Process
Measurement first. Always.
Most of the accounts we inherit are not underperforming because of bidding. They are underperforming because the thing they optimise towards is wrong, and nobody has checked in a year.
-
Measurement before media
We audit what your account is actually optimising towards. In most inherited accounts at least one conversion event fires wrong – duplicated, firing on page load, or counting a step nobody values. Until that is fixed, every bid strategy is learning from bad data.
You get: Tracking plan, event map, list of what is broken and in what order it gets fixed.
-
Rebuild the structure
Campaign structure follows the decisions you need to make, not platform defaults. Separate what you want to control separately – market, margin tier, product line, funnel stage – and merge what you do not, so the algorithm gets enough signal per unit.
You get: Rebuilt account, new naming convention, budget allocation you can read at a glance.
-
Test in a stated order
Creative first when creative is the bottleneck, audiences first when it is not, and never both at once on the same budget. Each test has a written hypothesis, a minimum sample and an end date agreed in advance, so nobody argues about the result afterwards.
You get: A test log you can see, with what we expected, what happened and what we concluded.
-
Report on payback, not activity
The monthly review opens with cost per acquisition, contribution and payback period, then explains what we changed and why. Where the funnel allows it we report on cohorts, because a good CPA on users who churn in week two is not a good CPA.
You get: Monthly written review, live dashboard, a standing 30-minute call if you want one.
Why measurement comes before media buying
Smart bidding is a learning system. It optimises towards whatever event you tell it to value, at whatever value you assign. Feed it a conversion that fires on every page load and it will faithfully find you the cheapest possible page loads. The campaign will look excellent and sell nothing.
This is the single most common thing we find, and it is not a sign of a careless team. Tracking decays: a developer ships a checkout change, a consent banner starts blocking a tag, iOS changes how a postback is attributed. Nobody notices because the dashboard still shows numbers. They are just the wrong ones.
What we do about attribution specifically
On the web that means server-side events where the platform supports them – Meta's Conversions API, Google's Enhanced Conversions – plus deduplication that actually works, and consent handling that does not silently drop a third of your conversions.
On mobile it means a SKAdNetwork conversion-value schema designed around a decision rather than copied from a template, postbacks configured so the data arrives before the budget decision needs making, and cohort reporting in AppsFlyer or Singular that runs to payback rather than to install.
How we decide what to test
In order of how much they usually move the number: the offer, the creative, the landing experience, the audience, then the bidding. Most accounts are tuned in exactly the reverse order, which is why they plateau – you cannot bid your way out of a weak offer.
Every test gets a written hypothesis, a minimum sample size and an end date before it starts. The reason is not process theatre: it stops the argument where a test that produced an inconvenient result gets re-litigated as "not run long enough".
Questions people actually ask
- Who actually works on the account?
- The person you meet on the first call. We are a small team by choice – nobody gets handed from a salesperson to a junior after signing, because there is no salesperson and there is no junior.
- How often do we talk?
- A written review every month, and a standing 30-minute call if you want one. Some clients prefer Slack and no calls at all, which is fine. What we do not do is fill an hour a week with a status update that could have been a paragraph.
- Who owns the accounts?
- You do, always. Ad accounts, analytics properties, tag containers and MMP dashboards stay in your name with us added as users. If we part ways you keep everything, including the campaign history, which is what makes the algorithms work.
- What if it is not working?
- We say so in the monthly review rather than waiting to be asked. If after three months the numbers do not support continuing, we will tell you that too. It has happened and it will happen again.
- Do you write the creative?
- We write ad copy and creative briefs, and we specify what a landing page needs. Producing video and design is quoted separately or handled by your team – we would rather be honest about that than pretend a media buyer is also a studio.
- How fast do things change?
- Tracking fixes land in week one. Structural changes take two to four weeks to rebuild and another two to exit the learning phase. Meaningful numbers arrive in month three. Anyone promising results in week two is either lucky or lying.
See the process on your own account.
The free audit is the first phase, done before you pay anything: we read the tracking and the structure, then write up what we would change and in what order.