The month everybody bids on is your most expensive month.

The short answer

A private vocational college in Greece, selling two-year programmes worth €3,000 to €6,000 each. Two adjacent six-month windows, June to November, one year apart, at effectively the same budget: €27,338.77 before, €26,874.18 after — a difference of 1.7%.

Before: an average cost per click of €0.34 and, against roughly €27,000 of spend, zero recorded conversions. After: €0.09 a click and 1,033 enquiry forms at €26.02 each.

The transferable part is none of that. It is the month-by-month table further down: the old account spent 44.8% of the whole six months in September and nothing in November — and September turned out to be the most expensive month per enquiry at €29.51, against €20.94 in July. The month with the most demand is the month with the most competition, and the auction charges you for the second one.

The most interesting number in the account is a zero

The account had been running for years before we touched it. Somebody had built campaigns, somebody was paying the invoices, and the school was enrolling students. What nobody could do was connect those three facts, because in the six months from June to November of the earlier year, against €27,338.77 of spend, the account recorded no conversions whatsoever.

That is worth reading carefully, because it is not the same as a bad conversion rate. A bad conversion rate is information. This was the absence of a measuring instrument: the forms on the site were not connected to anything, so the click left Google and vanished. Every judgement about the advertising — which programme to promote, which month to push, whether any of it worked — was being made from memory and gut feeling on top of about €55,000 a year.

So the first work was not the ads. It was rebuilding the landing pages, putting a CRM behind the forms and wiring the tracking, so that the second half of the sentence — what came back — existed at all. Everything below only means something because that came first.

Same budget, six months later

Both windows are June to November, one year apart, taken from the account's own export. The earlier period is the previous team's campaigns; the later one is after the rebuild.

The budget line is the one to hold on to while reading the rest. Nothing here was bought with more money. It is the same annual spend, moved around and pointed at things that were counted.

The finding: they were buying the peak, not the season

Split both windows by month and something appears that neither the totals nor any dashboard summary would ever show. Here are the two years side by side, with what each month actually produced in the later one.

Read the first column and the last one together. The previous year put almost half of six months of budget into September, spent €58.52 in June and nothing at all in November, and September is the month where an enquiry cost the most. July's enquiries cost €20.94; September's cost €29.51 — 41% more for the same thing.

The reason is not mysterious, which is what makes it useful. Every competing school in the country reaches the same conclusion about when families decide, and they all raise their bids in the same weeks. Demand and competition peak together, and you are charged for the second one. The intent of the person searching in July is not meaningfully lower; the price of reaching them is.

The shoulder months are where the correction shows up. Between the two windows, June plus November went from €58.52 to €3,596.87 — and those two months, at the end of the later window, delivered 123 enquiries at prices at or below the period average. The peak month still got the largest single share, because that is where the volume is. It just stopped getting half of everything.

Why the clicks got cheaper — and the part that isn't merit

A 74% fall in cost per click is the sort of number that ought to make a reader suspicious, so here is the mechanism and here is the caveat.

The mechanism is ordinary and documented: Google discounts what you pay for a click on an ad it judges relevant to the query. Tighten the chain from keyword to ad text to landing page and the same position costs less. That work was done — the pages were rebuilt at the same time as the campaigns, which is exactly the chain the platform is scoring.

The caveat is that not all of the drop is merit. Impressions rose 343% on flat spend, and that shape is the signature of a large, cheap display and video layer sitting on top of a small, expensive search core — not of search advertising reaching four times as many people. Cheap impressions drag a blended cost per click down regardless of whether anything improved.

We grade the display-and-video inference as an inference. It follows from the impression and spend figures, and it is not a claim we can make from a campaign-level export we do not have. It is in the article because it is the honest reading, and because a case study that only reports the flattering half of its own data is not evidence of anything.

The half that wasn't advertising

It would be easy to present 1,033 enquiries as an advertising result. It isn't one, and the honest accounting matters more here than the number does.

Alongside the campaigns we rebuilt the landing pages the ads pointed at, put a CRM behind the forms so that an enquiry became a record with an owner rather than an email in a shared inbox, wired the tracking so the platform could learn which clicks produced enquiries, and spent time in the school's own offices training the staff who had to work the follow-ups. The advertising bought the click. The rest of it is why anybody knows.

