Case study · Multi-location optometry

A new account’s first month is not a verdict.

An independent optometry group with two Cape Town branches launched a brand-new Google Ads account. The first month cost R2,535 per booking. Three months later it was costing R416. That fall is not a turnaround story — it is what a learning phase is supposed to look like.

Abstract illustration: a blurred field resolving into sharp concentric focus
Industry
Independent optometry — eye tests and eyewear
Market
Two branches in Cape Town, each with its own catchment
Starting point
A brand-new account, built from scratch, with no conversion history
Engagement
Account build and ongoing management
Headline result
Cost per booking fell from R2,535 to R416 across the first four months

The challenge

A brand-new account has a specific, unavoidable problem: it knows nothing. Google’s bidding automation works by learning which clicks turn into customers, and on day one it has no examples to learn from. It has to buy that knowledge, and the advertiser pays for it in early inefficiency.

That is the learning phase, and it is where a great many new accounts get killed — not because they were built wrong, but because somebody looked at month one, saw a terrible cost per conversion, panicked, and started changing things. Every change resets the learning. The account never gets the chance to become the thing it was going to become.

This account had two further complications. Two branches with genuinely separate catchments meant one blended campaign would have quietly starved whichever location was less competitive. And bookings largely happen by phone, which is the hardest kind of conversion to see clearly.

The approach

1
Separate the branches properly
Four campaigns, not one: a brand campaign and a generic eye-test campaign for each of the two locations. Each branch gets its own budget and its own geography, so neither can cannibalise the other and each one’s performance can actually be read.
2
Defend the brand cheaply
The brand campaigns run on a target impression-share strategy — the job there is simply to be present when somebody searches for the practice by name, and to do it at low cost. Those campaigns run at roughly 30% click-through rate, because the person searching already knows who they want.
3
Let the generic campaigns build data before asking them to be efficient
The generic eye-test campaigns were allowed to gather volume first. Asking an automated bidding strategy to hit an aggressive cost target before it has any conversion history to learn from is how you get an account that spends nothing and books nobody.
4
Hold your nerve through month one
The single most valuable thing done here was not a setting. It was resisting the urge to tear the account up in April, when the numbers looked bad and were supposed to.

The results

Cost per booking, month by monthCost per booking, month by monthWhat it cost to produce one booked eye test, as the campaigns came out of the learning phase.R2,535AprilR1,343MayR528JuneR416July**July is month-to-date (1–12 July 2026). Source: live Google Ads account.

The cost of producing one booked eye test fell by 84% across four months — not because the campaigns were rebuilt, but because they were left alone long enough to learn.

Bookings, month by monthBookings, month by monthThe same period, counted in bookings rather than rands.2April15May33June18July**July is month-to-date (1–12 July 2026) — 18 bookings in the first 12 days.

And the volume moved in the right direction at the same time, which is the part that matters. A falling cost per booking is easy to achieve by simply booking fewer people. This account booked more.

Cost per booking
R2,535 → R416
April to July, down 84%
Bookings per month
2 → 33
April to June, still climbing
Account CTR
~15%
brand campaigns run near 30%

What we are working on now

The next unlock on this account is not in Google Ads at all — it is on the website. The generic eye-test campaigns currently send people to general pages rather than to pages built to do one job: get an eye test booked. Dedicated landing pages are the clearest remaining opportunity, and until they exist the campaigns are being asked to compensate for a conversion problem that is not theirs to solve.

Phone-call tracking is the other live piece of work, because in this business a booking that happens over the phone is the most common kind of booking there is.

The takeaway

A new Google Ads account is not a machine you switch on. It is a machine that has to be taught, and the tuition is paid in the first month’s numbers.

If you judge a new account on its learning phase, you will kill accounts that were about to work. The discipline is not in the launch. It is in leaving it alone long enough to become what it is going to be — and knowing the difference between an account that is learning and an account that is genuinely broken.

Related reading

Want results like these in your account?

Every account is different. A free audit shows you exactly what is working, what is wasting budget, and where the gains are. No obligation.

Request a free audit

All figures are taken from live Google Ads accounts and were last verified on 14 July 2026. Client identity withheld by request. Results reflect specific market conditions and business context; individual outcomes vary.