A studio service
Every ad group a test, every month a result.
Most accounts have one ad per ad group that nobody has touched since launch. We run two to three variants in every ad group, rotate them evenly, evaluate on a fixed schedule with a minimum sample, and queue the next hypothesis — through the Google Ads and Microsoft Ads APIs, so it happens in hundreds of ad groups, not the three someone remembered to check.
N° 01The method
Hypothesis, sample, decision, next.
Ad testing fails in two ways: nobody runs it, or someone runs it and calls a winner after forty clicks. We fix both. The API creates the variants and sets rotation; a scheduled job evaluates each ad group only once it has passed a minimum click and conversion threshold; decisions are written to a log you can read.
What gets tested is not random. Each round has a named angle — the provider's credentials versus availability, 'same-day' versus 'accepting new patients', a price anchor versus a guarantee, the location name in the headline or not. The angle that wins in one ad group is promoted as the next hypothesis across its siblings.
01
Variant generation
Two or three responsive search ads per ad group, sharing the pinned facts (name, location, phone) and differing on one deliberate angle. Written by us, pushed by the API.
02
Even rotation, then optimise
Rotation set to even during the test window so the platform does not pick a favourite on day two. Switched back to optimised once a winner is declared.
03
Asset-level reads
Google now reports clicks and conversions per headline and description. We read those to learn which lines carry the ad, not just which ad won.
04
Campaign-level experiments
For anything bigger than copy — a bid strategy, a match-type change, a new landing page — we use the platforms' experiment tools with a proper traffic split, via the API.
05
Minimum-sample discipline
No decision below a click and conversion floor set per ad group. Low-volume groups run longer or borrow the winner from a sibling; they are never called early.
06
A log you can read
Every test: the angle, the variants, the sample, the result, and what was queued next. In plain language, in a sheet you own.
Investment
Clear scope. Clear price.
Flat management fees, quoted against account size — never a share of spend. Media is billed by the platform to your own card, so the number you see is the whole number.
Included in management
$0extra
Scheduled ad testing is part of every Google and Microsoft management retainer.
- Variants in every ad group
- Scheduled evaluation
- Test log
- Winners promoted across siblings
- Variants per ad group
- 2–3 live at any time
- Evaluation
- monthly, minimum 100 clicks
- New variants
- 1 per ad group per round
Not included
- Testing in accounts we do not manage
- Landing-page variants (separate)
Testing sprint
from $1,500one-off
Your team runs the account; we run a 60-day structured test programme inside it.
- Angle plan for every ad group
- Variants written & pushed via API
- Two evaluation rounds
- Asset-level findings report
- Handover of the winning copy
- Duration
- 60 days, fixed
- Ad groups
- up to 40
- Evaluation rounds
- 2
- Access needed
- standard access to your account
- Revisions
- 1 round on the angle plan
Not included
- Ad spend (billed by the platform to your card)
- Bids, budgets, keywords, or negatives — copy only
- Continuing after day 60 (move to a plan)
- Landing pages or website changes (quoted separately)
- Any guaranteed lead count or cost per lead
Landing-page tests
from $2,000one-off
Copy tests only go so far — test the page the click lands on.
- Two page variants designed & built
- Campaign experiment with traffic split
- Call & form tracking per variant
- Result report and rollout
- Pages
- 1 page, 2 variants
- Duration
- until 200 conversions or 60 days
- Build
- on your CMS or standalone
- Revisions
- 1 round per variant
Not included
- Ad spend (billed by the platform to your card)
- A third variant or a second page (quoted)
- Photography or video
- Rollout across other pages
- Any guaranteed lead count or cost per lead
Ad testing is not sold separately to clients we manage — it is simply how we manage. The sprint and landing-page options exist for teams who run their own accounts and want the discipline without the retainer.
How every package works
- One package covers one business and one ad account per network. A second brand or a second account is a second package.
- Ad spend is never inside the fee. Google and Microsoft bill your card directly; you can see it in your own billing at any time.
- The caps on each card are the scope. Going past a cap (more offices, providers, ad groups, hours, pages) is priced on the card or quoted before we do it; it is never silently absorbed.
- Structure-change hours cover requests you send us: new offers, new services, budget shifts, pauses. Routine optimisation we do on our own cadence does not count against them.
- Response time is for replies and pauses. New campaign builds and page work run on the delivery times shown, not on the response time.
- Month-to-month, 30 days' written notice either way. No setup fee is refunded once the build is delivered.
- You own the account, the data, and everything we build in it. We hold manager access and remove it on the last day.
- We do not answer your phones, book appointments, manage reviews, run social media, or write blog content. If you need those, we can point you to someone who does.
- Nothing here guarantees a number of leads, patients, or a cost per lead. What we guarantee is the work on the card, on the schedule on the card.
- Anything not written on the card is not in the package. If you are unsure whether something is included, ask before you buy; the answer is the same afterwards.
Questions
The answers we give most often.
- Doesn't Google already optimise my responsive search ads?
- It picks combinations of the headlines you gave it. It cannot invent the angle you never wrote, and it will happily favour a mediocre line if that is the best it has. Testing is about feeding it genuinely different hypotheses, then reading which ones the auction rewarded.
- How long does a test take?
- It depends on volume, not the calendar. A busy ad group can reach the sample floor in two weeks; a provider-name ad group with thirty searches a month may take a quarter. That is why it runs continuously and is evaluated by a rule, not by whoever is free on Friday.
- Can you do this through the API, or is it manual?
- Variant creation, rotation settings, evaluation, pausing and cloning are all done through the Google Ads API and the Microsoft Advertising API. Google's interface-only 'ad variations' tool is not exposed to the API, so we use multiple ads per group with even rotation plus campaign experiments — which is the more controllable method anyway.
- What do you test besides headlines?
- Descriptions, sitelink sets, callouts, the landing page, and — as separate experiments — bid strategy and match type. Copy is where we start because it is cheapest and fastest to read.
Next step
Find out what your ads could be saying.
Send us your account or a few of your current ads and the market. We'll reply with the angles we'd test first and what it would take to run them.