numbers below are what came back.

The results

In the twelve months to March 2020, NZ$160,012 of ad spend produced 2,014 leads at NZ$79.45 each, from 121,000 clicks and 22.4 million impressions.

Those numbers need one piece of context. These were multi-day guided trips costing several thousand dollars, frequently booked by two people travelling together. Against that, a lead at NZ$79 is cheap. It only has to convert occasionally to pay for itself many times over.

The interesting part is where those leads came from.

Everything that was not search produced half the leads for less than a third of the budget, at NZ$49.70 a lead against search at NZ$107.60. The Discovery test was the clearest case: NZ$11,774 bought 61,445 clicks at nineteen cents each and 221 leads at half the cost of a search lead, in a year when hardly anyone was running Discovery properly.

A cheaper lead is only better if it is the same lead, which is why the grading work mattered. Every campaign in that table could be judged on the kind of enquiry it produced and not just the price of it.

The cheaper channels held up when we graded them. We would not call that conclusive and we are not going to pretend otherwise, but it was consistent enough that we were comfortable moving budget toward display and discovery rather than treating them as cheap volume.

We will be straight about the number we like less. Across the three years, spend rose from NZ$86,143 to NZ$143,830 while cost per lead went from NZ$73 to NZ$93. Some of that is auction prices rising across the board. Most of it is deliberate: we were buying volume in expensive overseas markets, and lead count grew every year. Against a booking worth several thousand dollars, NZ$93 is still a number the business could comfortably live with, and if we had wanted a flattering cost per lead we would have stayed small in the cheapest market and served nobody well.

The audience we built before anyone asked for it

Early on we set up remarketing audiences on Meta, split between people who had requested a brochure and people who had been on the site but hadn't. Nobody asked us for them and there was no campaign waiting to use them. We built them because data only exists if you were collecting it, and you cannot go back later and collect the year you missed.

We tested messaging against those audiences and brought cost per lead down by 35%, then extended the campaigns from North America into Australia and the UK at no extra cost to the client.

Twelve to eighteen months later they came to us wanting to find more people like the ones already requesting brochures. We could start that week, because the audience was already built and already collecting. Set up on the day they asked, it would have been empty, and they would have waited months before it was usable.

That is why those leads were cheap. A lookalike is only as good as the seed audience behind it, and ours was seeded on people who had actually requested a brochure, not people who had merely visited a page. The split we put in at the start, converters separated from non-converters, is what made that seed possible.

This was a small programme running alongside the client's own ad hoc Meta activity, not a full channel. It is here because of what it shows: the groundwork we do early is what makes the clever thing possible later.

How it ended

The business changed hands in 2020 and the account went with the new owners. That is the one reason this story stops in September 2020, and it is worth saying plainly that it did not stop on performance.