There is a second axis here, and it is the point at which this article stops giving numbers. Once a CRM was in place, the school could remember who had asked and what they had asked about, reminders went out, and the sales process tightened. The client's own recollection is that the proportion of enquiries turning into enrolments improved substantially over that period.

The same rule removed two other numbers from this piece: a visitor-to-enquiry rate the client remembers but cannot retrieve a source for, and a set of ad-level click-through rates that are entirely plausible against the cost-per-click collapse but are not in any export we hold. Neither is published here. If they surface with a source, they will be.

Five things to do before you increase the budget

Everything above collapses into a short list, and none of it costs media money.

• Record the enquiry before you optimise the ad. An account with no conversion data is not being managed, it is being renewed. If the tracking is broken, that is the whole of the first month's work and any supplier who skips it is selling you a report you cannot check.
• Define a conversion once, in writing. One submitted interest form. Not a brochure download counted twice, not a page view, and not an enrolment — enrolments happen months later and often by phone, which makes them a poor signal to hand a bidding algorithm.
• Build the month-by-month cost-per-enquiry table. It exists in every account and it is almost never produced. It is the cheapest analysis in this article and it found the biggest thing in it.
• Fund the shoulder months. Zero in a month is a decision, and in the account here it was a €0 month that would have delivered enquiries below the period average.
• Stop reading blended averages. A single account-level CTR across search, display and video describes nothing that exists. Split the report by network before you draw any conclusion from it.

One piece of local evidence for the schools reading this from Greece, since it is easy to check. The searches a prospective student actually makes — ιδιωτικό ιεκ, ιεκ αθήνα — carry high advertiser competition in Google's own keyword data, while the searches a school owner might make to find help with marketing barely register at all. Your competitors are already in that auction. The open question is not whether to be in it; it is whether you can say what came out of it.

What managing it costs

Prices are not repeated here, because they live in one place with a visible review date and repeating them creates a second version that will eventually disagree with the first. The management fee for this system, what the build includes and where the floor sits are itemised in what it costs, and why.

The one commercial line worth stating on this page, because it decides whether any of the above is relevant to you: below roughly a thousand euros a month in media, a management fee is a bigger line than the advertising it manages. A school spending €300 a month is better served by fixing the tracking once and running the account itself than by paying anybody to manage it, and we will say so on the call rather than after the first invoice.

The system this article describes is System 09. The measurement half — the pages, the CRM, the tracking that turned an unmeasured €27,000 into 1,033 countable enquiries — is System 08, and in this project the two were bought together, which is the usual shape.

Sources, and how confident we are in each number

The figures above are not all the same kind of claim, so they are graded rather than presented as equivalent.

• Platform export — spend, impressions, average cost per click, conversions and the monthly rows, for both windows, from the advertising account's own reporting. The monthly file reconciles to the summary totals to the cent, which is why both are shown.
• Calculated — the budget shares, impressions per euro, cost per enquiry by month, and the 41% gap between July and September. All are arithmetic on the exported rows and can be re-derived from the tables on this page.
• Inference, labelled as such — that the impression increase reflects a display and video layer rather than search reach. It follows from impressions rising 343% on flat spend and from the blended CTR falling, and it is not confirmed by a campaign-level export.
• Client-reported, and therefore excluded — the improvement in close rate after the CRM went in, and a visitor-to-enquiry rate the client remembers. Both are plausible; neither has a retrievable source, so neither appears as a figure and there is no return-on-ad-spend claim built on them.
• Anonymised at the client's level — the school is described by category and country rather than by name. The numbers are unchanged.

A word on shelf life. Auction prices move, and an article quoting them gets cited for years. The two windows here are a matched pair a year apart, which is what makes the comparison fair; it does not make €0.09 a click a benchmark for anyone else's account, in any year. Treat the method as the transferable part and the numbers as one account's history. This page carries a visible review date — if you are reading it long after 7 August 2026, ask us what has changed.

MarketStarter is a HubSpot Solutions Partner and a Salesforce Partner, based in Greece, answering everywhere. The other article in this series itemises what an AI receptionist really costs, from our own invoices.