Role
Paid campaigns
When do you decide to switch a campaign off? If it is after three weak days, you are deciding on noise. If it is after three weeks, you have paid for three weeks of something that does not work. Between the two sits work that can be taken off the desk.
Intro call
Four questions we would start with
When is the decision to change a campaign made?
If it is every morning, you are changing more often than the data can say anything. Most daily optimisation is a reaction to variation, not to a trend.
What happens when ten creative variants have to be prepared?
Usually three get made, because that is what there is time for. A test on three variants is not a test, it is a choice between three random ideas.
How many people touch one campaign before it goes live?
Marketing, a designer, sometimes an agency and whoever owns the landing page. With four people a launch takes a week, however simple the campaign.
What do you do when the ad panel and the analytics show different numbers?
They always do. The question is whether you have it written down which number you use for which decision - or whether you work it out afresh each time.
We take over the production and the reporting, you keep the decisions
Preparing variants, assembling the campaign structure, daily monitoring and reporting can be taken off the desk. The budget, the audience and the decision on direction stay with the specialist.
- The budget is not set automatically. - The platforms' algorithms do their work inside a campaign, but the decision how much we spend and where is a human decision with accountability for the result.
- Creative variants are material, not a publication. - They are produced faster and in greater numbers, but they go through approval - especially when they carry the brand name.
- An alert has a threshold, not a sensitivity. - We react to a change that goes beyond normal variation, not to every deviation - otherwise the alerts stop meaning anything in the second week.
Scope
What can be taken off campaign work
Twenty example tasks. Each one has a human approval point. Below is a selection, and the scope we fit to your process.
Preparing creative and variants
- Variants of ad copy from one message (approved by: the specialist)
- Fitting the graphic formats to the platforms' requirements
- Proposed headlines to test, with the reason they differ
- Checking compliance with the platform's rules before publication
- A set of landing page variants to pair with the ad
Campaign structure
- Preparing the campaign structure to the agreed scheme
- Checking the naming and the tags that make later analysis possible
- Checking the tracking tags and parameters are correct
- Detecting overlapping audiences
- The list of things to check before launch
Daily running
- An alert when a deviation goes beyond normal variation (decision on how to react: the specialist)
- Detecting creative fatigue - falling effectiveness with rising frequency
- A signal when the budget will run out before the end of the period
- Pointing out campaigns with no conversions despite spend
- Monitoring ad rejections by the platform
Reporting and conclusions
- A report with one set of definitions for every platform
- Pointing out the divergence between the panel and the analytics, with the size of the difference
- Variant results with whether the difference is significant
- An archive of what you have tested, with the result
- Material for a campaign review (conclusions: the specialist)
A day's work
Today
- The morning goes on looking through panels and checking whether something broke
- As many variants get made as you manage to prepare before launch
- The campaign launch waits for graphics and approvals
- You notice a drop in conversions once the budget is spent
- You assemble the report from four panels, each counting differently
- You remember the results of a test from six months ago, or you do not
With AI assistance
- Deviations beyond the norm are reported to you, and the rest of the noise does not take your attention
- There are as many variants as a sensible test needs
- Preparing the structure and the tags stops being manual work
- Fatigued creative reports itself before it eats the budget
- The report has one definition for every platform and shows where the difference arises
- The test archive answers whether you have already checked this
- Variants tested
a test covers as many versions as it needs, not as many as there was time for
producing variants stops limiting the scope of the test
- Time to react to a drop
you react before the budget is spent, not after
the alert has a threshold based on the campaign's history
- A consistent report
one definition of a conversion for every platform
the divergence is shown as a difference in definitions rather than in results
- Knowledge between campaigns
you do not repeat a test from a year ago
results stay in the archive, not in the memory of whoever ran it
We start with reporting, not with automating spend.
- Step 1
We agree the definitions
What counts as a conversion, at which point and from which source. It is an hour's conversation that ends years of argument - and usually reveals that two departments measure different things, both correctly.
- Step 2
One report for every platform
Assembled automatically, with the definition of each number recorded. The conclusions are written by a person and that does not change.
- Step 3
Alerts with a threshold
Only with history can you set a threshold that tells variation from change. An alert without history is random and teaches the team to ignore it.
- Step 4
Producing variants
Last, because this stage needs the brand rules written down and a working approval process.
Common questions
Questions that come up at the first meeting
What about the brand when creative is produced in bulk?
That is why variants are material for approval rather than a publication. In paid campaigns the risk is higher than elsewhere, because the material reaches people who do not know you - and that is where the first impression is built.
Are we not buying ourselves redundancies in the team this way?
No, because the bottleneck in campaigns is rarely people - it is usually the number of things you manage to test before the budget runs out. A team that does not assemble reports by hand checks more hypotheses and knows sooner which one does not work. The work moves from operating panels to thinking about what you are actually testing.
Does this fall under the AI Act?
Running campaigns is not a high-risk use. The point to watch is targeting: profiling audiences has legal limits independent of the AI Act, and sensitive categories are out of reach whatever is technically possible.
Still have a question that is not here? Book a call
Next step
How much of your budget is spent before anyone notices a campaign has stopped working?
We will show you what running campaigns looks like when the report has one definition and the alert tells a change from variation.