Google Ads Case Study: When Your Best Keyword Converts Zero

Google Ads case study — private hire operator, Nice–Monaco corridor. 5 May – 18 July 2026. €583.18 total spend.

---

The constraints

The client operated in one of Europe's most competitive local service markets: airport transfers between Nice Côte d'Azur and Monaco, where luxury operators bid aggressively on a narrow keyword set.

The first constraint was budget. €583 across 74 days — roughly €7.90 a day — and it did not run continuously. The client paused and restarted spend according to cash flow.

The second constraint was the destination. The client's website was outdated: built years earlier, not maintained since, and never designed to convert. I didn't own it and couldn't rebuild it. Every click the campaign bought had to survive that landing environment — and as it turned out, most of them didn't.

Why exact match, manual bidding. Smart bidding requires volume and continuity to leave the learning phase. An account that restarts every few weeks never accumulates enough conversion signal for automated bidding to work; it just relearns, repeatedly, at the advertiser's expense. So I ran a pure exact-match structure with tight manual control — fewer keywords, no broad-match bleed, every euro traceable to a specific query.

Before you read the numbers

The site's WhatsApp button — the campaign's primary conversion action — was broken from the start of the engagement, and conversion tracking was misconfigured alongside it. Neither the website nor the tracking setup was in my remit. I found both and worked with the client's webmaster to fix them, because there was no point optimizing traffic into a funnel that couldn't receive it.

The consequence is that every figure below is a floor. In one two-week window in July, the account recorded 2 conversions while the client received 8–9 inbound WhatsApp calls. The direction of the measurement error is known. The magnitude isn't.

On attribution: the client receives effectively no inbound calls on this channel when ads are paused — a pattern observed across several separate engagement periods. That on/off reversal is the strongest attribution evidence available at this budget, and better than any modelled attribution I could claim.

So read €34.30 per contact as a ceiling on cost, not a measurement of it.

Headline results

• 2,793 impressions

• 254 clicks

• 9.09% click-through rate

• €2.30 average cost per click

• 17 recorded WhatsApp contact conversions

• 6.69% conversion rate

• €34.30 cost per recorded conversion

A €2.30 average CPC in this market is low. The exact-match structure is why: no spend leaked to loosely related queries, so the average wasn't dragged upward by traffic that was never going to convert.

The diagnosis

Partway through, one keyword didn't behave.

[réserver vtc nice] — "book VTC Nice," about as high-intent as a query gets — produced the best CTR in the entire account at 14.95%, and zero conversions.

That pattern is diagnostic. High CTR means the ad is working: people see it, they want it, they click. Zero conversions from high-intent traffic means the failure sits downstream of the ad. It is almost never a keyword problem and almost never a bidding problem.

I audited the destination and found two things:

• The WhatsApp CTA button was broken — and had been from the start. The site's primary conversion action never fired. The most motivated traffic in the account was arriving, trying to convert, and hitting a dead end.

• Conversion tracking was misconfigured, so the account's own numbers were unreliable on top of that.

Neither finding was surprising on a site nobody had maintained in years. What's worth noting is that neither was visible from inside Google Ads. The account looked healthy — solid CTR, reasonable CPC, conversions arriving. The only signal that anything was wrong was a single keyword whose numbers didn't add up.

Neither fix was my responsibility. I did them anyway, with the client's webmaster, because an account optimized into a broken funnel is wasted work regardless of how good the account is.

What the data proved — and what it didn't

This is the part most case studies skip.

I cannot claim the fix produced a measurable lift. Splitting the campaign around it, CTR rose from 8.14% to 10.77% and conversion rate from 6.21% to 7.34% — but average CPC rose from €2.03 to €2.64, and cost per recorded conversion got worse, from €32.77 to €36.03. The later window also straddles the fix date rather than sitting cleanly after it, so it isn't an "after" period at all.

On top of that, 8 and 9 conversions per window are small samples, and the underlying tracking was unreliable throughout. A swing of one or two conversions is noise, not signal.

So the honest conclusion is: the test was confounded, and I won't claim a result it can't support.

What the data does support

1. Budget interruption cost real money. Average CPC rose 30% on identical keywords, identical match types, the same market, within eight weeks. That is not competitive drift over that horizon. It is consistent with repeated learning-phase resets: every restart drops the campaign back to zero accumulated signal, and it pays more for the same traffic while relearning. This sits on 254 clicks, which is enough to take seriously — and unlike conversions, click and cost data were never in question.

2. The best keyword was the most starved. [transfert aeroport nice] produced the account's cheapest conversions — around €21 each — across both measurement windows, while consistently receiving the lowest CTR and the least traffic. Consistent across two independent periods, so it isn't a fluke of one sample.

3. Roughly 18% of spend returned one conversion. Three keywords consumed €104 of €583 and produced a single recorded conversion. Reallocated toward the €21 keyword, that spend supports an estimated 4–5 conversions instead of 1 — approximately a 20% lift at identical budget.

The recommendation

• Instrument the funnel properly before optimizing it further. Until reported conversions match observed inbound volume, every efficiency figure in this account is an estimate. That is the first thing I would fix on any comparable engagement — before touching a bid.

• Run the budget continuously, even at a lower daily rate. Continuity is worth more than intensity at this scale.

• Reallocate toward [transfert aeroport nice] and raise its CTR — its problem is ad copy, not keyword quality.

• Establish the repeat-booking rate. At a €90–150 fare and typical operator margin, a first ride does not comfortably cover a €34 acquisition cost. This channel is profitable only if customers return.

Revenue validation

At least one confirmed end-to-end booking was traced to the campaign: a corporate executive booking an airport-to-Monaco transfer — a premium route in the client's highest-value segment.

One booking is one data point, and I present it as one data point. It matters because it closes the loop: it confirms the funnel worked all the way through to revenue, not just to a WhatsApp message.

What I'd tell you in an interview

The case for hiring me off this account isn't the CTR. It's that I found a broken conversion path from a pattern in the keyword data rather than by accident, fixed it cross-functionally despite it sitting outside my scope, and then told the client I couldn't prove the fix worked.

I'd also say what this account still doesn't tell us. €34.30 per contact is only defensible alongside a repeat-booking rate, and I never had visibility into one. A single transfer at that acquisition cost is marginal on most fare and margin assumptions in this market. The channel works if customers come back; it doesn't if they don't, and I'd want that number before scaling spend.

Small budgets and old websites are good teachers. There's no volume to hide behind, no room for a channel quietly wasting 18% of spend, and no option to blame the traffic when the destination is the problem. What I learned running €583 scales up: instrument first, find where the money leaks, prove it, and be honest about the limits of your own measurement.

---

Ibrahim Daoud is a performance marketing consultant based in Nice, France, working in English, French and German. Full account data, screenshots and keyword-level breakdowns available on request.