A 40% cost reduction that came from
fixing the structure, not raising the budget
Who
CDF.immo.ch is
CDF.immo.ch is a rental platform in French-speaking Switzerland. Before January 2026 it had no paid search presence at all: tenant enquiries came from organic and direct traffic only.
comtogether designed and ran the full account architecture: the campaign segmentation strategy, conversion tracking, keyword management, and every bid strategy decision from the first day through to the portfolio migration.
A cold account
cannot self-correct
Three things were working against the account at launch, and none of them was budget. Bidding had no conversion signal to learn from. Broad matching was diluting quality scores and holding cost per click high. And pinned headlines were forcing a poor ad strength rating, which blocked the rotation Google needed to find the combinations that worked.
The question was whether the cost problem was structural or budgetary. A 33% drop in cost per acquisition in February, on an unchanged budget, answered it.
No conversion signal to learn from at launch, so the algorithm optimised for traffic rather than contacts.
Poor quality scores keeping cost per click high, on terms that were never going to produce a tenant enquiry.
Forcing a poor ad strength rating, which blocked the rotation Google needs to find the combinations that work.
Fix structure.
Build signal.
Then scale.
Three sequential fixes across the first four months built the foundation for compounding efficiency. The order mattered more than any single change: bidding cannot optimise toward a signal that does not exist yet, and a signal cannot form while ad rotation is blocked.
Unpinned the headlines in positions 2 and 3 across the lead campaign, added negative keywords, and ran maximise-clicks temporarily to gather signal. Ad strength moved from poor to average, cutting wasted impression share.
Moved from maximise-clicks to target cost per acquisition once a first conversion signal existed. The efficiency gain landed immediately, in the first full month of the new strategy.
Unified all three campaigns under one portfolio strategy with a cost-per-click ceiling. A shared conversion signal across campaigns, rather than three isolated ones, let the algorithm find efficiency across the whole account.
Three campaigns,
two markets,
one signal
Three geo-segmented search campaigns across two markets, each with its own keyword set and bid signal, so an enquiry in one town never competed against another for the same budget.
WhatsApp tracked as a conversion action from day one, alongside form submissions and calls. In this market WhatsApp turned out to carry 39% of all contacts, so leaving it untracked would have meant bidding on a partial picture.
Responsive search ads with headlines left unpinned after the launch fix, letting Google rotate freely. The best-performing variant took the highest impression share of every combination tested.
A portfolio bid strategy with a cost-per-click ceiling, introduced only once each campaign had produced enough conversions to contribute a trustworthy signal to the shared pool.
Every fix compounded.
Cost fell every month
it should have.
From no paid presence at all to a steady contact flow in seven months, with cost per acquisition falling at each structural change rather than drifting. July data is partial, to 22 July. All figures come from the client's Google Ads account for January through July 2026.
The channel split is the part worth carrying into the next account of this kind: 60% of contacts arrived through the form, 39% through WhatsApp, and 1% as calls straight from the ads.
What this
case proves
A Google Ads account launched cold will not correct itself through budget alone. It needs a deliberate sequence: fix ad quality first, build a conversion signal before enabling automated bidding, then unify campaigns under a portfolio strategy once there is enough data for a shared signal to mean something.
Tracking WhatsApp from day one was the decisive call. For a regional rental portal it is the dominant contact channel, and without it the bidding would have been optimising against 60% of the actual contacts while believing it saw all of them.