Online advertising is the one channel where a whole budget can be spent without ever knowing what it produced. All our work consists of making that impossible.
Cost per click is the most discussed and least useful measure. A twenty-cent click that never buys costs more than a three-euro click that buys one time in ten. The only question that matters is the cost of acquiring a customer and what that customer returns over their lifetime. Until those two numbers are measured, you are not steering a campaign, you are watching it run.
So we always start with conversion tracking, before the first ad. A form sent, a call made, a quote signed, a sale: each event must be reported, with its value, and survive the disappearance of third-party trackers. That requires server-side collection and not only in the browser. It is technical, invisible, and it is what separates a steered campaign from an endured one.
Finally, the best ad does not rescue a bad landing page. Cost of acquisition is decided as much after the click as before: load time, promise kept between ad and page, short form, visible proof. We refuse to launch a campaign towards a page we judge to be losing, because that is spending your budget to prove something we already know.
Every conversion counted right, server-side, before any decision.
Events set, a value attributed to each, server-side collection so they survive blockers and the end of third-party trackers. We then check that the reported number matches the real number of enquiries received: a thirty per cent gap is common in hasty installations, and it distorts every decision that follows.
Campaigns built on intent, with negative keywords worked as hard as the rest.
Campaigns on declared intent, structured by theme rather than by product. We work negative keywords as much as positive ones: on an account we take over, a third of the budget usually goes to unrelated queries, and cleaning that up returns the third before we have even touched bidding.
Start from your real customers and let the algorithm learn from a clean signal.
On Meta and LinkedIn, the audience matters more than manual targeting. We start from your real customers, derive lookalike audiences, and let the algorithm work on a clean signal. Targeting that is too narrow starves learning: below fifty conversions a week a campaign does not learn, and things must be grouped.
Three to five variants per campaign, renewed before they wear out.
Three to five variants per campaign, refreshed before fatigue. On social formats a creative loses half its effectiveness in three to six weeks: it is not the audience tiring, it is frequency rising. So we produce continuously rather than waiting for results to collapse.
One page per promise, repeating the ad word for word.
One page per promise, never the home page. The ad's promise must reappear word for word in the page's title, otherwise the visitor thinks they have landed wrong and leaves in two seconds. Form reduced to essentials: every extra field costs enquiries, and the rest can always be asked for later.
Budget moved each week towards what performs, one decision at a time.
Budget moved each week towards what produces, clean stops on what does not. We do not touch bids daily: a bidding algorithm relearns after every change, and fidgeting costs more than it earns. One decision a week, backed by enough data not to be chance.
A campaign is judged neither on the third day nor in the twelfth month. Here are the real milestones, and what it is reasonable to expect at each.
Tracking set and checked against reality: we compare what the tool reports with what you actually receive. Until the two agree we launch nothing, because every decision taken afterwards would rest on a false figure.
Simple structure, contained budget, three creative variants. We deliberately start narrow on the most qualified queries: widening is easy, repairing the record of an advertiser who sprayed wide is less so.
We let the campaigns learn without disturbing them, correcting only clear drift. It is the hardest period to hold: the figures are ugly and the temptation to change everything is strong. Giving in restarts learning from zero.
Enough data to decide: what produces gets more, what does not stops. It is also the moment we tell you whether the channel is profitable for you. Some trades are not, and it is better to learn that in six weeks than in a year.
Creative refreshed before fatigue, gradual extension of the audiences that work, split tests on landing pages. A well-held account improves slowly and steadily; an account left alone degrades in three months.
Google Ads and Meta opened under your company, with your payment method. We are invited in as managers. The day you leave you lose neither the history, nor the audiences, nor the accumulated learning, which are often worth more than a quarter's budget.
The list of tracked events, their value, and how they are reported. It is the document that lets you verify a figure instead of believing it, and it serves every other marketing decision you take.
Ads, visuals and video delivered as open files, with the tested variants and their results. You know what worked and why, which is worth more than a folder of images with no memory.
Spend, cost per acquisition, attributed revenue, and the decision taken for the following month. One page. If the channel was not profitable that month, the first line says so.
It depends on the cost per click in your sector, which ranges from twenty cents to fifteen euros. The useful rule is not an amount but a volume: enough conversions a week for the algorithm to learn, around fifty. If your budget cannot reach that, it is better to concentrate on one campaign than to sprinkle four.
No, never, and that is deliberate. An agency paid a percentage of the budget has an interest in you spending more, which is exactly the opposite of your interest. You pay Google and Meta directly, at cost, on your account. We bill our design and steering work.
Clicks, yes. Reliable decisions, no. It takes four to six weeks for campaigns to learn and to accumulate enough conversions to compare. Judging after a week means drawing a conclusion from too small a sample, and that is how a campaign that was about to become profitable gets stopped.
Then we will tell you before spending, not after. Some markets are held by players whose monthly budget exceeds your annual revenue: entering head-on is lost before it starts. There are often angles left, a geographic niche or a sharper intent, but sometimes the honest answer is that this channel is not yours.
Leave us your website and a word about your project. A director studies it and replies to you personally within twenty-four hours: what holds your visitors back, what we would do, and whether your house can join us. It is the first door of admission, and it commits you to nothing.